Make the firm itself a Claude project
A shared Claude project loaded with the firm's policies, values, governance and knowledge sources, queryable as the first-line answer to internal questions, has emerged as a reproducible deployment pattern across engagements.
A reproducible deployment pattern has emerged across multiple engagements: build a single shared Claude project that contains the firm’s own policies, values, governance, document templates and structured knowledge, then direct staff to query it as the first place to ask any internal question. Most engagements end up calling it “Ask [Firm]”. The pattern combines three other heuristics into a packaged delivery shape.
The first is context. AI’s usefulness depends almost entirely on the context it has access to (Useful AI is a context problem). A shared project loaded with the firm’s actual material — not just gestured-at via connectors — is the simplest way to give every staff member a baseline of useful context from the moment they open the tool.
The second is policy delivery. Loading the firm’s AI usage policy into the project’s system instructions, and directing staff to ask the project about its own rules, makes the policy self-enforcing through use (Embed the AI policy in the AI itself). The same project that answers operational questions answers governance questions, with the policy interpretation woven into both.
The third is values. Embedding the firm’s stated purpose, ways of working and communication standards in the project’s instructions means the tool’s responses reflect those standards by default. The mechanism is genuinely a new one — see AI as an operational interpreter of purpose, vision and values — and the Ask X pattern is the most direct way to make it concrete.
The deployment shape, distilled from the engagements where it has worked: a top-level instruction defining the firm’s persona and standards; a cascade-of-sources skill at the project root that searches lightweight markdown content first (Structure documents for AI consumption, not just human reading) then heavier document stores; the firm’s full policy and governance content in retrievable form; the document templates and skills the firm wants its work produced through. It is the first thing every new staff member is told to use, and the answer they get from it is the firm’s answer.
The pattern’s value is that it is the simplest credible artefact a firm can build that justifies the AI subscription on its own. Without it, broad rollout depends on each user finding their own use cases through trial; with it, every user has a useful starting point from day one.