2025-01-31_brainforge_x_televero_-_ai_transformatio_e6e022ff

Meeting Title: Brainforge x Televero - AI Transformation Initiatives Date: 2025-01-31 Meeting participants: Uttam Kumaran, Amyadams, Ray, Brian Tucker, Connor Fenn, Scott_Harmon

---

WEBVTT

1 00:00:59.140 ⇒ 00:01:00.380 Connor Fenn: Hey! Good morning!

2 00:01:01.900 ⇒ 00:01:03.150 Scott_Harmon: Hey? Are you, Connor?

3 00:01:03.790 ⇒ 00:01:05.117 Connor Fenn: I am. Hey, Scott!

4 00:01:05.459 ⇒ 00:01:07.419 Connor Fenn: Nice to meet you.

5 00:01:07.420 ⇒ 00:01:08.170 Uttam Kumaran: Good morning!

6 00:01:08.450 ⇒ 00:01:10.050 Scott_Harmon: Yeah, I’ve heard a lot about you.

7 00:01:10.800 ⇒ 00:01:14.376 Connor Fenn: Yeah, I I know we got a lunch set up soon. I’m looking forward to it.

8 00:01:14.900 ⇒ 00:01:18.830 Scott_Harmon: Yeah, yeah, do we that we’re trying to? Right?

9 00:01:18.830 ⇒ 00:01:21.669 Connor Fenn: We talked, we did talk.

10 00:01:22.040 ⇒ 00:01:24.540 Scott_Harmon: So it’s still kind of it’s still kind of in the.

11 00:01:25.050 ⇒ 00:01:27.020 Uttam Kumaran: Planning phase. Yes.

12 00:01:27.300 ⇒ 00:01:30.840 Scott_Harmon: Okay? And yeah, you got some real estate background, I guess.

13 00:01:30.840 ⇒ 00:01:32.050 Connor Fenn: I do. Yep.

14 00:01:32.050 ⇒ 00:01:37.120 Scott_Harmon: Sweet, sweet. Yeah, no, that’s my favorite topic. So, looking forward.

15 00:01:37.990 ⇒ 00:01:40.529 AmyAdams: Am I echoing, or is it? Can you all hear me? Okay, with that.

16 00:01:40.530 ⇒ 00:01:41.629 Uttam Kumaran: That was great.

17 00:01:41.630 ⇒ 00:01:42.070 Connor Fenn: We’re good.

18 00:01:42.070 ⇒ 00:01:43.750 Scott_Harmon: Doing. Great Hi, Amy!

19 00:01:44.080 ⇒ 00:01:52.050 AmyAdams: Thank goodness, I always come prepared with my headphones, and I never know when the echo is, gonna be there. Okay. Hello! Good morning to y’all.

20 00:01:52.050 ⇒ 00:01:52.750 Uttam Kumaran: Good morning!

21 00:01:53.060 ⇒ 00:01:54.689 Scott_Harmon: How are things up in Dallas.

22 00:01:55.400 ⇒ 00:02:01.269 AmyAdams: Good Dallas is doing good. Oh, we got like 2 to 4 inches of rain in the middle of the night.

23 00:02:01.970 ⇒ 00:02:06.269 AmyAdams: It wasn’t last night 2 nights ago, so I’m not complaining. I like the rain.

24 00:02:06.270 ⇒ 00:02:08.130 Scott_Harmon: Oh, good, good, good.

25 00:02:08.139 ⇒ 00:02:12.489 Ray: Hey? Hey? Good morning rays on here. I’m just not on video right now.

26 00:02:12.490 ⇒ 00:02:13.639 Scott_Harmon: Oh, Hi, Ryan, it’s Scott!

27 00:02:13.640 ⇒ 00:02:14.010 Uttam Kumaran: I agree.

28 00:02:14.450 ⇒ 00:02:16.189 Scott_Harmon: Good day. Talk to you again.

29 00:02:16.190 ⇒ 00:02:17.214 Ray: Good to see you guys.

30 00:02:20.260 ⇒ 00:02:20.610 Brian Tucker: Got it.

31 00:02:20.610 ⇒ 00:02:21.260 AmyAdams: Hey.

32 00:02:21.970 ⇒ 00:02:31.609 AmyAdams: hi! Brian, Brian is on! Brian Ray are from my team. I’m so glad that they’re on so that we can better get in sync and aligned.

33 00:02:32.112 ⇒ 00:02:54.410 AmyAdams: okay. So we left off last meeting. We had kind of talked about televiro, and then the last 10 min we were kind of jumping in to the initiatives, and we got to look at the no auditing tool for like a couple seconds, and but I also wanted to dive a little bit deeper into our other initiative, the converting patients, and maybe what y’all could help, what kind of pilot we could come up with for that.

34 00:02:54.850 ⇒ 00:03:03.210 AmyAdams: But, Brian, I know that you’re on. We had some preliminary questions about.

35 00:03:04.010 ⇒ 00:03:23.999 AmyAdams: I guess, like the work and where the work was gonna be happening like, would you all be coming into our server and just making sure that everything like we’re secure cause we have taken a lot of steps so far to protect our data. And we just wanna make sure that nothing would happen. But yeah, Brian, would you like to expand or anything on that.

36 00:03:26.531 ⇒ 00:03:43.390 Brian Tucker: I don’t know that we necessarily need to get into that conversation. Right now. I think, really, what we wanna focus on, and Amy and I talked about this a little bit last night is we talked about a proof of concept, potentially, with

37 00:03:43.520 ⇒ 00:03:51.179 Brian Tucker: how we could automate reviewing notes and ensuring that we have some compliance.

38 00:03:52.080 ⇒ 00:04:06.410 Brian Tucker: And I think the other thing that we want to talk about. So one of the business problems that we’re really wanting to challenge, really wanting to solve right now is around our conversion from referrals to new patients.

39 00:04:06.410 ⇒ 00:04:06.780 AmyAdams: Good.

40 00:04:07.207 ⇒ 00:04:10.622 Brian Tucker: To to their 1st appointment, and then further

41 00:04:11.270 ⇒ 00:04:19.930 Brian Tucker: getting them back in for follow up appointments right now there is a tremendous amount of manual work that’s being done there.

42 00:04:20.628 ⇒ 00:04:30.019 Brian Tucker: And it’s done by onshore resources and offshore resources. And that includes calling patients, which, of course, has about a 95%,

43 00:04:31.090 ⇒ 00:04:33.019 Brian Tucker: no answer rate.

44 00:04:33.310 ⇒ 00:04:41.340 Brian Tucker: And so we’re really, just openly looking at our our process for getting patients into their 1st appointment

45 00:04:41.590 ⇒ 00:05:08.910 Brian Tucker: and getting them into their follow, ups up through the 4th appointment and figuring out ways that we can use potentially artificial intelligence to help us with that. Even an an artificial intelligence agent, and also, combine that with some automation that we already have in place. And if we need to extend that we will. So that’s really what we’re interested in. And it, you know, we’re we’re we’re open to. You guys, help us think through some of these things.

46 00:05:08.910 ⇒ 00:05:12.789 Scott_Harmon: So, Brian, if I could just click, click on that, as they say.

47 00:05:12.790 ⇒ 00:05:13.420 Brian Tucker: Yeah.

48 00:05:13.760 ⇒ 00:05:19.230 Scott_Harmon: For just a second. You know Amy did a great job in our introductory call of kind of giving us an overview of

49 00:05:19.540 ⇒ 00:05:26.830 Scott_Harmon: your processes and kind of the workflows you use, both automated and human to support those.

50 00:05:28.340 ⇒ 00:05:33.109 Scott_Harmon: Could you or anyone just help us with the current conversion rate?

51 00:05:33.540 ⇒ 00:05:39.630 Scott_Harmon: I think I had it somewhere around 25%. And there’s there’s actually a few stages. If you break down

52 00:05:40.090 ⇒ 00:05:50.549 Scott_Harmon: conversion. There were, I think there were actually a couple of stages that maybe you look at from a metrics perspective. Can you just kind of expand a little bit? Give us a baseline of what is that today.

53 00:05:51.040 ⇒ 00:05:57.399 Scott_Harmon: And and if there’s a couple of stages where you see the fall off, because I know you guys.

54 00:05:58.310 ⇒ 00:05:59.099 Brian Tucker: Yeah, no. Absolutely.

55 00:05:59.100 ⇒ 00:06:02.400 Scott_Harmon: Very data driven. Very process, oriented.

56 00:06:02.930 ⇒ 00:06:03.430 Brian Tucker: Yeah.

57 00:06:03.430 ⇒ 00:06:10.809 Scott_Harmon: And if you could just unpack that a little for us and give us where you’re at today, that would help inform how we think about it. I think.

58 00:06:10.810 ⇒ 00:06:26.100 Brian Tucker: Right. So I think Amy has the numbers that she can share with us. But let me just kind of walk you through it super high level. So there’s a step that goes from when the patient is sent to us. Okay, and that patient can be sent to us with a referral

59 00:06:26.320 ⇒ 00:06:27.800 Brian Tucker: or without a referral

60 00:06:28.408 ⇒ 00:06:52.359 Brian Tucker: and when they’re sent to us we have an automated outreach sequence that goes through that, that the purpose of this process, this outreach process is to get them to complete their new patient information, which we call our registration process. Internally, we don’t say that externally, because that word registration is heavy. So we have some number of patients that are sent to us, whether they we get a referral from their

61 00:06:52.580 ⇒ 00:06:56.540 Brian Tucker: care provider, or they’re just sent here by their care provider.

62 00:06:57.250 ⇒ 00:07:02.539 Brian Tucker: And we want to get them to register. So that’s 1 percentage that we want to increase. Okay.

63 00:07:02.640 ⇒ 00:07:30.159 Brian Tucker: then we have those patients that register, and there’s a drop off between registering and getting them into their 1st appointment. That’s the second step. Okay? Then the 3rd step is is that once we schedule them for their 1st appointment, actually getting them to show up for their 1st appointment. Right? So in these steps we have a bunch of patients that don’t register. We have a bunch of patients that register that go into a black hole

64 00:07:30.930 ⇒ 00:07:43.670 Brian Tucker: and then we have the patients that actually get scheduled that don’t show up for their 1st appointment, and then finally, from there we want to get them into at minimum 3 consecutive appointments. We want to get them into more.

65 00:07:43.670 ⇒ 00:08:03.900 Brian Tucker: but getting them to the 4th appointment is really a key metric for a lot of healthcare providers, because once you get them to their 4th appointment. You’ve got them pretty established with you, and they’re going to continue to come back. Okay? So those are really the 4 key stages that we want to look at going forward, and I think Amy has most of the

66 00:08:04.040 ⇒ 00:08:06.539 Brian Tucker: conversion rate around that, don’t you, Amy?

67 00:08:09.360 ⇒ 00:08:11.390 Scott_Harmon: Oops. Sorry, Amy, I’m not hearing you.

68 00:08:11.390 ⇒ 00:08:12.250 Brian Tucker: Yeah.

69 00:08:12.250 ⇒ 00:08:13.829 Uttam Kumaran: Amy, you’re on mute. Yeah.

70 00:08:18.050 ⇒ 00:08:28.169 Scott_Harmon: Nope, not not yet not getting Nope nothing yet.

71 00:08:29.010 ⇒ 00:08:36.619 Ray: Hey? I’m just gonna jump in here while Amy comes back on board. So the conversion rate through the 1st 2 steps is about 40%.

72 00:08:36.909 ⇒ 00:08:38.750 Scott_Harmon: That’s what I have right now, right.

73 00:08:38.750 ⇒ 00:08:46.809 Ray: We get 50 ish percentage points to the registration, and then we have a drop off to registration. And then we have another.

74 00:08:47.170 ⇒ 00:08:53.729 Ray: You know 5, 5 ish, 3 to 5 ish percent that may not show up, even if they’re scheduled.

75 00:08:53.860 ⇒ 00:08:58.359 Ray: So that’s roughly where we’re at. It’s better than average. But

76 00:08:58.840 ⇒ 00:09:02.920 Ray: I mean guys getting the referral, I have 50% more referrals.

77 00:09:02.920 ⇒ 00:09:03.950 AmyAdams: More referrals.

78 00:09:05.570 ⇒ 00:09:06.350 AmyAdams: Hello!

79 00:09:08.920 ⇒ 00:09:09.970 AmyAdams: Don’t hear me now.

80 00:09:09.970 ⇒ 00:09:10.490 Scott_Harmon: Yeah.

