Running a Weekly Business Review With Your Copilot

Most owners review the business when something goes wrong. By then the revenue has already moved, the customer has already left, and the automation has been misfiring for a fortnight. The alternative is not a bigger dashboard; it is a short, boring, repeated habit. Our guide to what an AI copilot does inside a CRM exists mostly to make this habit cheap enough that you actually keep it.

Why weekly, and why the same order

A week is short enough that you can still act — a deal that stalled on Wednesday is still warm on Monday — and long enough that daily noise averages out. The fixed order matters more than the frequency. If you look at the same five things in the same sequence, you notice changes without doing any analysis, because your memory of last week is doing the comparison for you.

The agenda

Thirty minutes, five sections, asked as questions rather than assembled as reports.

1. Inbox — how many conversations came in, how many are still open, what the median first-reply time was, and which are older than your response target.
2. Pipeline — what entered and left each stage, total value in play, and every deal with no activity for two weeks.
3. Bookings and revenue — appointments booked, cancelled and no-showed, plus the value that represents against the week before.
4. Automations and templates — which ran, which errored, which templates were sent and how many got a reply.
5. AI performance — how many conversations the AI handled alone, how many it handed to a human, and what it cost.

The fifth item is the one new to most teams and the one that improves fastest. Every handoff is a gap in the knowledge base, and reading a handful of them tells you exactly what to add — the method is set out in building a knowledge base your AI can actually use.

“

Ask the same five questions in the same order every Monday and you will spot the change before the numbers do.

— be digital ai team

Look for movement, not levels

The absolute numbers matter less than the direction. Forty new enquiries is neither good nor bad; forty when you had sixty for three weeks running is a signal worth twenty minutes of thought. Ask for every figure alongside the previous period, and treat anything that moved more than about a fifth as something to explain rather than note.

Be equally suspicious of good news. A sudden jump in bookings is sometimes a marketing win and sometimes a duplicate-entry bug. The question "what changed to cause this?" is the same either way, and the copilot can usually answer it by showing where the extra volume came from.

End with decisions, not observations

A review that produces a list of interesting facts has failed. Each section should end in one of three outcomes: nothing to do, a change you make now, or a question you will answer next week. Write the changes down with a name against them — even in a two-person business, an unassigned action is an action nobody takes.

Where a change means editing the CRM itself, let the copilot draft it and then read the draft properly before approving. Familiar weekly rhythm is exactly when people start clicking approve without looking, which is why how approval gates keep an AI from breaking your CRM matters more in week twelve than in week one.

Make it repeatable

After three or four runs the agenda stops changing, and that is the moment to stop typing it. Save each section as a named routine so the definitions stay identical week to week — turning repeated work into reusable AI skills covers how to write them so they still make sense in six months. From then on the review is five invocations and your own judgement, which is the only part that was ever worth your time.

What to do with the record

Keep the output somewhere you can look back at — a document, a note, anything durable. Six weekly reviews in a row make a quarterly review write itself, and a year of them is the clearest history of the business you will ever have: not just what the numbers were, but what you decided and why. That trail is worth more than any single week's analysis.