Use case: one Custom AI per client

The build, per client

Create a Custom AI named for the client and connect:

• Their voice profile — built from their published material with the voice builder, then reviewed against how they actually want to sound.
• Their positioning and strategy docs — the brand one-pager, the audience description, the offer.
• Their content board — past posts, upcoming campaigns, the swipe file for this account, as live sources.
• Their reference creators — the two or three accounts this client wants to learn from, attached as an ideas reference.

Instructions carry the account rules:

You write for [client], a [what they do] speaking to [audience]. Voice per the attached profile. Never mention competitors by name. Every post maps to one of the three content pillars in the strategy doc. Flag anything that contradicts the positioning doc instead of writing it.

Why sealed rooms beat one big brain

• No bleed. Client A's slang never shows up in client B's LinkedIn post, because B's AI has never seen A's material.
• Instant switching. Open the client's AI and you're fully in their world — no warm-up, no "let me re-read the brand guide."
• Receipts for review. The context line under each draft shows exactly which client documents produced it. When a client asks "why this angle?", you have the trail.
• Onboarding is an artifact. New teammate on the account? They chat with the client's AI, which carries the whole context you'd otherwise explain across three calls.

Scaling it

The same pattern works at any size: a freelancer with three clients, an agency with thirty. Workspace teammates see the same client AIs, so the account team drafts against one shared standard. And when you land a new client, duplicating your best-structured client AI and swapping the knowledge gets you 80% of the setup in minutes.