Brainforge x UrbanStems: Forecast Project Proposal
Date: March 17, 2026 Source: Granola Meeting ID: 1f9ec868-d63b-4464-98e3-d59c09c7dc77 URL: https://notes.granola.ai/t/1f9ec868-d63b-4464-98e3-d59c09c7dc77
Participants:
- Uttam Kumaran (note creator) from Brainforge uttam@brainforge.ai
- Zack Gibbs from UrbanStems zack.gibbs@urbanstems.com
- Robert Tseng from Brainforge robert.tseng@brainforge.ai
UrbanStems Forecasting Project Discussion
- Zack interested in custom-built forecasting solution over SaaS
- Prefers agent orchestration approach using LangChain
- Already experimented with N8N for reviews automation with good success
- Believes multi-agent approach offers better control than out-of-box SaaS
- Current SNOP process entirely manual via Google Sheets
- Very little auditability when mistakes occur
- Big cost implications from forecasting errors
- Unique SOPs make custom solution more attractive
- Robert’s experience at Ruggable (post-$100M revenue)
- Required 2 full-time analysts for model maintenance
- AI agents could eliminate need for daily manual oversight
- Baseline 10% accuracy achievable, then AI handles drift monitoring
- Brainforge capabilities alignment
- Moved from N8N to Mastra (JavaScript) and similar frameworks
- Experience with agent-based database querying and analysis
- Open source approach with better observability than N8N
Cost Optimization Opportunities
- Current BI tooling costs breakdown shared
- Polyatomic: Major expense for NetSuite data extraction
- Redshift: Paying for unused data storage
- Looker: Opportunity to reduce licenses further
- Potential 20% cost reduction achievable in 1-2 quarters
- Polyatomic replacement with Hevo/Stitch connectors
- DBT Cloud easily replaceable with open source version
- Redshift optimization through data cleanup
- Robert’s perspective: Growth focus more valuable than 20% BI cost cuts
Timeline and Next Steps
- No new projects until post-Mother’s Day (end of May)
- Currently focused on 30 distinct items, 10-12 reporting-based
- Valentine’s Day lessons learned still being processed
- Optimal project window: June-September (less busy demand period)
- Marketing team shake-up expected July timeframe
- New leadership changes will settle by August-September
- Growth marketing opportunities may emerge then
- Zack to lead internal process definition before engaging Brainforge
- Uttam to follow up around June timeframe for SNOP project