Where do you let the loop close?

Last week I asked where you are happy to let an AI loop finish a job on its own, and where a human still needs to be the one who decides. Here is the practical version, and it is simpler than it sounds.

Stop thinking about what the AI knows, and start thinking about what it is allowed to do at the end. Almost every task finishes one of two ways: the loop hands you something to check, or it takes an action in the world without you in between. That difference is where nearly all of the risk sits.

Take a VAT return. An AI that drafts it and leaves it for you is doing something very different from one that files it with HMRC on your behalf. The words look alike, but one hands you a decision and the other has already made it. The same line runs through an email drafted rather than sent, and a payment flagged rather than made.

You already govern people this way. You would let someone you trust draft a letter unwatched, but not move a client's money without a second pair of eyes. The task decides how much rope you give it, and you work that out by asking how hard it would be to undo.

So this week, sort the AI steps creeping into your firm by that one test, and put a human gate wherever the loop can send, file, or pay on its own. That is the line Lexendo is built around: the loop finds the risk, and you stay the one who decides.

Next week I will look at what happens when the loop gets an early number wrong, because a loop does not make a single mistake, it builds on it.

This is Post 2 of Series 2 (The AI That Acts) in The AI Footnote, my weekly note on AI in UK accountancy.

Lexendo

Find the risk across your client portfolio, before HMRC does.

Six domains of UK tax and compliance coverage. 12,580 cited UK sources. Audit Intelligence mapped to ISA 240, 315, 520, 550 and 570. 30-day money-back guarantee.

If Lexendo doesn't find a risk worth acting on in your first 30 days, full refund.