81 00:09:14.720 ⇒ 00:09:15.669 Uttam Kumaran: You’re back on the head.

82 00:09:15.820 ⇒ 00:09:17.509 Scott_Harmon: Oh, God! You’re on mute again.

83 00:09:18.520 ⇒ 00:09:19.870 AmyAdams: I’m sorry I’m.

84 00:09:19.870 ⇒ 00:09:20.339 Scott_Harmon: Sure you are.

85 00:09:20.340 ⇒ 00:09:21.040 AmyAdams: Hear me now.

86 00:09:21.710 ⇒ 00:09:22.510 Brian Tucker: I can hear you.

87 00:09:23.070 ⇒ 00:09:33.359 AmyAdams: Ray was doing a great job. That’s exactly from like end to end referral to getting them into their 1st appointment. We we convert like 35, 40.

88 00:09:34.260 ⇒ 00:09:35.650 Scott_Harmon: Got it, and.

89 00:09:35.650 ⇒ 00:09:41.739 AmyAdams: And but referral to registration is like we, we drop 50% right there.

90 00:09:41.740 ⇒ 00:09:48.449 Scott_Harmon: Gotcha. So and again, this is perfect. And I think you know, we could probably exchange more detailed.

91 00:09:48.840 ⇒ 00:09:53.719 Scott_Harmon: You know, data as part of a of a pilot. I don’t want to go too deep here, but

92 00:09:55.980 ⇒ 00:09:59.979 Scott_Harmon: just one comment. It sounds like, you know, you have a team of

93 00:10:01.670 ⇒ 00:10:07.549 Scott_Harmon: of people. I think you mentioned onshore and offshore. Amy, last time said there were several of them in Texas, and

94 00:10:07.830 ⇒ 00:10:10.429 Scott_Harmon: I’m guessing that they get assigned

95 00:10:10.690 ⇒ 00:10:21.430 Scott_Harmon: from the point of the referral. Someone has these lists of people, and they’re they’re using both systems, in other words, email and reminders and stuff as well as outbound calling.

96 00:10:21.690 ⇒ 00:10:25.942 Scott_Harmon: So people have sort of a group of people they’re trying to convert.

97 00:10:26.730 ⇒ 00:10:31.829 Brian Tucker: Okay? So so you’re yeah, that’s right. So the so the step from

98 00:10:32.070 ⇒ 00:10:38.840 Brian Tucker: receiving the patient information whether that is with or without a referral to

99 00:10:38.970 ⇒ 00:10:44.459 Brian Tucker: getting the patient through the registration process is 100% automated today.

100 00:10:44.680 ⇒ 00:10:53.930 Brian Tucker: we don’t really. I mean, we do a little bit of outbound calling to try to get those people that we’re doing outreach to. But it is all everything.

101 00:10:54.270 ⇒ 00:10:54.740 Scott_Harmon: Oh!

102 00:10:54.740 ⇒ 00:10:58.440 Brian Tucker: Them. Putting it into them. Getting registered is automated now.

103 00:10:58.440 ⇒ 00:10:58.820 Scott_Harmon: Wow!

104 00:10:58.820 ⇒ 00:11:03.910 Brian Tucker: We are. We have in the past done calling out to these patients. But I will tell you.

105 00:11:04.210 ⇒ 00:11:22.039 Brian Tucker: 95% of the patients don’t answer the phone right? And that is just the way the world is today. Right? So that is that. And so if if we put you know we. We have some people that are calling particular clients right? That are high visibility, that we want to try to get better conversion. But there’s no way.

106 00:11:22.380 ⇒ 00:11:22.820 Scott_Harmon: Got it.

107 00:11:22.820 ⇒ 00:11:28.499 Brian Tucker: That we want to invest the time and energy to call people when there’s a 95%, no answer rate. And.

108 00:11:28.630 ⇒ 00:11:31.739 Scott_Harmon: I’m sure there’s something that we can do. There.

109 00:11:31.890 ⇒ 00:11:54.889 Brian Tucker: Not only in the automation because it’s like a, it’s like a 5 touch automated texting process. We don’t get their emails as part of the referral or part of the outreach. So all we get is their cell phone number. So I am sure, both in that outreach process, whether it be in in some way of contacting the patient. Whether that be phone or texting, I am sure there are things that we can tighten up in there. There’s no doubt about it.

110 00:11:55.160 ⇒ 00:11:56.600 Scott_Harmon: Gotcha gotcha.

111 00:11:56.600 ⇒ 00:12:03.020 Brian Tucker: And I’ll tell you one other thing that’s super super super. Important to know is when we do get a hold of a patient

112 00:12:03.190 ⇒ 00:12:10.630 Brian Tucker: in whatever way that we get a hold of them. The answer that they give us 99% of the time is, I’m not ready.

113 00:12:10.860 ⇒ 00:12:31.970 Brian Tucker: And that phrase, I’m not ready. Means a lot of things right I don’t have the money right now. I don’t have the insurance coverage right now, right it’s too expensive. I’m nervous. I’m scared. There’s a stigma with behavioral health, right? There’s a laundry list of reasons why they’re I’m not ready and we can probably help you figure that out. But I think that is really

114 00:12:32.410 ⇒ 00:12:34.019 Brian Tucker: part of the answer which is.

115 00:12:34.020 ⇒ 00:12:34.390 AmyAdams: So.

116 00:12:34.390 ⇒ 00:12:35.239 Brian Tucker: How do we get them ready.

117 00:12:36.980 ⇒ 00:12:38.420 Scott_Harmon: Gotcha. So

118 00:12:40.020 ⇒ 00:12:46.600 Scott_Harmon: you. I recall that you’ve got a contact management system that you use to kind of, you know, automate.

119 00:12:46.720 ⇒ 00:12:50.980 Scott_Harmon: that 1st step the registration process, or there’s some system that kind of

120 00:12:51.640 ⇒ 00:12:53.479 Scott_Harmon: sends the text and the reminders.

121 00:12:53.768 ⇒ 00:12:58.680 Brian Tucker: Yeah, we have. It’s all automation that we build ourselves homegrown. It goes out through twilio and.

122 00:12:58.680 ⇒ 00:12:59.430 Scott_Harmon: Gotcha. Okay.

123 00:12:59.430 ⇒ 00:13:05.999 Brian Tucker: Yeah, I think I think what’s important to know from my perspective is is that the percentage of people that come and they register

124 00:13:06.740 ⇒ 00:13:23.690 Brian Tucker: for whatever reason we get up to the plate and we swing the bat and we get them to register. It’s the other folks that don’t register that we need to figure out why it is that they’re not registering and why they’re not. Whatever is, the reasons are that they’re not ready and try to warm them up to get them all ready.

125 00:13:23.690 ⇒ 00:13:34.989 Scott_Harmon: So just as a thought exercise for maybe a couple of minutes. If I could get you to react to a thought exercise which you’ve probably already had. This is nothing, you know. This is standard, you know. AI kind of thinking, if

126 00:13:35.808 ⇒ 00:13:44.269 Scott_Harmon: if you had an AI expert, I’m not you, these, these people that you’re engaging in the registration process. They could basically be with them

127 00:13:44.710 ⇒ 00:13:51.410 Scott_Harmon: and hold their hand. Answer any question they had like, what? What in your mind would be the ideal helper

128 00:13:53.680 ⇒ 00:13:57.330 Scott_Harmon: and and kind of how would that helper behave.

129 00:13:58.450 ⇒ 00:14:01.120 Brian Tucker: That’s a great. So that’s a great question.

130 00:14:01.150 ⇒ 00:14:24.368 Brian Tucker: So one of the things that we actually did was is we took all of the patient requests that we get from our patients. Okay, and it’s something like 20,000 requests for things like scheduling insurance questions, billing questions, etc, etc. We took all those, and we actually de identified all the data, ran it through AI and generated some FAQ questions. Okay, now,

131 00:14:24.710 ⇒ 00:14:37.500 Brian Tucker: realize that this is done on a very basic primary level. We didn’t do any agent building or anything like that. We just basically took it and said, Show me what the most and and we put those on our website. Okay, now.

132 00:14:37.510 ⇒ 00:14:38.330 Brian Tucker: all right.

133 00:14:38.470 ⇒ 00:15:05.490 Brian Tucker: could that be exploited further and create an agent that it could essentially help these patients get all their questions answered without ever have talking to a human, absolutely right. I think a lot of it is information. You know, people are going to do their research online, and they want to do it in an objective influence, uninfluenced way and get the answers to those questions. And I can tell you. We haven’t even scratched the service on what we could do to give them the information that they need.

134 00:15:06.310 ⇒ 00:15:15.580 Scott_Harmon: Okay. So so just to play that back you’ve done some experimenting with. I’m just gonna call it an FAQ agent, which is fantastic place to start.

135 00:15:19.580 ⇒ 00:15:25.565 Scott_Harmon: could you react to the idea of a more what I’m going to call a concierge based registration process?

136 00:15:26.670 ⇒ 00:15:42.430 Scott_Harmon: you know, you go to some nice retail outlet with a concierge that just meets you at the door and walks you in, and instead of making you fill out the forms and answer all the questions it says, Hey, I’m going to do this for you. Don’t worry about it. I’m going to take care of all this like a concierge shopper or a concierge.

137 00:15:42.580 ⇒ 00:15:46.079 Scott_Harmon: And so, instead of having me have to fill out web forms, and.

138 00:15:46.570 ⇒ 00:15:50.450 Scott_Harmon: you know, do something that might, I just might find

139 00:15:51.150 ⇒ 00:15:55.490 Scott_Harmon: inconvenient or problematic. I just have a helper. Do everything for me.

140 00:15:56.160 ⇒ 00:16:00.189 Scott_Harmon: Can you react to that? That notion of a more concierge based approach.

141 00:16:01.720 ⇒ 00:16:18.189 Ray: So I’ll let me just jump in is, listen. I I think people might need a little bit more help, but that would suggest that we have a high abandonment rate. That means they went to the registration, and they had difficulty, and they would could benefit from a concierge.

142 00:16:19.333 ⇒ 00:16:23.410 Ray: The fact that it’s concierge or not is not gonna get someone to.

143 00:16:24.080 ⇒ 00:16:28.229 Ray: Yeah, I don’t. We’re lose. We’re not getting people to the link.

144 00:16:28.720 ⇒ 00:16:32.330 Ray: So from the time they pick up the after visit, Summary

145 00:16:32.520 ⇒ 00:16:37.109 Ray: and the doctor says, Hey, go to Televero. Here’s their information.

146 00:16:38.080 ⇒ 00:16:45.480 Ray: They just don’t take the action to come to the link. Now, certainly Scott does fall off like people abandon the registration.

147 00:16:45.590 ⇒ 00:16:53.100 Ray: then we get them to register. Then we have fall off to the schedule. Then we get scheduled to 1st appointment fall off. But the biggest one

148 00:16:53.430 ⇒ 00:16:58.800 Ray: is in the 50% of the people who just don’t take the doctor’s instructions.

149 00:16:58.960 ⇒ 00:17:02.719 Scott_Harmon: So that so they’re not even engaging or starting the process. So.

150 00:17:02.720 ⇒ 00:17:03.260 Ray: That.

151 00:17:03.400 ⇒ 00:17:10.090 Scott_Harmon: If I just played that back, common sense would tell you that instead of sending them a link, or you could always a B test this, but

152 00:17:10.950 ⇒ 00:17:13.910 Scott_Harmon: you know an agent reaches out to chat with them.

153 00:17:14.050 ⇒ 00:17:22.159 Ray: Well, we have people. So Brian, a little update, someone is trying to reach out and call. So we we’re reaching out via text.

154 00:17:22.310 ⇒ 00:17:26.659 Ray: And we’re reaching out via phone. So Sarah’s team is reaching out. But with a.

155 00:17:26.810 ⇒ 00:17:29.199 Ray: you know, 8% connection rate.

156 00:17:29.200 ⇒ 00:17:29.940 Brian Tucker: That’s right.

157 00:17:29.940 ⇒ 00:17:30.990 Ray: So.

158 00:17:31.687 ⇒ 00:17:34.120 Ray: You know, we need to look at.

159 00:17:35.380 ⇒ 00:17:59.699 Ray: We need to look at all of those and see what we can optimize. 1st of all, these reach outs are not automated. They’re not even using the latest Sdr processes where there’s an automatic dialer. And there’s A AI voice message being left behind. And then, you know, cause they’re only gonna connect, you know, with 200 phone calls. If they connect with 10 people that would. That’s gonna be it. So we’re not using that there.

