AI data brain

AI infrastructure. CS operations

ongoing

the situation

meeting notes in the meeting tool, customer data in the CRM, conversations in the messaging platform. each living in its own place. getting a full picture of a customer meant pulling from all three.

what i built

the architecture

designed a unified intelligence layer with a central data warehouse. three live data pipelines feed it continuously:

  • meeting tool. recordings classified and synced nightly as CS, sales, or internal
  • CRM. continuous sync of companies, deals, emails, service stages, and subscriptions
  • messaging platform. confirmed customer channels, synced nightly
data sourcesmeeting toolCRMmessagingwarehousedatawarehouseAI agentactionsdocsalertsreports

on top of the warehouse, an AI agent queries, synthesizes, and acts on the data. it writes back to the docs platform, sends alerts via messaging, and generates reports on demand.

customer 360 view

built a single view combining all signals. one row per customer, updated nightly.

last meetingmeeting toollast emailCRMrenewal dateCRMmessagingmessaging platformopen to-dosdocs platformcustomer360one row per customer · updated nightly

this view powers three use cases: CSM morning briefings, automated risk alerts for accounts with no contact in 30+ days, and weekly digest emails per CSM. all generated automatically.

meeting intelligence

built classification logic that distinguishes customer meetings from internal ones, and CS conversations from sales calls. based on participant domains, title keywords, and CRM customer matching. the enriched view joins meeting data and CRM data across 23 columns.

onboarding tracker

piloted with three live accounts. each customer gets a docs page with their account status overview and open to-dos extracted from transcripts, messages, and emails. the AI agent auto-updates each page as new signals arrive. goal: prove the concept, tune the output, then roll out as the standard onboarding workflow.

the result

  • any question about the business. customers, pipeline, account health. answerable in seconds
  • CSM workflows automated: account briefs, risk signals, and weekly digests generated on a schedule
  • hundreds of meetings classified and enriched, searchable as structured data
  • internal documentation self-maintaining. the AI agent updates project pages nightly
  • organization set up to make decisions from a single source of truth
← back