Context Engine
Ask about any customer or deal and get an answer from your own records, with its source.
The problem
Your customer history is split across the CRM, inboxes, call notes and proposals. For the full picture, someone searches four places or asks whoever's been there longest.
78%
Of enterprises are struggling to connect AI to the systems they already run.
Zapier's survey of more than 500 enterprise leaders, October 2025. Read the source
We connect it to the systems you already use.
What you get
- Documents, proposals and call notes in one place
- Connected to your CRM
- The same access rules as your CRM
- Inside Claude, ChatGPT and your other AI tools
- Skills for weekly jobs, like a Monday pipeline review
How it runs
- 01
Connect
We connect your CRM and bring in your documents and call notes.
- 02
First version
It works on your own data within days.
- 03
Sign-off
Your team confirms the facts it will rely on.
- 04
Live
It goes into your AI tools and updates on a schedule.
Case study
Context Engine
One place for the team and its AI tools to ask about any account, deal or conversation.
100%
Tom Hartley, Account Executive, Sales access
What's in the pipeline for Q4?
£48k across seven open deals.
The largest is the Copperfield relaunch at £12k, now in Proposal.
| stage | deals | value |
|---|---|---|
| discovery | 3 | £14k |
| proposal | 2 | £21k |
| negotiation | 2 | £13k |
| total | 7 | £48k |
- 7 open deals, October to December
- Deal: Copperfield relaunch
- Deals and accountsCRM
- EmailsCRM
- Calls and meetingsCRM
- Proposals and contractsDocument storage
- Approved factsSigned off by the team
- Margin and pricing notesFinance only
Technical detail
- Hybrid RAG
- Retrieval-augmented generation that combines keyword and vector (semantic) search over documents with SQL over CRM records, so pipeline figures are counted, not guessed.
- Citations and provenance
- Every answer carries the IDs of the records it used, shown as citations the reader can check.
- Role-based access
- Access tiers are enforced by row-level security in the database, so a question only reaches what the asker's role allows.
- Approved knowledge
- Facts the team has signed off, like the ICP and pricing rules, are held as the source of truth and repeated word for word.
- Model Context Protocol (MCP)
- The same engine answers inside the team's AI assistants, and any LLM can sit behind it.
- CRM sync
- Workflow automation syncs deals, companies, contacts, emails, notes, calls and meetings from the CRM every hour.
Price
From £5,000 · Working on your data in days, live in 3 to 6 weeks
Includes 30 days of fixes after go-live.
No VAT is added to these prices.
Questions
- Which CRMs do you work with?
- HubSpot, Salesforce and Pipedrive, or any CRM that lets other software connect to it.
- Where does our data live?
- On accounts in your name, so it stays yours whoever runs it.
- Is our data used to train AI models?
- Everything runs on business accounts in your name, which keep your data out of AI training.
Built with
RAG and hybrid retrieval over documents and CRM, provenance on every answer, permissions that follow existing access, MCP access for your AI tools.
Next step
Give your team one screen for its weekly job, built on your own data.