160 00:17:59.750 ⇒ 00:18:06.740 Ray: The other is is we really haven’t analyzed the data to see when we should be sending the text messages ideally

161 00:18:06.870 ⇒ 00:18:09.150 Ray: by age, group by gender.

162 00:18:09.410 ⇒ 00:18:15.000 Ray: I mean, we don’t even know. I mean, maybe people only respond on Friday evenings. So that’s part of it.

163 00:18:15.360 ⇒ 00:18:18.340 Ray: And then to Brian’s, you know. Other point is.

164 00:18:18.580 ⇒ 00:18:24.860 Ray: you know, how do we engage them? Differently so we can get them in

165 00:18:25.542 ⇒ 00:18:37.890 Ray: to the practice as well? Now listen ideally. We would have a person standing at the checkout desk at every one of the clinics that refer to us, you know, wrap their arms around them and say, Give me your appointment. We we can’t do that. But

166 00:18:38.020 ⇒ 00:18:42.839 Ray: I mean, if there was something something that simulated that

167 00:18:44.950 ⇒ 00:18:47.439 Scott_Harmon: Of course. Yeah, we could. We could also

168 00:18:47.440 ⇒ 00:18:53.180 Scott_Harmon: went. I, just, I’m so just to summarize the last 8 min. Here, we’re moving further.

169 00:18:53.780 ⇒ 00:18:59.169 Scott_Harmon: Let’s call it up the funnel to the very top of the funnel, which is where your attention has drawn us.

170 00:18:59.470 ⇒ 00:19:00.720 Scott_Harmon: saying, Look.

171 00:19:00.910 ⇒ 00:19:06.279 Scott_Harmon: I’m going to paraphrase for you, Ray, the most problematic numbers. We don’t get them to engage at all.

172 00:19:06.640 ⇒ 00:19:12.189 Scott_Harmon: and that that those do engage. Yes, there’s some fallout from the registration process. But

173 00:19:12.710 ⇒ 00:19:17.129 Scott_Harmon: maybe the most acute problem is that people aren’t even engaging at all.

174 00:19:17.280 ⇒ 00:19:17.670 Ray: The.

175 00:19:17.670 ⇒ 00:19:19.400 Scott_Harmon: And yeah, that’s.

176 00:19:19.960 ⇒ 00:19:21.430 Ray: That’s the biggest number.

177 00:19:21.430 ⇒ 00:19:28.199 Uttam Kumaran: So, Scott, if I could, if I could just go through a couple of things. So I think there’s a couple of like key problems that I think

178 00:19:28.470 ⇒ 00:19:34.320 Uttam Kumaran: you know, we deal with both for different clients. But I do think there’s a lot of opportunity. So one thing is.

179 00:19:34.550 ⇒ 00:19:55.640 Uttam Kumaran: there’s 2 2 ways to look at sort of AI phone calling, one is like, you reduce the amount of people. Maybe you need to make phone calling. These phone calls can happen at different times. You can also, you know, sort of reduce the cost of having a person on right. This doesn’t seem more about like we can get more people to make calls. This is more about, can we improve the quality of those calls?

180 00:19:55.640 ⇒ 00:20:08.020 Uttam Kumaran: So part of this is like, okay, do do we understand what the current objections are? Do we understand which objections typically lead to? What conversion rate? So there’s part, partly a data exercise. Secondary is

181 00:20:08.090 ⇒ 00:20:29.309 Uttam Kumaran: getting all that feedback is one thing. But then getting your team, who’s calling to implement that and then to measure is a is another exercise. Right? So the nice thing about using AI, and there’s a couple of tools that we have experience. The folks on our team doing in sort of voice. AI is that you can immediately implement those changes and a B test phone calling.

182 00:20:29.651 ⇒ 00:20:47.849 Uttam Kumaran: right? And this kind of goes into my next point, which is, you guys definitely need to do a B testing on the site as well. A lot of our clients, we help with like event based data modeling. So like amplitude mixed panel stuff where you can track events on the website. But you guys should definitely rule out a lot of your hypothesis on

183 00:20:48.174 ⇒ 00:21:04.409 Uttam Kumaran: where are people coming out of the funnel? What are they clicking on? That leads to? Higher conversion rates? And that’s something that you guys should definitely be doing a B testing not only on the website itself and the registration process but also in the phone calls.

184 00:21:04.410 ⇒ 00:21:16.059 Uttam Kumaran: And then lovely thing about these AI Voice agents is, not only will they go through the exact script and sort of handle the objections necessarily. Similarly, like you guys automated

185 00:21:16.160 ⇒ 00:21:37.103 Uttam Kumaran: with with twilio, sending the message out at a particular time. Same thing with calls. As soon as they get a referral, a call can get made or at a particular time. And so I feel like, what I’m hearing is that there just needs to be sort of a concise testing plan to test all these variables, and sort of isolate what works and and what lift each can get.

186 00:21:37.630 ⇒ 00:21:40.761 Uttam Kumaran: which is great, like I that is something that

187 00:21:41.320 ⇒ 00:21:46.760 Uttam Kumaran: you know, you guys have the the infrastructure already for you know, that’s that’s what I’m hearing.

188 00:21:49.840 ⇒ 00:22:01.749 Brian Tucker: Yeah. And and of course, you know, we have the laundry list of patients who never registered right? So I’ll tell you this, we even take that list every 6 months.

189 00:22:02.666 ⇒ 00:22:18.169 Brian Tucker: or, you know, 3 times a year, and we go back, you know, 4 months, or whatever. And we do another outreach process, and you know we’ll get 50 patients that for whatever reason they’re ready now, right and and and they’ll they’ll take the step and register

190 00:22:18.720 ⇒ 00:22:20.050 Brian Tucker: right? So

191 00:22:20.680 ⇒ 00:22:40.760 Brian Tucker: you know, I I guess it’s to me it’s a lot like marketing, you know. You’re gonna meet with a client. You’re gonna step up to the home plate. You’re gonna swing them back. You’re gonna get a home run right then there’s other ones that you have to warm them up and get them around the basis to get them to score right? Whatever percentage it is, we’re getting half of those people, or whatever the number is, we’re getting them to

192 00:22:40.810 ⇒ 00:22:51.030 Brian Tucker: register. So we’re hitting a home run to get into registration. Now, what happens after that is a different conversation. But how do we warm those other people up and get them around the basis, some number of them.

193 00:22:51.170 ⇒ 00:22:55.259 Brian Tucker: because that’s a differentiator. I mean, if we could do 10%, there.

194 00:22:55.450 ⇒ 00:22:55.870 Uttam Kumaran: Yeah.

195 00:22:55.870 ⇒ 00:23:06.160 Brian Tucker: That would be huge. Right? We’re we’re above the industry standard on getting them registered the conversion rate. It’s very low in in behavioral mental health because of the stigma, but.

196 00:23:06.160 ⇒ 00:23:06.720 Uttam Kumaran: Yeah.

197 00:23:06.930 ⇒ 00:23:08.660 Brian Tucker: I don’t think that we’re doing.

198 00:23:09.180 ⇒ 00:23:19.470 Brian Tucker: We haven’t even. We’re we’re in, you know. We’re crawling on our hands and knees, you know, before even preschool on what we could potentially leverage in technology to.

199 00:23:19.830 ⇒ 00:23:23.079 Brian Tucker: you know, drive that drive that rate up in some way.

200 00:23:24.560 ⇒ 00:23:46.659 AmyAdams: So right? So Brian, like mentioned, like, we have an automated system right now. As soon as we get the fact, the referral fax. We just have like automated system, like sending messages every 2 days. Oh, we received your referral. Please. Register at the link below like, and then, if they don’t register at the link below 2 days later they say, Hey, where we

201 00:23:46.730 ⇒ 00:23:57.539 AmyAdams: flow, but maybe that we can use a watch or leverage AI to help us with that, making it a more personalized experience. Like, I think we’re, we’re okay with, like.

202 00:23:58.090 ⇒ 00:24:10.700 AmyAdams: changing what we currently have in place in helping AI with these patient reach outs, because maybe that’s where in itself we’re not being concierge enough when we do our 1st initial reach out.

203 00:24:10.700 ⇒ 00:24:12.670 Uttam Kumaran: Yeah. So there’s there’s

204 00:24:12.870 ⇒ 00:24:31.800 Uttam Kumaran: yeah, there’s 2 components. There’s 1 like, it’s the quality of the messaging. Whether that’s voice, whether that’s a text. Definitely. I think there’s an opportunity for AI. But I think even easier than AI is just to isolate variables and run testing right things. You, you explained a couple of things. One is like time of day day of week

205 00:24:32.270 ⇒ 00:24:35.979 Uttam Kumaran: frequency. Right? Those are all great variables to test. And

206 00:24:36.130 ⇒ 00:24:51.730 Uttam Kumaran: you know, we run a lot of experiments for clients on the data side very similar to like. It’s very similar to this. And and what Ray, what you mentioned, it’s just like marketing. Right? How many typically in marketing, they say, like Coke, has to get 9 impressions into your brain for you to recognize a brand.

207 00:24:51.730 ⇒ 00:24:52.330 Brian Tucker: Right.

208 00:24:52.570 ⇒ 00:24:55.120 Uttam Kumaran: It’s a you know. It’s a very similar concept where

209 00:24:55.230 ⇒ 00:25:12.310 Uttam Kumaran: you need to understand. Is it 3 text? Is it a call? Is it a follow up, and then it works. And again, you just have to run sort of multivaria testing. The nice thing is, I think there’s opportunity for AI to improve the quality of the engagements. But ultimately the AI isn’t gonna help structure the test.

210 00:25:12.310 ⇒ 00:25:32.899 Uttam Kumaran: The test is like we need to isolate some variables and and run those right. And and again, this isn’t across the entire set. This is taking a subset of folks running them through one test, running a couple through another test and then, in addition, leveraging AI for the quality of the engagement itself. Whether you can personalize the messaging, personalize the phone, call handle objections. Better

211 00:25:32.920 ⇒ 00:25:36.080 Uttam Kumaran: make a call faster. Things like that.

212 00:25:38.600 ⇒ 00:25:40.500 Brian Tucker: Okay, that’s great.

213 00:25:40.940 ⇒ 00:25:41.650 Brian Tucker: So that’s.

214 00:25:42.050 ⇒ 00:25:46.630 Uttam Kumaran: Yeah, that’s kind of like what I’m thinking. I mean for me, I think in terms of like

215 00:25:46.870 ⇒ 00:26:10.959 Uttam Kumaran: I would love to. At least, I think there’s we kind of talked about a lot. So I think for us, if we can put maybe a testing plan in front of you, I think that’s 1 deliverable that we can get your feedback on in terms of a proof of concept. You know we have. We can definitely demo you the AI sort of phone capabilities. You guys may already be familiar with that. But I I do think that this is this seems like

216 00:26:11.080 ⇒ 00:26:22.459 Uttam Kumaran: more of a data problem. And then additionally leveraging AI for the quality of the engagements. Right? How do you make the text messages more personal? The calls better. Yeah.

217 00:26:22.700 ⇒ 00:26:26.120 Scott_Harmon: Yeah, I just wanna, this is fantastic. Just

218 00:26:26.310 ⇒ 00:26:29.730 Scott_Harmon: a couple decreedal questions, Amy, do you happen to have

219 00:26:30.000 ⇒ 00:26:35.109 Scott_Harmon: an example of the link that gets sent, and where, like the landing page, I’m just curious. It’s a bit

220 00:26:35.690 ⇒ 00:26:39.579 Scott_Harmon: again. I’ve just done so much of this in a marketing context like.

221 00:26:39.580 ⇒ 00:26:42.139 Brian Tucker: Yeah, Scott, so I can send you that.

222 00:26:42.440 ⇒ 00:26:43.139 Scott_Harmon: Yeah. The other thing.

223 00:26:43.140 ⇒ 00:26:51.979 Brian Tucker: Yeah. The other thing I want to point out, too, is is that while we’re talking a lot about the outreach process and engaging with the patient directly. I can also tell you that.

224 00:26:52.700 ⇒ 00:27:11.629 Brian Tucker: When they do come to our website right, there’s a massive opportunity there to have them engage with something on the website because the page think about it. Right? You you come to me and I say, oh, yeah, you were referred to me for behavioral health. Yeah. Okay. Here, here’s the form sign up. Well, wait a minute.

