Before You Give the Client AI’s Recommendation, Ask These Questions

AI can make a client meeting look easier than it is. It can sort a receivables report, point to slow-paying accounts, and draft a few actions. That is useful preparation. It is not a recommendation yet.

In a Forbes contributor article, Mike Milan put the boundary plainly: “The math finds the opportunity. AI helps build the options.” The advisor and owner still have to decide whether an option fits this business, this moment, and the evidence in front of them.

That distinction matters when a client has asked, “What should we do next?” A polished AI response can create the illusion that the hard part is over. The hard part is deciding what to do, who will do it, what would change the decision, and how you will know whether it worked.

Before you give the client an AI-generated recommendation, stop and ask these five questions.

1. What pressure is the client trying to relieve right now?

Start with the client’s words, not the first pattern on a dashboard. “Cash is tight,” “we cannot keep waiting on customers,” and “I need room to make payroll” are different conversations even when they all point toward receivables.

This is the Burning Issue: the pressure that deserves attention first. It keeps the meeting from becoming a tour of every ratio and every possible concern. If the owner is worried about payroll next week, a broad discussion about annual growth may be true and still be beside the point.

Ask what is urgent, what decision is actually pending, and what happens if nothing changes this month. The answer gives the recommendation a job to do.

2. What evidence supports the suggested action?

Once the client names the pressure, look for the Fuel Source. The Fuel Source is the financial or operating fact creating that pressure. AI may flag a concentration of overdue invoices. The advisor still needs to check the aging detail, payment history, terms, disputes, delivery status, and the cash calendar.

“Collect receivables faster” is not enough evidence. Which invoices? Which customers? Are they truly overdue? Has the work been accepted? Is a dispute holding up payment? A recommendation gets stronger when it is tied to facts that the client can inspect.

Imagine this: a receivables recommendation that needs a second look

Imagine this. A contractor’s owner says cash feels tight heading into the next payroll. An AI review highlights $84,000 in invoices more than 60 days old and suggests putting every late customer on hold until payment arrives.

That may be one option, but it is not a finished recommendation. The advisor looks closer. One large invoice is waiting on a signed change order. Another is tied to a customer with a long history of paying after a monthly billing review. A third has already been promised for payment next Tuesday. The remaining invoices have no documented follow-up.

The useful next step may be a short, owner-approved collection plan: resolve the change-order documentation, confirm the Tuesday payment in writing, and assign follow-up on the invoices with no response. The point is not that this plan will produce a particular amount of cash. The numbers describe a modeled opportunity, not realized cash. The owner and advisor still need to decide what is appropriate and carry it out.

3. What would make this action the wrong choice?

Every recommendation needs a disqualifier. Ask what new fact, constraint, or tradeoff would cause you to change course.

In the receivables example, an account hold could be the wrong move if it interrupts work that will fund the next payroll, if a legitimate dispute has not been resolved, or if the customer’s contract requires a different process. The plan may also change if the owner has already committed the expected collection to another obligation.

This question gives the client a way to challenge the recommendation without turning the conversation into an argument. It also keeps AI output from becoming a script. A suggestion is only as good as the conditions under which it holds.

4. Who owns the action, and what result do we expect?

A recommendation without an owner usually becomes a meeting note. Name the person responsible, the next action, the date, and the expected result before anyone leaves the room.

For the contractor, the owner may take responsibility for approving customer contact and resolving the change order. The controller may own the follow-up list. The advisor may set the review date and make sure the cash forecast reflects only receipts that have support behind them.

Write the expected result in plain language: “By next Tuesday, we expect written confirmation or a specific objection on the three invoices without a documented follow-up.” That is an expectation to test, not a promise that cash will arrive. If financing enters the conversation, the lender retains underwriting authority; an advisor’s analysis does not approve credit.

5. When will we compare actual results with what we expected?

Set the review before the action starts. At the review, compare actual receipts, customer responses, and cash timing with the expectation you recorded. Then decide what changed, what did not, and whether the next action still makes sense.

This is where the advisory work improves. Maybe the customer paid as expected. Maybe the change order was the real blockage. Maybe the follow-up revealed a billing error. The point is to learn from the client’s actual situation, not to defend the original AI draft.

AI can help organize the evidence and draft options. It cannot take responsibility for the relationship, the tradeoffs, or the result. The advisor’s role is to make those things visible before the client acts.

Turn a suggestion into a decision

Clear Path To Cash® is built for the meeting after the numbers are on the table. Advisors use the Advisor Membership to move from the client’s pressure to evidence, an advisor-reviewed recommendation, a named action, and a follow-up that tests what happened. Learn about Clear Path To Cash Advisor Membership.

That is a better standard than handing over an answer that merely sounds complete.