---
title: Client Brain — RAG Database for Client Data
type: article
created: '2026-02-09'
updated: '2026-02-09'
source_docs:
- raw/2026-02-09-impromptu-zoom-meeting-120957196.md
tags:
- ai-tools
- rag
- client-data
- knowledge-management
- automation
layer: 2
client_source: null
industry_context: null
transferable: true
---

# Client Brain — RAG Database for Client Data

## Overview

**Client Brain** is an internal AI tool under development that centralizes all client data into a single queryable knowledge base. The goal is to give any team member instant access to the full history and context of any client relationship — without digging through emails, call recordings, or shared folders manually.

The tool is built on a **Retrieval-Augmented Generation (RAG)** architecture, meaning queries return AI-generated answers grounded in actual source documents, with citations back to the originating records.

## What's Ingested

As of the Feb 9 2026 discussion, the following data sources have been loaded into the RAG database:

- **Fathom call recordings** — 500–600 meeting transcripts ingested
- **Emails** — all client-related email threads (with deduplication to handle CC'd copies)
- **Shared documents** — all files from client shared folders
- **Slack** — planned addition; comments and threads to be pulled in

## How It Works

Once the query engine is complete, users will be able to ask natural-language questions such as:

- *"What did we say about [client] last week?"*
- *"Who said [X] and when?"*
- *"Give me a summary of Crazy Lenny's over the past seven days."*

The system will return a synthesized answer **plus links to the source documents** the answer was drawn from, so users can verify or dig deeper.

## Current Status

> As of Feb 9 2026: data ingestion is complete. Mark is actively building the query engine.

- [x] Fathom transcripts ingested
- [x] Emails ingested (deduplication applied)
- [x] Shared drive documents ingested
- [ ] Query engine — in progress
- [ ] Slack integration — planned

## Intended Use Cases

| Use Case | Description |
|---|---|
| Client onboarding handoff | New team members can instantly get up to speed on a client's history |
| Meeting prep | Pull a summary of recent activity before a client call |
| Cross-team visibility | Anyone can see what's happening with any client without being CC'd on every email |
| Institutional memory | Reduces dependency on any one person's knowledge of a client relationship |

## Design Notes

- **Deduplication was non-trivial** — emails are frequently copied to multiple people, so the ingestion pipeline had to identify and collapse duplicate records before indexing.
- **Email noise** — raw email threads contain a lot of low-signal content ("hi, how are you?"). The query layer will need to handle this gracefully, either by filtering at ingest or by relying on the LLM to ignore noise when summarizing.
- **Source attribution** — a key design requirement is that answers always surface the documents they came from, keeping the tool trustworthy and auditable.

## Strategic Context

Client Brain is part of a broader initiative to use AI agents to automate operational work and free up team members for higher-value strategic activity. The vision is that instead of spending time hunting for context, account managers and strategists can ask a question and immediately have the full picture.

> *"The future is these agents and optimizations, there's no doubt."* — Mark Hope, Feb 9 2026

Related tools in development:
- [[wiki/knowledge/ai-tools/orbit-abm-platform]] — ABM campaign management platform
- [[wiki/knowledge/ai-tools/ppc-optimization-bot]] — Google/Amazon ads monitoring and optimization

## Related Docs

- [[wiki/meetings/2026-02-09-impromptu-zoom-org-structure-staffing-ai-tools]] — Source meeting where Client Brain was demoed and discussed