225 00:27:11.730 ⇒ 00:27:29.980 Brian Tucker: You know, I got a lot of questions. I don’t know anything about this. I’m scared. I’m nervous. We’re not handling any of this right? So that’s why we’re getting. We’re getting half the people that are like. Yep, I need. No, I need to go into somebody. I need to get my meds managed. Yep, I know I got depression. I need to see somebody. I’m ready. Let’s go. Let’s do it. And then the other half of them are like, hold on.

226 00:27:29.980 ⇒ 00:27:46.660 Brian Tucker: you know. I walked into your car dealership, and you’ve got me in the finance room signing the papers. And I what you know I mean, that’s literally what it’s like, right? So there’s a huge opportunity inside of that, too, just giving them the opportunity to educate themselves and engage there as well.

227 00:27:46.660 ⇒ 00:27:51.630 Scott_Harmon: And I’m you know there’s just so many permutations here. There isn’t any one magic bullet. And right

228 00:27:51.630 ⇒ 00:27:57.179 Scott_Harmon: it’s all over this. But I had another client where they, instead of sending a

229 00:27:57.290 ⇒ 00:28:02.810 Scott_Harmon: a text or an email with a click to register link, which is, you know, we get those all the time.

230 00:28:03.220 ⇒ 00:28:06.910 Scott_Harmon: 50% conversion rates just pretty standard on yeah

231 00:28:06.910 ⇒ 00:28:11.630 Scott_Harmon: links. People are kind of conditioned against clicking links. But it’s.

232 00:28:11.630 ⇒ 00:28:11.980 Brian Tucker: Yeah.

233 00:28:11.980 ⇒ 00:28:17.029 Scott_Harmon: Instead of instead of saying, Hey, click! This link, just say, Hey, my name is my name is George.

234 00:28:17.030 ⇒ 00:28:17.580 AmyAdams: Okay.

235 00:28:17.970 ⇒ 00:28:22.229 Scott_Harmon: We got a referral, you know. We’d love to help you.

236 00:28:22.710 ⇒ 00:28:24.119 Scott_Harmon: How can I help you know

237 00:28:24.350 ⇒ 00:28:27.600 Scott_Harmon: it? Just it conversationally, instead of saying, click the link.

238 00:28:27.750 ⇒ 00:28:38.199 Scott_Harmon: In other words, there’s some people that just don’t want to click a link. They’re not ready to your point. They feel like it’s a bridge that they’re crossing. They’re not ready to cross, but if they think it’s a person they’re talking to.

239 00:28:38.740 ⇒ 00:28:42.209 Scott_Harmon: they may go. What referral like? What do you mean? Oh.

240 00:28:42.210 ⇒ 00:28:42.900 Brian Tucker: Right.

241 00:28:42.900 ⇒ 00:28:48.840 Scott_Harmon: You know, Dr. Dr. Fred said you could get a referral, and we work with them. And oh, what do you do like?

242 00:28:49.220 ⇒ 00:28:57.079 Scott_Harmon: And so they’re conversationally just, I’ll say disarming. That’s a little bit, you know, just answering questions. That

243 00:28:57.410 ⇒ 00:29:02.759 Scott_Harmon: the call to action of clicking a link is a is a big ask is what I’m saying, yeah, for a lot of people.

244 00:29:02.760 ⇒ 00:29:03.810 Brian Tucker: Absolutely.

245 00:29:03.810 ⇒ 00:29:08.459 Scott_Harmon: Maybe you maybe get better engagement rates if they could just start talking to you.

246 00:29:08.750 ⇒ 00:29:13.500 Brian Tucker: And they may just drop out and go screw you. Leave me alone, you know, right.

247 00:29:13.700 ⇒ 00:29:19.580 Scott_Harmon: You’re freaking me out. But I just wonder if you get a better engagement rate to a conversational text.

248 00:29:21.470 ⇒ 00:29:24.270 Scott_Harmon: You know. Maybe not, you know. Maybe it’s the same, you know.

249 00:29:24.530 ⇒ 00:29:26.780 Scott_Harmon: Just leave me alone. You’re bugging me.

250 00:29:26.780 ⇒ 00:29:31.410 Brian Tucker: Yeah, if you if if you guys will put your cell phone number there in the chat.

251 00:29:31.590 ⇒ 00:29:46.519 Brian Tucker: I’ll I’ll just put you in the outreach process, and and you’ll get so whoever puts their number in there, I’ll just add you in there right now, and I’ll just. I’ll just manually push you into the job, and then you’ll get. You’ll get one today, and then you’ll probably get one like every 2 days after that.

252 00:29:46.520 ⇒ 00:29:46.980 Uttam Kumaran: Yeah.

253 00:29:46.980 ⇒ 00:29:47.755 Brian Tucker: Right

254 00:29:49.080 ⇒ 00:29:52.730 Brian Tucker: And then that way you can kind of see what’s going on. I mean.

255 00:29:53.830 ⇒ 00:29:54.180 Scott_Harmon: I mean.

256 00:29:54.180 ⇒ 00:29:57.516 Brian Tucker: As as crude as it is. It works, but you know.

257 00:29:58.270 ⇒ 00:29:59.850 Scott_Harmon: Just socially, yeah, yeah.

258 00:29:59.850 ⇒ 00:30:06.719 Scott_Harmon: My dentist switched to a model. Now, where I get a text from one of the the booking people.

259 00:30:06.720 ⇒ 00:30:07.500 Brian Tucker: -

260 00:30:07.500 ⇒ 00:30:15.819 Scott_Harmon: And and versus a reminder. And I just because I there’s a social thing. I I don’t blow them off as much. I’m like, oh, yeah, I’m busy, or whatever. So

261 00:30:16.370 ⇒ 00:30:21.370 Scott_Harmon: you know, I I don’t know. It’s just a theory, but it could be again. There’s.

262 00:30:21.370 ⇒ 00:30:21.930 Brian Tucker: All right.

263 00:30:21.930 ⇒ 00:30:27.610 Scott_Harmon: Tom’s point. And you guys are so data focused, this will be easy to take. And then you can ab test things and like.

264 00:30:27.610 ⇒ 00:30:28.410 Uttam Kumaran: Exactly.

265 00:30:28.470 ⇒ 00:30:58.180 Uttam Kumaran: And then the last point I’ll make is the AI tools do a great job at giving you like incredibly rich data responses per message all of that track, and then like. So I think you know, in terms of next steps, like, maybe one. I want to take a couple of these ideas, and maybe at least we have a couple of demos around voice and text that would love to share with you. I think another thing we could do is at least put a proposal together, I do think, for sort of how you would run some of these tests.

266 00:30:58.517 ⇒ 00:31:01.469 Uttam Kumaran: And then maybe we can come up with what a proof of concept

267 00:31:01.600 ⇒ 00:31:20.729 Uttam Kumaran: could look like again. It’s great that we have a key, Kpi, which is the conversion rates and that’s perfect because we can definitely measure lift. I know we didn’t talk about the note summaries thing, but I do think that this is really rich enough to to take on first, st and I’m sure if if this works out, we can, you know, move on to.

268 00:31:21.320 ⇒ 00:31:34.700 Scott_Harmon: Well, I I don’t know. Let me let me check with the team. I do want to spend a couple of minutes here on the note. Summary thing I would like to know which of the 2, if we were gonna contemplate a pilot or proof of concept which one you’d prefer. 1st

269 00:31:34.880 ⇒ 00:31:38.640 Scott_Harmon: the note summary thing. Maybe we could just spend a few minutes on it.

270 00:31:39.228 ⇒ 00:31:43.210 Scott_Harmon: I don’t know if you all saw the little demo that Uton put together last time.

271 00:31:43.825 ⇒ 00:31:47.110 Scott_Harmon: If not, can you share it real quickly, Tom.

272 00:31:47.110 ⇒ 00:31:48.009 Uttam Kumaran: Yeah, I can. Yeah.

273 00:31:48.010 ⇒ 00:31:51.440 Scott_Harmon: Just want to again. Just make sure I understand the business drivers

274 00:31:51.640 ⇒ 00:31:54.949 Scott_Harmon: there, like what success would look like.

275 00:31:55.663 ⇒ 00:32:01.329 Scott_Harmon: And then have you guys tell us which one you think makes most sense to drill into, to focus on.

276 00:32:02.880 ⇒ 00:32:11.140 Uttam Kumaran: Yeah, so this is a demo. We worked on. It’s just all synthetic data. But basically, again, we took like a.

277 00:32:11.350 ⇒ 00:32:34.960 Uttam Kumaran: you know just a little bit of, you know. We just thought, okay, what would a patient sort of record analysis or or summary sort of situation look like, basically, one thing you may want to do is sort of take a look at an individual record and get outputted. A couple of actions that that need to be taken. For example, like it could be a scoring about how the sessions went. It could be a piece of analysis which is.

278 00:32:34.960 ⇒ 00:32:35.449 AmyAdams: You know.

279 00:32:35.450 ⇒ 00:32:38.107 Uttam Kumaran: There was a missed scheduled session.

280 00:32:38.660 ⇒ 00:32:44.690 Uttam Kumaran: and like, okay, here are some action. Here are some actions that need to be taken. You can go drill down into a specific

281 00:32:44.800 ⇒ 00:32:45.835 Uttam Kumaran: patient.

282 00:32:47.730 ⇒ 00:32:53.820 Uttam Kumaran: kind of just threw together a couple of ideas here. But yeah, this is something definitely doable.

283 00:32:53.980 ⇒ 00:33:00.959 Scott_Harmon: So what are the use cases and value props around analyzing records? That would be the top of your queue.

284 00:33:04.350 ⇒ 00:33:08.280 Brian Tucker: Okay? So that’s a great question. So there’s a laundry list of them.

285 00:33:09.107 ⇒ 00:33:14.399 Brian Tucker: But I will say, one thing right now is we have 2 resources

286 00:33:14.630 ⇒ 00:33:22.400 Brian Tucker: that essentially review all of the notes done by our providers for consistency and quality.

287 00:33:22.630 ⇒ 00:33:32.119 Brian Tucker: And you know, a grammar even right? We would like to put a process in place that takes. Yeah, okay. So

288 00:33:32.220 ⇒ 00:33:59.939 Brian Tucker: as you can imagine, the variability in these notes is deep and wide, right? And we currently do not have, for whatever reason, examples of what right looks like for a correct, complete, and compliant, as I would call fit for purpose. Note right? In other words, it’s good enough to meet the insurance company requirements, and they’re not gonna flag it. And it’s good enough for our clinical quality to be able to meet that

289 00:34:00.380 ⇒ 00:34:14.169 Brian Tucker: now that’s a separate issue. But really what we would like is a way to look at these notes and ensure they meet, whatever the criteria is that we determine, so that a human would only need to look at

290 00:34:14.480 ⇒ 00:34:29.169 Brian Tucker: some output that says, You know, these don’t match it. And this is the reason why, where it would spot a pattern and be able to go back to that person and work with that person, to put to create the corrective behavior in place, as you can imagine. Right now.

291 00:34:29.330 ⇒ 00:34:36.040 Brian Tucker: this is a hundred percent manual. And it is we need to replace this repetitive work.

292 00:34:36.260 ⇒ 00:34:42.479 Brian Tucker: and the and the the tribal knowledge that’s associated with that and get that somehow.

293 00:34:42.600 ⇒ 00:34:45.099 Brian Tucker: So that it’s, you know, completely

294 00:34:45.429 ⇒ 00:34:47.993 Brian Tucker: auto on, you know, replace with a.

295 00:34:48.710 ⇒ 00:34:49.310 Scott_Harmon: So that.

296 00:34:49.310 ⇒ 00:34:49.900 Brian Tucker: So, yeah, right.

297 00:34:49.909 ⇒ 00:34:58.969 Scott_Harmon: The value prop here from a metrics or financial perspective, or compliant like, what’s the business pain you’d be

298 00:34:59.129 ⇒ 00:35:00.899 Scott_Harmon: looking to solve or.

299 00:35:02.650 ⇒ 00:35:03.000 Brian Tucker: I mean.

300 00:35:03.000 ⇒ 00:35:03.750 AmyAdams: Now we have.

301 00:35:03.750 ⇒ 00:35:04.330 Brian Tucker: Yeah.

