Writing on grounded knowledge.
Notes on grounding, honest abstention and sovereignty for teams that hold other people's confidential data.
Your Team Is Already Pasting Client Data Into ChatGPT
For a confidentiality-bound firm, putting client data into a consumer chatbot can be an unauthorized disclosure. Banning it does not work. Here is what the sanctioned, sovereign alternative actually has to be, stated honestly.
Tommy Jamet · readWhy Your AI Assistant Should Say 'I Don't Know'
Weak context can push an LLM's wrong-answer rate from 10% to 66%. The hardest problem in a company brain isn't answering. It's knowing when to abstain.
Tommy Jamet · readBuilding the Company Brain: Architecture for the Agent Era
Why RAG over documents isn't the answer, and what to build instead. A year of architecture lessons from shipping a Company Brain in production.
Tommy Jamet · readWhat Intelligence Analysts Know About Evidence That Most Teams Get Wrong
Analysts don't collect more data. They weigh what they already have: corroboration over conviction, source diversity over seniority, trajectory over volume. A grounded company brain runs on the same discipline.
Tommy Jamet · readHow to Capture Your Firm's Knowledge Without Adding Another Tool
The knowledge that matters is already being written in Slack and email. The problem is not that you need another place to put it. It is that none of your tools talk to each other, and the context decays before anyone can use it.
Tommy Jamet · readThe Real Cost of Losing Context Between Client Meetings
Within days of a client conversation, most of the nuance is gone. The decision survives. The reasoning behind it does not. For a firm whose product is judgment, that loss is the expensive one.
Tommy Jamet · readWhy Your Company Keeps Forgetting What It Already Knows
The knowledge that runs your firm lives in senior people's heads and scattered notes, then quietly decays. The fix is not another tool. It is a company brain that remembers, and shows its evidence.
Tommy Jamet · read