Brainforge Knowledge

BEFORE_AFTER_FRAMEWORK

6 min read

Before/After Framework: Agent Value Definition

Purpose: Every agent must demonstrate net new value created, not just speed improvements Requirement: All agents must include a “Before/After: Net New Value Created” section Last Updated: 2025-01-16

Why This Framework Exists

Problem: It’s easy to build agents that speed up existing work, but that doesn’t displace anything useful. We only want to build agents that create net new value—work that wasn’t happening before.

Solution: Every agent must demonstrate:

  1. BEFORE: What should happen but never does (and why)
  2. AFTER: What the agent does automatically
  3. OUTCOME: Net new value created (pipeline, insights, capabilities)

Criteria: If we can’t identify clear “before” tasks that never happen, the agent isn’t creating net new value—it’s just automating something that’s already working.

Framework Structure

BEFORE (Without Agent)

What We Say We Should Do:

  • List 3-5 tasks that should happen but don’t
  • Use “We should…” format (aspirational statements)

What Actually Happens:

  • ❌ NEVER HAPPENS - [Why it fails]
  • ❌ NEVER HAPPENS - [Why it fails]
  • ❌ NEVER HAPPENS - [Why it fails]

Why It Fails:

  • Primary reason (e.g., No time, too manual)
  • Secondary reason (e.g., No owner, no one responsible)
  • Tertiary reason (e.g., Always deprioritized, always next quarter)

AFTER (With Agent)

What Agent Does Automatically:

  • ✅ [Task 1] - [How it does it]
  • ✅ [Task 2] - [How it does it]
  • ✅ [Task 3] - [How it does it]

Human Role:

  • What human reviews/adjusts
  • What human approves/deploys
  • What human does with output

OUTCOME: Net New Value Created

Value That Didn’t Exist Before:

  1. NEW [Category 1] (e.g., PIPELINE, INSIGHTS, CAPABILITY)

    • Specific value created
    • Another specific value
    • Impact/metric

  2. NEW [Category 2]

    • Specific value created
    • Another specific value
    • Impact/metric

  3. NEW [Category 3]

    • Specific value created
    • Another specific value
    • Impact/metric

Key Metric: Without agent → [Current state]. With agent → [New state].

Value Proposition: [One sentence: What net new value does this create? Why does this displace work that wasn’t happening?]

Examples

✅ Good Example: Event Follow-Up Agent

BEFORE: “We should follow up on all event contacts within 24-48 hours with personalized emails” Reality: NEVER HAPPENS - No time, too manual, always falls by the wayside AFTER: Agent automatically parses event data, generates personalized emails, segments contacts, creates handoff summaries OUTCOME: NEW PIPELINE from event contacts that were previously ignored (0% → 100% follow-up)

Why This Works: Event follow-up wasn’t happening before—agent creates net new value (pipeline from ignored contacts).

❌ Bad Example: Email Drafting Agent (Speed Only)

BEFORE: “We should draft outreach emails faster” Reality: Emails are already being drafted (happening successfully) AFTER: Agent drafts emails faster OUTCOME: Faster email drafting (but emails were already being drafted)

Why This Fails: Emails were already being drafted successfully—agent only speeds up existing work, doesn’t create net new value.

How to Fix: Find what’s NOT happening that should be:

  • “We should personalize emails with company research” (but we don’t have time)
  • “We should track qualification signals in emails” (but we never remember to)
  • “We should follow up with non-responders automatically” (but we forget)

Framework Checklist

When creating a new agent, ask:

BEFORE Section:

  • Are these tasks that genuinely never happen? (Not just “happen slowly”)
  • Is there a clear reason why they fail? (Time, owner, priority)
  • Would these tasks create value if they happened? (Not just nice-to-have)

AFTER Section:

  • Does the agent automate these tasks completely?
  • Is the human role clear (review/approve/deploy)?
  • Can this scale without scaling headcount?

OUTCOME Section:

  • Is the value truly NEW? (Didn’t exist before)
  • Is it measurable? (Can quantify before vs after)
  • Is it meaningful? (Creates pipeline, insights, or capabilities)

Final Check:

  • If we couldn’t build this agent, would this work still not happen? (If yes, agent creates net new value ✅)

How to Use This Framework

For New Agents

  1. Start with the “BEFORE” section:

    • What should happen but never does?
    • Why doesn’t it happen?
    • Would it create value if it did?

  2. If you can’t identify clear “NEVER HAPPENS” tasks, the agent isn’t valuable:

    • Don’t build agents that just speed up existing work
    • Find what’s not happening that should be

  3. Define the “AFTER”:

    • What does the agent do automatically?
    • What’s the human’s role?
    • Can it scale?

  4. Quantify the “OUTCOME”:

    • What net new value is created?
    • Can you measure before vs after?
    • What’s the impact?

For Existing Agents

  1. Review each agent:

    • Does it have a clear “BEFORE/AFTER” section?
    • Does the “BEFORE” show tasks that never happen?
    • Does the “OUTCOME” show net new value?

  2. If not, update the agent:

    • Add the framework section
    • Clarify what net new value it creates
    • If you can’t identify net new value, deprioritize the agent

Common Patterns

Pattern 1: “Should” Tasks That Never Happen

BEFORE:

  • “We should [task]” → Never happens
  • “We should [task]” → Never happens
  • “We should [task]” → Never happens

AFTER: Agent does all of these automatically

OUTCOME: NEW value from work that wasn’t happening

Pattern 2: Manual Processes That Are Too Time-Intensive

BEFORE:

  • Task takes too long manually → Never done
  • Task requires specialized knowledge → Never done
  • Task requires tracking multiple sources → Never done

AFTER: Agent does it automatically in minutes

OUTCOME: NEW value from previously impossible work

Pattern 3: Work That Has No Owner

BEFORE:

  • Everyone’s busy, no one owns this → Never happens
  • Falls between roles → Never happens
  • Always “next quarter” → Never happens

AFTER: Agent owns this work automatically

OUTCOME: NEW value from previously unowned work

Red Flags (Agent Doesn’t Create Net New Value)

❌ “It speeds up [existing work]” - Only speed improvement, not net new value ❌ “It makes [existing process] easier” - Only convenience, not net new value ❌ “It improves [existing output] quality” - Only optimization, not net new value ❌ “[Existing work] happens faster now” - Only efficiency, not net new value

✅ “This work wasn’t happening before” - Net new value ✅ “We couldn’t do this at scale before” - Net new capability ✅ “This creates pipeline we didn’t have before” - Net new value ✅ “This gives us insights we never had before” - Net new value

Template

See gtm/agents/.claude/agents/_template.md for full agent template with Before/After framework included.

Last Updated: 2025-01-16
Requirement: All agents must include this framework section

Graph View