302 00:35:04.643 ⇒ 00:35:24.076 AmyAdams: We have our 2 people who like audit like every single note, and like that, would be like they wouldn’t have to. Wouldn’t have to audit every single note. Just be the ones that are flagged, that oh, this one may not be up to our standard, that and we would put something into place, but also maybe denial rates like

303 00:35:24.530 ⇒ 00:35:31.565 AmyAdams: If the if it’s not less denial rate, we have better quality notes, we’re getting less denials from back from insurance

304 00:35:32.660 ⇒ 00:35:39.320 AmyAdams: but also like to explain a little bit further, like, there’s certain parts in our notes that.

305 00:35:39.460 ⇒ 00:36:04.330 AmyAdams: and it’s all like just free form text, like a certain subjective box, needs to have exactly what the therapist is working on with the patient is each box being used for what it’s supposed to be used for? And is all the information there is the treatment plan box. Is it actually what the treatment is going to be? And did it change from last last week’s session? Like is the risk assessment, 100% cleared. That’s 1 of the

306 00:36:04.330 ⇒ 00:36:12.540 AmyAdams: biggest sections in our note that I would want to make sure that is monitored. Is that risk? Assessment? Asked every time.

307 00:36:12.700 ⇒ 00:36:18.799 Scott_Harmon: So it ideally. You know, this is very similar Utam, that project we’re we’re doing a similar project, not similar

308 00:36:19.290 ⇒ 00:36:22.440 Scott_Harmon: in In A, with a customer support team

309 00:36:22.560 ⇒ 00:36:28.839 Scott_Harmon: and the customer support team has these experts. When they have, when they have problems, they write up these

310 00:36:29.580 ⇒ 00:36:42.580 Scott_Harmon: these knowledge notes that for the Csrs. And then they just put them in a Google share. They’re they’re just just these notes that they’ve written up. And they’re they’re not consistently written, and they don’t use proper terminology and blah blah.

311 00:36:42.770 ⇒ 00:36:46.169 Scott_Harmon: And so what we’re doing there is, we’re basically

312 00:36:46.890 ⇒ 00:36:51.839 Scott_Harmon: providing an agent agent, so that when they’re typing things at the actually the point of

313 00:36:52.260 ⇒ 00:37:00.230 Scott_Harmon: note, you know, creating the note. They’re they’re getting sort of corrected in real time, like, Hey, you could word this better, or you left this out, or

314 00:37:00.410 ⇒ 00:37:06.945 Scott_Harmon: you know as they’re as they’re going through. Let’s let’s make up a maybe there’s 5 or 6 sections of your note that

315 00:37:07.340 ⇒ 00:37:12.820 Scott_Harmon: clinician needs to enter instead of just filling in a box. There’s actually a helper saying, Wait a minute.

316 00:37:13.030 ⇒ 00:37:16.809 Scott_Harmon: We left this out, or that’s different than the last time, whatever. It’s just kind of

317 00:37:17.020 ⇒ 00:37:21.990 Scott_Harmon: holding their hand and actually writing the text. So the text is written according to a set of standards. And

318 00:37:22.210 ⇒ 00:37:29.980 Scott_Harmon: and so that’s once you do that, now that the knowledge is clean because you’ve you’ve

319 00:37:30.160 ⇒ 00:37:35.140 Scott_Harmon: you fixed it at the point of entry the other way to do it is just audit it, and

320 00:37:35.500 ⇒ 00:37:42.189 Scott_Harmon: if there’s confusion or something wrong, you just go back and fix it, or try and clear it up, or make it better. You can do both.

321 00:37:42.570 ⇒ 00:37:43.450 Scott_Harmon: But

322 00:37:45.410 ⇒ 00:37:49.229 Scott_Harmon: you, you know. Probably the ideal way to really fix it is to fix it at the point of

323 00:37:49.770 ⇒ 00:37:54.650 Scott_Harmon: of when the person is entering their notes. Do you have a point of view on that or or.

324 00:37:55.870 ⇒ 00:37:57.109 AmyAdams: No. II,

325 00:37:57.610 ⇒ 00:38:06.940 AmyAdams: that that is a good idea. That’s the best way to catch it before it, that it’s already done, and then have to go back and communicate with that provider. Oh, like

326 00:38:07.320 ⇒ 00:38:11.590 AmyAdams: you know it, it’d be more instant gratification, for sure.

327 00:38:12.090 ⇒ 00:38:20.110 Scott_Harmon: Yeah, I mean, what you’ve got is, therapists are probably fantastic therapists, and some some are maybe better writers and note takers than others, you know. And so

328 00:38:20.270 ⇒ 00:38:22.739 Scott_Harmon: one way to think about this is, if you put

329 00:38:23.190 ⇒ 00:38:27.759 Scott_Harmon: put like a perfect note writing person next to your therapist

330 00:38:28.627 ⇒ 00:38:32.659 Scott_Harmon: you could almost imagine assistant being in the session.

331 00:38:32.830 ⇒ 00:38:37.020 Scott_Harmon: you know, making the notes perfectly while the therapist is doing the therapy.

332 00:38:37.160 ⇒ 00:38:41.980 Scott_Harmon: and so at the end, you know, they’ve got this perfect, you know, perfectly formatted.

333 00:38:41.980 ⇒ 00:38:46.700 AmyAdams: Would it be transcription in the background? It’d be like listening to the session.

334 00:38:47.330 ⇒ 00:38:51.049 Scott_Harmon: I you could do either, you know you could do both.

335 00:38:53.300 ⇒ 00:38:53.769 AmyAdams: Yeah, yeah.

336 00:38:53.770 ⇒ 00:39:01.219 Scott_Harmon: Ideally, it would be listening. But you could. You could start short of that and just take the note taking session.

337 00:39:01.370 ⇒ 00:39:04.570 Scott_Harmon: And instead of having them type things into a form

338 00:39:04.780 ⇒ 00:39:09.760 Scott_Harmon: they’re they’re conversing with. An AI helper, and

339 00:39:09.940 ⇒ 00:39:15.200 Scott_Harmon: you know they’d be typing it in, and they helper would come back and go. Maybe we could word that better. What do you think about this.

340 00:39:15.200 ⇒ 00:39:16.669 AmyAdams: Better. What do you think?

341 00:39:18.280 ⇒ 00:39:19.020 AmyAdams: Okay.

342 00:39:21.320 ⇒ 00:39:32.219 AmyAdams: okay, we. I think I hit on this last time. We are like in trials with AI transcription tool right now. But we still want this, this.

343 00:39:32.370 ⇒ 00:39:35.019 AmyAdams: this AI note taking

344 00:39:35.575 ⇒ 00:39:53.900 AmyAdams: audits, because whenever we, if we were to do that or and go with that direction? Are are they abusing the AI like, what if the the diagnosis codes aren’t right? Or you know, I just I just want to make sure that we have like something in place for. But I yeah.

345 00:39:54.640 ⇒ 00:39:58.861 Ray: Yeah, let me let me just pile on a little bit, just to clarify the problem. So

346 00:39:59.590 ⇒ 00:40:04.079 Ray: we we are already working a transcription pilot

347 00:40:04.230 ⇒ 00:40:20.770 Ray: which hop operates outside. So that that’s not currently an opportunity area. But once the note is created, we are manually reviewing the notes for compliance around specific areas that are important.

348 00:40:21.804 ⇒ 00:40:38.000 Ray: And they’re important for a variety of reasons. They’re important to the patient’s care. But they’re also important to the insurance company. So what we would need is an AI tool that looked at the unstructured data in a patient’s note and queried for conditions. That which

349 00:40:38.240 ⇒ 00:40:46.649 Ray: would require manual intervention as opposed to a hundred percent review. That’s really what we would be

350 00:40:46.790 ⇒ 00:40:54.290 Ray: looking at. I’m trying to get our highly trained clinicians out of reviewing people’s notes.

351 00:40:54.500 ⇒ 00:41:00.240 Ray: a hundred percent of them, and would like them to only be looking at 2% of the ones that are out

352 00:41:00.480 ⇒ 00:41:03.209 Ray: risk for not being in compliance.

353 00:41:03.630 ⇒ 00:41:06.720 Uttam Kumaran: Yeah, you just want to do the 80 20 basically on that.

354 00:41:06.720 ⇒ 00:41:09.159 Ray: I wanna do. 98, too. But yes.

355 00:41:09.640 ⇒ 00:41:10.120 Uttam Kumaran: Yeah.

356 00:41:10.770 ⇒ 00:41:17.570 Scott_Harmon: And again it sounds like the measure here would be out of compliant notes like you’d be.

357 00:41:17.710 ⇒ 00:41:24.410 Scott_Harmon: Your goal is to a reduce the time it takes to figure that out. But B, obviously, and reducing.

358 00:41:24.410 ⇒ 00:41:24.810 Scott_Harmon: yeah.

359 00:41:24.810 ⇒ 00:41:26.600 Ray: Client notes with.

360 00:41:26.600 ⇒ 00:41:35.020 Ray: And Scott compliance is, is not a is not a cliff right? It’s really it’s degrees of risk.

361 00:41:35.430 ⇒ 00:41:40.040 Ray: Right? I mean. Compliance is whether you there’s a requirement.

362 00:41:40.829 ⇒ 00:41:46.419 Ray: There’s a requirement in there, and it’s really the strength. So what we’ll do is we’ll be looking at weaker

363 00:41:46.870 ⇒ 00:41:50.479 Ray: notes. And those conditions could change

364 00:41:50.640 ⇒ 00:41:55.959 Ray: the the other half once we set this up, not only will it be highlighting

365 00:41:57.556 ⇒ 00:42:04.039 Ray: so if you think about it, the the core of this is, we need to query

366 00:42:04.150 ⇒ 00:42:05.920 Ray: the completed notes.

367 00:42:07.075 ⇒ 00:42:13.270 Ray: For a number of reasons. The 1st one is for strength of compliance, maybe, is a better way to say it.

368 00:42:13.610 ⇒ 00:42:32.249 Ray: and then we will eventually need it to query for other things which could be high risk patients. So, for example, we might say query notes, and tell us which patients have. 2 comorbidities are on this medication, and between the ages of 5 and 17.

369 00:42:32.800 ⇒ 00:42:33.490 Scott_Harmon: Gotcha.

370 00:42:33.690 ⇒ 00:42:40.430 Ray: Right. The next criteria might be is, take a look at someone who has had a high blood pressure.

371 00:42:40.880 ⇒ 00:42:48.289 Ray: etc. So we need a way of querying these raising alerts so that those alerts, then

372 00:42:48.510 ⇒ 00:42:52.979 Ray: trigger manual reviews or modifications of the Care Plan.

373 00:42:53.390 ⇒ 00:43:04.019 Scott_Harmon: Got it. So the scoring dashboard who Tom just showed you. If we change that to be, let’s imagine a compliance score. I’ll make up a range of one to 5.

374 00:43:04.600 ⇒ 00:43:11.210 Scott_Harmon: And let’s let’s say, or maybe you’ve already got a scoring thing, but basically only the

375 00:43:11.620 ⇒ 00:43:20.820 Scott_Harmon: the ones that were most at risk, or least compliant, would be flagged. And then you could read a bunch of notes about what was not compliant about it, you know, over to the right.

376 00:43:21.020 ⇒ 00:43:23.890 Scott_Harmon: And and I’m I’m guessing that

377 00:43:24.100 ⇒ 00:43:30.670 Scott_Harmon: noncompliance impacts your revenue collection from the provide, from the from the insurance companies and.

378 00:43:31.013 ⇒ 00:43:39.260 Ray: It, it could. So really, the the ramifications are, it is they get denied, and we have to rebuild it. So it’s manual intervention.

379 00:43:40.650 ⇒ 00:43:44.499 Ray: That’s really the biggest thing I mean, these aren’t like compliance like.

380 00:43:45.090 ⇒ 00:43:54.349 Ray: or or they require us to submit a copy of the patient record. It. It requires additional work, and it could be denials, you know. I’ll tell you.

381 00:43:56.453 ⇒ 00:44:02.270 Ray: and this is in the area. I mean our overall denial rate after rebill is only 2%,

382 00:44:02.940 ⇒ 00:44:07.479 Ray: which which is phenomenal. But I want to eliminate the manual review

383 00:44:08.557 ⇒ 00:44:13.560 Ray: which is going into making these the cleanest notes possible to get them through

384 00:44:13.780 ⇒ 00:44:19.919 Ray: our 1st time. So our 1st time approval rate with insurance is about 96%

385 00:44:20.802 ⇒ 00:44:28.639 Ray: but that’s a function of some tools that are built into our billing software, plus having a supervisor review

386 00:44:28.940 ⇒ 00:44:32.010 Ray: clinicians, notes which is wasting a lot of time.

