Answers
Shared memory for multi-partner consultancies: what the firm standardises, what stays yours
In a consultancy each partner has their own clients and their own way of running an engagement, and both are worth keeping — so the useful question is not how to make every partner's AI answer the same way, it is which part of the answer belongs to the firm. Curated memory draws that line explicitly: what the firm sells — the method, the standards, the things it will not do, the register it writes in — is written once and sanctioned, and it reaches any AI any partner connects. How each partner runs their own account stays in their own project, unstandardised on purpose.
Last updated September 2, 2026
Standardising everything is the failure, not the goal
The reflex is a firm-wide playbook telling every partner how to work, and it dies for a reason: the partner's judgement about their own clients is exactly what those clients are paying for. But leaving all of it to each partner means the firm has no answer of its own — two engagements describe the same method differently, a standard the firm believes it has gets applied in two versions, and nobody notices until a client compares.
Two levels, and only one of them is sanctioned
The firm's level holds what is true across every engagement: how the method is named and staged, the standards the work answers to, what the firm will not do, the register it writes in. That is written once, a person at the firm approves it, and from then on it is served to whichever assistant each partner has connected. The partner's level holds the engagement — this client's constraints, this account's history, the call made last month — and it is read by that work alone.
One client's material never reaches another's work
Isolation is by project, not by policy: a memory belongs to the project it was written in, and the read that opens an assistant's work is scoped to the project it is working in. A partner cannot pull another partner's account into their own reading, and the firm-level norms are the only thing everyone shares — which is the point, because they are the only thing everyone is supposed to share.
What it looks like in practice
Two partners describe the firm's diagnostic phase to two clients in the same week. One calls it four weeks with two workshops; the other, three weeks with interviews. Neither is wrong about their own engagement, and neither is quoting the firm — there is nothing to quote. The method exists, in the sense that both partners can describe it from memory. It has just never been written anywhere an assistant could read before drafting a proposal.
Questions people ask about this
- Does this force every partner to work the same way?
- No, and the split is what prevents it. Only what the firm decided is sanctioned firm-wide; a partner's method with a particular client lives in that engagement's project and is never promoted to a firm norm by an assistant. If the firm wants something standardised, a person decides that deliberately — it never happens as a side effect of one partner writing things down.
- Our partners use different AI tools. Does the firm level still reach everyone?
- It does. The memory sits outside the assistants and is read by whichever one each person connected, so partners on different tools still share one set of firm norms. This is the case that makes the design obvious: in a firm where nobody was told what to install, the shared level has to be the part that does not depend on what anyone installed.
- Who writes the firm level, if every partner is busy?
- It accumulates as a side effect of the work rather than as a documentation project. Assistants write what was decided while the engagement is running, and a person approves or corrects it in a short review afterwards. The firm level is what survives that review more than once — not a document somebody was asked to produce over a weekend.
Where this is verifiable
Product documentation on this site (How it works, Install), the sanctioned product spec in the repository, and the Privacy Policy for what is stored, per project, and who can reach it. Everything described here is behaviour the tools apply today, not roadmap.
https://www.arroway.app/en/answers/shared-memory-for-multi-partner-consultancies