What the assistant can and can't answer

The assistant is deliberately narrow. Knowing where it stops is what makes the answers inside that boundary worth trusting.

What it can and can’t do

• It only knows what’s been logged. If something was never captured in a log, it tells you so honestly rather than guessing. It won’t invent figures; any number or date it gives comes from your records.
• It’s focused on the client you opened it against. For portfolio-wide questions (like “which clients are at risk?”) it will look across your clients, and when it does it says so clearly.
• It advises, it doesn’t act. It won’t resolve risks, log interactions or edit records for you.
• It asks when something’s genuinely unclear: for example, if a name could match two people, instead of guessing.

Before you start

• Log something first. For a brand-new client with nothing logged, the assistant will prompt you to log your first interaction. It can’t brief you on a client it has no record of. The more your team logs, the sharper it gets.
• It uses credits. Each question uses a small number of credits from your workspace’s balance, and you’re only charged once you get an answer. If your workspace is out of credits, you’ll see a prompt to upgrade or top up under Settings → Billing (an owner handles this). See Credits and billing.

Next

• How to ask: Ask the assistant.