387 00:44:32.450 ⇒ 00:44:36.500 Scott_Harmon: Okay. But that’s really well framed. That’s I think I get it.

388 00:44:36.610 ⇒ 00:44:42.130 Scott_Harmon: So we want to be conscious of time. There’s these 2 areas. I think we know a lot more about now.

389 00:44:42.684 ⇒ 00:44:49.430 Scott_Harmon: What we usually do is put together a proposal for some kind of a limited proof of concept

390 00:44:49.540 ⇒ 00:44:56.080 Scott_Harmon: that does involve some some money, but but not a much, just sort of just enough to coverm’s cost, and.

391 00:44:56.570 ⇒ 00:45:00.269 Scott_Harmon: you know, get in there and let you play around with something that’s working?

392 00:45:00.747 ⇒ 00:45:04.750 Scott_Harmon: Which which of these 2 areas would you want us to start with.

393 00:45:10.070 ⇒ 00:45:13.270 Ray: I think conversion rate is really the would be number one.

394 00:45:13.700 ⇒ 00:45:14.390 Scott_Harmon: Gotcha.

395 00:45:17.930 ⇒ 00:45:23.949 Scott_Harmon: Okay, so Utam and Connor, I do think I think what we should do is brainstorm. We

396 00:45:24.350 ⇒ 00:45:28.270 Scott_Harmon: we had a very rich conversation. We you could probably imagine

397 00:45:29.170 ⇒ 00:45:33.739 Scott_Harmon: a bunch of different things, but but some A B testing, I think, along.

398 00:45:34.090 ⇒ 00:45:36.480 Scott_Harmon: or a little, you know, kind of a structured experiment.

399 00:45:36.480 ⇒ 00:45:37.060 Uttam Kumaran: Yeah.

400 00:45:38.105 ⇒ 00:45:38.980 Scott_Harmon: Maybe

401 00:45:39.320 ⇒ 00:45:47.529 Scott_Harmon: may make sense. Why don’t we brainstorm on how we could put together something simple and straightforward? We always like to use the walk before you run.

402 00:45:48.216 ⇒ 00:45:52.419 Scott_Harmon: You know, kind of model, and then send you back.

403 00:45:53.090 ⇒ 00:46:01.830 Scott_Harmon: you know, a proposal. Let you dig into some specifics, and then you could tweak it, because it may not be exactly what you want. Is that the right next step for us to take? Now

404 00:46:06.780 ⇒ 00:46:09.840 Scott_Harmon: I’m getting a yes from Amy. Yep. Oh, wait! You’re muted again, Amy.

405 00:46:09.840 ⇒ 00:46:10.930 Uttam Kumaran: You’re muted again.

406 00:46:14.550 ⇒ 00:46:23.200 Uttam Kumaran: That’s okay. So yeah, let’s take that on. I think we do have. I think, the data stuff. I think you’ll probably way more familiar with in terms of just running a test. I think

407 00:46:23.280 ⇒ 00:46:53.169 Uttam Kumaran: we’ll come up with some great tests that I think would at least help answer some questions, but again gearing towards a conversion rate, is perfect. And then for the voice and text, AI, we’ll also come up with a couple of ideas on things we can test ideally. We’re not doing too much at once. But I I wanna demonstrate the capabilities. And then that’ll kind of give us some more clarity. So let’s come up with like, yeah, what? What would be a great timeline for a proof of concept? And we’ll get back to you next week.

408 00:46:55.590 ⇒ 00:46:58.800 Scott_Harmon: That sound good ray. Brian and Amy.

409 00:46:58.800 ⇒ 00:47:12.869 Ray: That that sounds great. A. You know, Amy’s gonna be leading this in the point we just wanted to jump on. Get us on the right problem set initially. But let’s get moving. We we have a goal to be doing, 2 or 3 of these a quarter. These projects.

410 00:47:13.350 ⇒ 00:47:19.030 Scott_Harmon: Fantastic. Well, we we like, we like to jump in and see if we can come alongside and help out.

411 00:47:20.810 ⇒ 00:47:25.029 Uttam Kumaran: Okay, perfect. Well, thanks everyone for the time. Today, I really appreciate it.

412 00:47:25.253 ⇒ 00:47:31.719 Scott_Harmon: Tom and Connor, can I get you to stay on the zoom real quick, just so we can wrap up amongst ourselves and get if we’re all finished. Okay?

413 00:47:31.720 ⇒ 00:47:32.120 Uttam Kumaran: Definitely.

414 00:47:32.120 ⇒ 00:47:33.650 Scott_Harmon: Thank you, Brian. Thank you, Amy.

415 00:47:34.210 ⇒ 00:47:38.220 Brian Tucker: Welcome guys have a great weekend. Thank you. Appreciate yours, bye.

416 00:47:41.250 ⇒ 00:47:47.240 Scott_Harmon: Oh, Amy, we still can’t hear you it bye.

417 00:47:49.900 ⇒ 00:47:52.109 Scott_Harmon: boy, she really has trouble with her sound.

418 00:47:52.410 ⇒ 00:47:53.859 Uttam Kumaran: Some set of trouble. Yeah.

419 00:47:54.680 ⇒ 00:48:00.349 Scott_Harmon: So may I just think, maybe just let you talk a little bit about what kind of a project

420 00:48:01.830 ⇒ 00:48:09.820 Scott_Harmon: you know. Again, you’ve done a lot of to me. This is this is just really standard sales stuff. I mean, it’s not sales, but it’s just. It’s a sales pipeline stuff.

421 00:48:10.750 ⇒ 00:48:14.749 Scott_Harmon: Can you expand a little more on what you might propose, and we could

422 00:48:15.330 ⇒ 00:48:17.630 Scott_Harmon: just bat that around a little. What’s your dog’s name?

423 00:48:17.840 ⇒ 00:48:20.659 Uttam Kumaran: His name is Finn, and he needs to go outside. But.

424 00:48:20.660 ⇒ 00:48:21.869 Scott_Harmon: He’s fine. Hey, Fan.

425 00:48:24.520 ⇒ 00:48:48.079 Uttam Kumaran: yeah. So 2 things, one, the the conversion rate, like marketing measurement. That’s something we do for a lot of clients actually like taking a look at their website, looking at conversion funnels, what’s working what’s not which is great, because I doesn’t seem clear that they’re doing any of that. And that’s actually like way easier than a lot of the AI stuff. I think. Second is

426 00:48:48.498 ⇒ 00:49:05.760 Uttam Kumaran: additionally, I think they have their. I don’t think their concern really is around. We want to replace, like all of our people, or like we wanna curb hiring, it’s more about the quality of the of the actual interactions. And I. So I think there as well, I think they probably I got the text as well. I mean, it’s

427 00:49:06.170 ⇒ 00:49:27.270 Uttam Kumaran: it’s kind of a lot and you’re totally right. It’s like not. It’s not personalized at all. But I do think that there’s something we could do to assist them, and they’re already on twilio, which which makes the plugin really easy. To assist them on like what is a test between call and text that they can receive from an agent. That leads to a higher conversion rate in terms of

428 00:49:27.570 ⇒ 00:49:32.420 Uttam Kumaran: like a proof of concept. I think that’s what we would go for we wouldn’t even really need.

429 00:49:32.950 ⇒ 00:49:34.060 Scott_Harmon: I’m not. I’m not

430 00:49:34.440 ⇒ 00:49:43.389 Scott_Harmon: help explain more explicitly what proof of concept you might propose, because it’s not clear to me a bunch of words, but I didn’t get what what you’d be doing.

431 00:49:43.390 ⇒ 00:49:53.329 Uttam Kumaran: Yeah. So the the 1st proof of concept test would be to help them run one test on their website where they can A B test basically.

432 00:49:53.330 ⇒ 00:49:54.440 Scott_Harmon: A B test. What?

433 00:49:55.051 ⇒ 00:50:01.080 Uttam Kumaran: Whether it’s copy, whether it’s a change in like the funnel process for registration.

434 00:50:01.230 ⇒ 00:50:05.279 Uttam Kumaran: It’s not clear that they have any sort of AV testing capabilities on their site in order to.

435 00:50:05.280 ⇒ 00:50:05.950 Scott_Harmon: Hmm, okay.

436 00:50:05.950 ⇒ 00:50:09.120 Uttam Kumaran: Just route when people get to the registration page.

437 00:50:09.620 ⇒ 00:50:09.930 Uttam Kumaran: But.

438 00:50:10.430 ⇒ 00:50:13.560 Scott_Harmon: So let me let me take the other side of that.

439 00:50:13.560 ⇒ 00:50:14.230 Uttam Kumaran: Sure.

440 00:50:14.430 ⇒ 00:50:15.300 Scott_Harmon: That.

441 00:50:17.760 ⇒ 00:50:21.960 Scott_Harmon: And I I’m not sure I believe everything. But I’m just gonna I’m just gonna be the opposite of that.

442 00:50:22.236 ⇒ 00:50:22.790 Uttam Kumaran: Sure, sure.

443 00:50:23.040 ⇒ 00:50:29.320 Scott_Harmon: I think that they have a pretty good sense whether they’ve done full blown. Ab, you know, scientific ab testing or not

444 00:50:29.680 ⇒ 00:50:34.170 Scott_Harmon: of how where their current process

445 00:50:34.790 ⇒ 00:50:42.309 Scott_Harmon: like where abandonment might happen. Right? I think they’ve probably spent a fair amount of time on their webpage layout. I’m guessing they have a fair amount of

446 00:50:42.650 ⇒ 00:50:45.289 Scott_Harmon: a fairly fairly well well-informed view of

447 00:50:46.730 ⇒ 00:50:48.800 Scott_Harmon: where the form does and doesn’t work.

448 00:50:49.330 ⇒ 00:50:52.200 Scott_Harmon: Could you a be tested and get more scientific?

449 00:50:52.320 ⇒ 00:50:59.980 Scott_Harmon: Absolutely. But I don’t think that there’s going to be any big Aha in that. And and

450 00:51:02.880 ⇒ 00:51:07.639 Scott_Harmon: I think it’s broken because you could narrow the focus and go.

451 00:51:07.760 ⇒ 00:51:14.190 Scott_Harmon: People are just not engaging. Fuck the website. They’re not going to the website. Why would I test my website

452 00:51:15.020 ⇒ 00:51:20.409 Scott_Harmon: that nobody’s going to? Right? It seems to me like you’re you’re miss. We’re we’d be missing the

453 00:51:21.100 ⇒ 00:51:24.379 Scott_Harmon: the main problem, which is, people aren’t engaging.

454 00:51:26.550 ⇒ 00:51:30.540 Scott_Harmon: And so I’m only doing that just to draw.

455 00:51:30.540 ⇒ 00:51:40.319 Uttam Kumaran: No, no, I hear you, I so I guess my perspective is, even if we get them to engage they still, the website is the still, the linchpin.

456 00:51:40.320 ⇒ 00:51:41.130 Scott_Harmon: Problem. That’s.

457 00:51:41.130 ⇒ 00:51:46.560 Uttam Kumaran: But I’m unsure how we can solve the registration problem without an understanding

458 00:51:46.760 ⇒ 00:51:50.940 Uttam Kumaran: like if they even have web metrics and like, if they can do any sort of testing there.

459 00:51:50.940 ⇒ 00:51:56.840 Scott_Harmon: But but but the problem is, I don’t think I don’t think you’re listening to what they said. Well enough like.

460 00:51:56.840 ⇒ 00:52:14.229 Uttam Kumaran: No, I know. I know. They said that they have a they they miss 10% of people that get a phone call. They don’t pick up, I know, but I can’t. We can’t even measure that if they don’t have like a data like I don’t even know whether they’re able to look at the funnel on the website. And who’s coming? From what source and what their conversions are like? How can we.

461 00:52:14.230 ⇒ 00:52:19.990 Scott_Harmon: Not their problem. They said, 60% of the people never get to the website. So why are we focusing on their website.

462 00:52:19.990 ⇒ 00:52:24.910 Uttam Kumaran: Because in order to make sure that we can get them to the website, they need to be able to measure that right.

463 00:52:25.730 ⇒ 00:52:29.859 Scott_Harmon: Well, I know people are getting the website. It’s they want it to go up.

464 00:52:30.420 ⇒ 00:52:37.330 Uttam Kumaran: But that’s what I’m but that’s what I’m saying. I don’t think they know how people are getting to the website, through what channel and what the conversion rates are.

465 00:52:37.840 ⇒ 00:52:45.840 Uttam Kumaran: they said roughly. 10% of people that get any phone call register. But they don’t. They didn’t explain anything about anything in between.

466 00:52:47.760 ⇒ 00:52:49.030 Connor Fenn: I see what you’re saying. So you.

467 00:52:49.030 ⇒ 00:53:05.690 Uttam Kumaran: Do people who get a text and then a phone call, end up going. Do people that get just get a phone call, end up converting like none of that data was clear. So for us to affect anything, I I basically need to know that, like the current state of the world. They seem to have rough metrics around like 10%.

468 00:53:05.950 ⇒ 00:53:15.489 Scott_Harmon: So let me restate it. Let me restate. Let me see if I can restate it so we can agree with agree with this. The 1st thing we need, either from them, or if they don’t have it, we can.

469 00:53:15.490 ⇒ 00:53:16.040 Uttam Kumaran: Correct.

470 00:53:16.040 ⇒ 00:53:21.960 Scott_Harmon: Can do some. A B testing is what are the exact numbers of?

471 00:53:22.970 ⇒ 00:53:27.159 Scott_Harmon: I’m going to call it engagement where people get a touch

472 00:53:27.740 ⇒ 00:53:31.109 Scott_Harmon: and click on a link and go to a website like

473 00:53:31.560 ⇒ 00:53:34.389 Scott_Harmon: we, we may have 3 different ways for them.

474 00:53:34.610 ⇒ 00:53:35.290 Uttam Kumaran: Yeah.

475 00:53:35.290 ⇒ 00:53:40.140 Scott_Harmon: To to get them to our website. What are the performance of each one of those 3 ways.

476 00:53:40.140 ⇒ 00:53:41.969 Uttam Kumaran: You just need to know. There’s attribution.

477 00:53:41.970 ⇒ 00:53:47.930 Scott_Harmon: Received a phone call during a certain hour. It could be. They received a text worded a certain way it could be.

478 00:53:48.120 ⇒ 00:53:51.269 Scott_Harmon: you know. We sent a hot air balloon over their house.

479 00:53:51.560 ⇒ 00:53:56.980 Scott_Harmon: You just want to get a more detailed breakdown of that. I’m going to call that the top of the funnel.

480 00:53:57.130 ⇒ 00:53:57.750 Uttam Kumaran: Yeah.

481 00:53:57.910 ⇒ 00:54:00.820 Scott_Harmon: And and and so.

482 00:54:01.030 ⇒ 00:54:11.450 Scott_Harmon: the theory being, we need a little bit more precise data about if we’re folks in the top of the funnel. What are you doing now? Yeah, what’s working? And what’s not? Okay. So we agree on that.

483 00:54:11.450 ⇒ 00:54:29.790 Uttam Kumaran: I’m willing. Yeah, I’m willing to ditch the we need to make site modifications, I guess. What I what more was I? I was saying is, I can’t. We can’t make the modifications at the highest point in the funnel. If there’s no attribution strategy to the conversion, and it didn’t seem

484 00:54:30.480 ⇒ 00:54:33.580 Uttam Kumaran: I’m not getting the sense that that’s there, like.

485 00:54:33.580 ⇒ 00:54:41.630 Scott_Harmon: Thank you. Okay. Now, I’m closer to agreeing with you, not quite fully agreeing with you. But so he, let’s just consider. Here’s the other problem that

486 00:54:42.310 ⇒ 00:54:45.619 Scott_Harmon: people don’t want to click on a link, no matter when you call them.

487 00:54:46.070 ⇒ 00:54:46.560 Uttam Kumaran: Sure.

488 00:54:46.560 ⇒ 00:54:50.799 Scott_Harmon: You could test calling them at 18 different times during the week, you could test

489 00:54:50.960 ⇒ 00:54:54.279 Scott_Harmon: blue emails and green emails. You could test.

490 00:54:54.620 ⇒ 00:54:59.689 Scott_Harmon: you know, phone calls with a voice of Middle Eastern like you could test anything you want.

491 00:54:59.950 ⇒ 00:55:05.750 Scott_Harmon: If the call to action, the Cta is click this link to register. There’s just some

492 00:55:06.150 ⇒ 00:55:09.449 Scott_Harmon: super high resistance clicking on a link. So

493 00:55:10.310 ⇒ 00:55:15.580 Scott_Harmon: I think we should also test a different call to action.

494 00:55:15.980 ⇒ 00:55:18.150 Scott_Harmon: Sure and clearly I’ve

495 00:55:18.310 ⇒ 00:55:30.849 Scott_Harmon: already tip my hand. That I think one is the conversational is better than the e-commerce sign up. We’ve all been doing for the last 15 years, which has a built in fu rate.

496 00:55:30.850 ⇒ 00:55:31.420 Uttam Kumaran: Yeah.

497 00:55:31.420 ⇒ 00:55:32.629 Scott_Harmon: That you can.

498 00:55:32.780 ⇒ 00:55:33.760 Scott_Harmon: You can.

499 00:55:34.380 ⇒ 00:55:36.760 Scott_Harmon: You can measure the fu rate all you want.

500 00:55:37.190 ⇒ 00:55:42.730 Scott_Harmon: I’m not trying to. That was a little bit dismissive. We should know we should better understand the Fu rate.

501 00:55:42.730 ⇒ 00:55:43.809 Uttam Kumaran: I get it, I get it.

502 00:55:43.810 ⇒ 00:55:44.320 Scott_Harmon: But.

503 00:55:44.930 ⇒ 00:55:50.870 Uttam Kumaran: So you’re saying more like instead of even just look at like, are we driving more people to the site? Broadly

504 00:55:51.230 ⇒ 00:55:54.414 Uttam Kumaran: as the proof of concept like are PE, are more people just getting.

505 00:55:54.660 ⇒ 00:55:57.939 Scott_Harmon: I don’t want to use. I want to get rid of the site altogether. Fuck the website.

506 00:55:57.940 ⇒ 00:56:05.310 Uttam Kumaran: So you’re so then. So then you’re so then. But I I guess this wasn’t also clear to me is like I was gonna ask, but whether we could just handle

507 00:56:05.460 ⇒ 00:56:13.190 Uttam Kumaran: the registration, it doesn’t seem clear that are, there’s a phone call to go register on the site, or is the phone call to actually do handle the registration.

508 00:56:13.190 ⇒ 00:56:16.910 Scott_Harmon: It is not. It’s just to get people on the site. That’s my point, like.

509 00:56:16.910 ⇒ 00:56:19.070 Uttam Kumaran: Okay. Okay. Okay.

510 00:56:19.070 ⇒ 00:56:21.610 Scott_Harmon: They have a web based registration problem.

511 00:56:21.610 ⇒ 00:56:22.699 Uttam Kumaran: Okay, you want to come.

512 00:56:23.060 ⇒ 00:56:23.900 Uttam Kumaran: So base.

513 00:56:24.350 ⇒ 00:56:26.619 Scott_Harmon: Getting people to that web based

514 00:56:26.910 ⇒ 00:56:36.020 Scott_Harmon: that website. And through it is super low conversion. I’m suggesting that the problem is that you have a web-based

515 00:56:36.560 ⇒ 00:56:42.520 Scott_Harmon: registration process. Get rid of it. I’d rather a B test, a different flow.

516 00:56:42.720 ⇒ 00:56:45.759 Uttam Kumaran: So, so saying, How can we get.

517 00:56:46.570 ⇒ 00:56:53.220 Scott_Harmon: I’m I’m exaggerating to make a point. How can we get a few more people to go to your fucked up process?

518 00:56:53.220 ⇒ 00:56:54.369 Uttam Kumaran: Yes, I get it. I get it.

519 00:56:54.370 ⇒ 00:56:56.380 Scott_Harmon: Where they’ll just abandon, anyway.

520 00:56:56.790 ⇒ 00:57:00.899 Scott_Harmon: Why don’t we? A B test? A process that where I never see a website.

521 00:57:00.900 ⇒ 00:57:03.029 Uttam Kumaran: Yes, okay. I hear you.

522 00:57:03.030 ⇒ 00:57:03.400 Scott_Harmon: If they.

523 00:57:03.400 ⇒ 00:57:19.030 Uttam Kumaran: If they’re open to if they’re good with, like we, that registration via site is like, not required, then 100 that should happen in text or through the voice AI, and that it should collect the details necessary in order to

524 00:57:20.080 ⇒ 00:57:21.790 Uttam Kumaran: take the next step. Basically.

525 00:57:22.030 ⇒ 00:57:22.560 Scott_Harmon: Yeah.

526 00:57:22.560 ⇒ 00:57:26.220 Scott_Harmon: Guess my concern is that that I know you could

527 00:57:26.900 ⇒ 00:57:32.389 Scott_Harmon: come up with some A B tests of of 5 ways right off the top to improve

528 00:57:32.580 ⇒ 00:57:34.020 Scott_Harmon: their current process.

529 00:57:34.500 ⇒ 00:57:35.450 Uttam Kumaran: I hear you, I hear you.

530 00:57:35.450 ⇒ 00:57:46.190 Scott_Harmon: I’ve done it a million times. I’ve run big sales teams that just analyze the shit out of this thing. Do a B testing all day long. There’s no doubt that what you’re talking about can improve incrementally.

531 00:57:46.560 ⇒ 00:57:51.220 Scott_Harmon: But I’m just wondering if it’s really going to have an impact on the big number, which was.

532 00:57:51.400 ⇒ 00:57:56.210 Scott_Harmon: We have a we have a terrible conversion rate like, like, would you move it up? 1%.

533 00:57:56.210 ⇒ 00:57:57.040 Uttam Kumaran: I hear you, I hear you.

534 00:57:57.040 ⇒ 00:57:58.290 Scott_Harmon: At the end of the day. They’re like.

535 00:57:58.620 ⇒ 00:58:02.889 Scott_Harmon: okay, we got 1%. That’s just not compelling, you know, that’s what I worry about.

536 00:58:02.890 ⇒ 00:58:13.739 Uttam Kumaran: Then let’s do that. Let’s I think the proposal should be. We move the registration process to a new medium that will have a higher conversion rate given. It’s conversational that can be via phone or text.

537 00:58:14.400 ⇒ 00:58:16.170 Scott_Harmon: Connor Connor, what do you think.

538 00:58:16.802 ⇒ 00:58:31.650 Connor Fenn: Just with the registration. And this being like telehealth, what details goes into that? Because our like, will there be any concerns about doing that registration process, not on the website, because there’s certain information that’s.

539 00:58:31.880 ⇒ 00:58:34.930 Connor Fenn: you know, specific to that person that they can’t.

540 00:58:34.930 ⇒ 00:58:37.800 Scott_Harmon: I don’t know. Did you? Clicked on the text yet? I haven’t.

541 00:58:37.990 ⇒ 00:58:46.759 Uttam Kumaran: I haven’t. Yeah, I have to look. But security. Wise like we. I’m not. We’ll run this all on their stuff. They’re already using twilio for part of this.

542 00:58:47.100 ⇒ 00:58:50.970 Uttam Kumaran: like we can leverage whatever stack I’m sure they’re they’re used to.

543 00:58:51.160 ⇒ 00:58:56.689 Uttam Kumaran: I I think that’s a if they’re if they’re if they’re I don’t. I just don’t know whether they were like.

544 00:58:57.740 ⇒ 00:59:00.330 Scott_Harmon: I don’t either. I was. I was talking more from my

545 00:59:00.460 ⇒ 00:59:02.630 Scott_Harmon: like. If it were me so.

546 00:59:03.110 ⇒ 00:59:08.770 Uttam Kumaran: I know we should oppose that if they’re good to just say like the registration via web isn’t required.

547 00:59:09.366 ⇒ 00:59:16.829 Uttam Kumaran: If there’s for some, I’ll need to look if it’s for some reason, like, okay, it needs to happen this way, maybe for document upload or something.

548 00:59:19.430 ⇒ 00:59:21.749 Uttam Kumaran: I don’t know. Maybe there’s something there. But.

549 00:59:22.730 ⇒ 00:59:29.809 Scott_Harmon: I mean frankly, if it was, I’m I’m a little bit, Chatbot obsessed right now, but if if I just got a Chatbot reaching out to me, and then I could

550 00:59:30.180 ⇒ 00:59:34.669 Scott_Harmon: have a long running conversation with my chat Bot, that I could even move over to my

551 00:59:35.570 ⇒ 00:59:38.680 Scott_Harmon: PC. Fill out a form or answer a question like

552 00:59:40.420 ⇒ 00:59:44.140 Scott_Harmon: kind of a conversational chat. Bot! That may span.

553 00:59:44.410 ⇒ 00:59:48.160 Scott_Harmon: you know. Maybe it takes a week. You know, I’ve got a couple questions. I don’t come back.

554 00:59:48.660 ⇒ 00:59:51.700 Scott_Harmon: The Chatbot comes back and says, Hey, you know I

555 00:59:52.260 ⇒ 00:59:55.679 Scott_Harmon: you know I saw that you wanted this, you know. Can I help you with that like.

556 00:59:56.710 ⇒ 01:00:01.509 Uttam Kumaran: Yeah, I mean, I just clicked on it. It looks like you. They may need to upload id insurance.

557 01:00:02.240 ⇒ 01:00:08.240 Uttam Kumaran: But maybe it’s like a again. This is where I think if it’s if it’s maybe it’s a dual process.

558 01:00:09.870 ⇒ 01:00:16.060 Uttam Kumaran: like, maybe collect some things, and it’s like in order to finish your application. Just go to the site like we can come up with that test.

559 01:00:16.220 ⇒ 01:00:19.220 Uttam Kumaran: But maybe that’s I think that’s good for the proposal.

560 01:00:19.620 ⇒ 01:00:23.310 Scott_Harmon: Yeah, I’m looking at. I just clicked on the link I’m looking. There’s a lot of reading.

561 01:00:24.120 ⇒ 01:00:26.579 Uttam Kumaran: Yeah, I mean, it’s it’s not. It’s not good at all.

562 01:00:26.580 ⇒ 01:00:27.309 Scott_Harmon: It’s not.

563 01:00:27.310 ⇒ 01:00:32.570 Uttam Kumaran: It’s like, it’s just like, I think, this, yeah, it’s just like it’s just really bad.

564 01:00:32.780 ⇒ 01:00:36.280 Uttam Kumaran: That’s why I was like, there’s probably some edge and just making this like.

565 01:00:36.280 ⇒ 01:00:40.270 Scott_Harmon: Did you? Did you walk through? So it’s here’s click to get started. I’m just doing it on my mobile.

566 01:00:40.270 ⇒ 01:00:42.749 Uttam Kumaran: Yeah, yeah. Keep going. Keep going. You’ll see it.

567 01:00:44.930 ⇒ 01:00:46.499 Scott_Harmon: 5 min to complete.

568 01:00:46.680 ⇒ 01:00:49.189 Scott_Harmon: Okay, you’ll be required. Oh, yeah, shit.

569 01:00:49.190 ⇒ 01:00:51.829 Uttam Kumaran: Yeah, it’s like, it’s really, really not good.

570 01:00:51.830 ⇒ 01:00:54.150 Scott_Harmon: Okay? So it’s a form. It’s a

571 01:00:54.430 ⇒ 01:00:57.899 Scott_Harmon: it’s a wizard I call these wizard based sign. Up flows.

572 01:00:57.900 ⇒ 01:00:58.760 Uttam Kumaran: Yes.

573 01:00:58.760 ⇒ 01:00:59.819 Scott_Harmon: Man that.

574 01:01:01.420 ⇒ 01:01:06.460 Scott_Harmon: Yeah, there’s a ton of resistance to those that getting people through those gates is just tough.

575 01:01:06.460 ⇒ 01:01:13.650 Uttam Kumaran: No, I mean, like not when you click on the next page. The the next step is all the way at the bottom. And there’s a Youtube video.

576 01:01:14.100 ⇒ 01:01:16.890 Connor Fenn: So that’s why I’m just like notice that you got to.

577 01:01:16.890 ⇒ 01:01:17.540 Uttam Kumaran: I have.

578 01:01:17.540 ⇒ 01:01:19.660 Connor Fenn: Just to go to the next step.

579 01:01:19.660 ⇒ 01:01:28.049 Uttam Kumaran: But I do. I do think, Scott, what you’re saying is like, bite off the the biggest problem this stuff we can solve. If they go with us.

580 01:01:28.260 ⇒ 01:01:33.519 Uttam Kumaran: I think that they’re probably more interested in seeing like a big.

581 01:01:33.520 ⇒ 01:01:38.859 Scott_Harmon: Yeah, I mean, I just wonder if you just said, Look, we’re gonna again replace the whole thing. They don’t have to go through this site.

582 01:01:38.860 ⇒ 01:01:39.570 Uttam Kumaran: Yeah.

583 01:01:39.570 ⇒ 01:01:47.459 Scott_Harmon: And and let’s just see what chat bot based. Registration, what number. Let’s a b test that.

584 01:01:47.460 ⇒ 01:01:50.130 Uttam Kumaran: Sure. Sure. Yeah. Fire replacement. Yeah.

585 01:01:50.390 ⇒ 01:01:53.070 Scott_Harmon: And, by the way, you’ll get a lot better

586 01:01:54.360 ⇒ 01:02:01.060 Scott_Harmon: fallout metrics to your point in a Chatbot, because you’ll know, like every exactly like every question in

587 01:02:01.260 ⇒ 01:02:02.130 Scott_Harmon: like you’ll know.

588 01:02:02.130 ⇒ 01:02:11.109 Uttam Kumaran: Horrible. Yeah, you’ll. It’ll yeah, exactly. And you can basically have you, just you just test individual steps, combinations of steps.

589 01:02:11.110 ⇒ 01:02:17.699 Scott_Harmon: You could tell the chat bot like you could. You could rearrange the chat bots flow almost dynamically right like

590 01:02:18.400 ⇒ 01:02:19.560 Scott_Harmon: like once.

591 01:02:19.846 ⇒ 01:02:29.029 Uttam Kumaran: Yes, exactly. No. What? Yeah. Based on what’s saying it. It rearranges the following the next set of messages. So it’s not a tree. It’s sort of like builds as it goes.

592 01:02:29.030 ⇒ 01:02:33.430 Scott_Harmon: Right. And you could just, I can imagine. I mean, this is above my pay grade. But you could have some

593 01:02:35.030 ⇒ 01:02:38.610 Scott_Harmon: prompting logic which just tweaks the chat bots.

594 01:02:38.720 ⇒ 01:02:40.480 Scott_Harmon: Conversational.

595 01:02:40.480 ⇒ 01:02:50.379 Uttam Kumaran: That’s exactly what we do for the ABC. One. It’s like you could say 3 things it needs. You could say one thing. The next question is based on the previous set of responses.

596 01:02:50.530 ⇒ 01:02:51.770 Scott_Harmon: Right? Right? That’s.

597 01:02:51.770 ⇒ 01:02:57.000 Uttam Kumaran: It’s it’s between the goal and the outcome. It’s like a black. It’s sort of just like weaves in and out.

598 01:02:57.340 ⇒ 01:02:57.710 Scott_Harmon: Got it.

599 01:02:57.710 ⇒ 01:02:58.410 Uttam Kumaran: Oh, great!

600 01:02:58.410 ⇒ 01:03:04.400 Scott_Harmon: So let’s send them a note. I’ll I can. I’ll just wrap it, you know. Just keep it simple. Hey?

601 01:03:04.990 ⇒ 01:03:09.740 Scott_Harmon: What would you think of. You know, we’re thinking about creating a proposal for an alternative

602 01:03:10.600 ⇒ 01:03:16.429 Scott_Harmon: registration approach would be a conversational chat bot that would work across mobile and desktop or whatever.

603 01:03:16.650 ⇒ 01:03:21.379 Scott_Harmon: and we’d be looking to test the following, and I’d like to a B test it versus their current.

604 01:03:21.380 ⇒ 01:03:21.970 Uttam Kumaran: Yeah.

605 01:03:22.470 ⇒ 01:03:25.879 Scott_Harmon: And I think I think their current is kind of

606 01:03:26.080 ⇒ 01:03:29.789 Scott_Harmon: so bad that it might be kind of easy to get a higher score like

607 01:03:30.880 ⇒ 01:03:37.900 Scott_Harmon: like, what if we did a pilot for 2 months? And and instead of a 30% conversion, we got a 50% conversion.

608 01:03:38.100 ⇒ 01:03:38.780 Uttam Kumaran: Yeah, I mean.

609 01:03:38.780 ⇒ 01:03:40.190 Scott_Harmon: Think of! Think of the money.

610 01:03:40.620 ⇒ 01:03:44.510 Uttam Kumaran: Totally. The other thing is like they have 5 steps on here.

611 01:03:44.670 ⇒ 01:03:48.949 Uttam Kumaran: I wanna know what each of these steps the conversion rate is.

612 01:03:49.190 ⇒ 01:03:49.900 Scott_Harmon: Right.

613 01:03:49.900 ⇒ 01:03:55.499 Uttam Kumaran: Alright and but also it looks like, basically you need insurance id and credit card.

614 01:03:56.190 ⇒ 01:03:58.880 Uttam Kumaran: I feel like we could collect all of that through.

615 01:03:58.880 ⇒ 01:03:59.400 Scott_Harmon: Right.

616 01:04:00.510 ⇒ 01:04:01.469 Uttam Kumaran: So, if it’s nothing.

617 01:04:01.470 ⇒ 01:04:06.239 Scott_Harmon: You could say you could say, would you like to take a picture of it, you know? Just

618 01:04:06.360 ⇒ 01:04:11.140 Scott_Harmon: take a picture of your credit card now, and I’ll you know, scan it in for you and blah blah blah.

619 01:04:11.140 ⇒ 01:04:13.710 Uttam Kumaran: Okay, like this, better. Yeah, okay.

620 01:04:13.930 ⇒ 01:04:21.079 Scott_Harmon: Alright. Let me draft an email. I’ll send it to you first, st and then we can, once you, you know. If you like it we can. I’ll put it in the slack thing, and then.

621 01:04:21.080 ⇒ 01:04:21.425 Uttam Kumaran: Okay.

622 01:04:21.770 ⇒ 01:04:22.919 Scott_Harmon: You said it, okay.

623 01:04:23.240 ⇒ 01:04:24.020 Uttam Kumaran: Oh, good!

624 01:04:24.200 ⇒ 01:04:25.930 Scott_Harmon: Connor, so great to meet you.

625 01:04:26.050 ⇒ 01:04:27.609 Connor Fenn: Yeah, it’s great to meet you, too. Hopefully. We’ll.

626 01:04:27.610 ⇒ 01:04:31.019 Scott_Harmon: Yeah, this will be. This will be fun right right down your fairway. I think you’re gonna keep.

627 01:04:31.020 ⇒ 01:04:45.994 Uttam Kumaran: This is good. This is good. We do the data stuff so often. So I’m like, Oh, perfect. But you’re right about like should just dish this the site, because the site is actually not core to their business at all. The site is so useless. They’re like.

628 01:04:46.300 ⇒ 01:04:50.260 Scott_Harmon: You know, the more and more as I get into AI, you could use AI.

629 01:04:50.770 ⇒ 01:04:54.810 Uttam Kumaran: Lipstick on a sass pig. But a lot of times the problem is, the sass pig.

630 01:04:55.370 ⇒ 01:04:56.000 Uttam Kumaran: Yeah.

631 01:04:56.000 ⇒ 01:04:58.939 Scott_Harmon: Like we got all these effed up websites

632 01:04:59.200 ⇒ 01:05:02.149 Scott_Harmon: that people don’t like. Just skip them, you know, just like.

633 01:05:02.150 ⇒ 01:05:08.000 Uttam Kumaran: Like everything is conversational. There’s no ui. That’s the sort of big Archie thing is like ui sort of like.

634 01:05:08.400 ⇒ 01:05:10.440 Scott_Harmon: Yeah. And then the the fun part is.

635 01:05:11.380 ⇒ 01:05:17.090 Scott_Harmon: you know, when you start introducing voice to it, you could just imagine literally conversational, you know person.

636 01:05:17.720 ⇒ 01:05:21.550 Scott_Harmon: Oh, okay, I think we’re. I think we’re in sync. I’ll I’ll draft the email.

637 01:05:22.650 ⇒ 01:05:25.369 Uttam Kumaran: Thanks. Guys, yeah, appreciate it.

638 01:05:25.610 ⇒ 01:05:26.270 Uttam Kumaran: Bye.