> **The question:** At our quarterly update, the finance director said a new AI tool will take over invoice matching and that our team should get the same work done with about forty percent fewer hours by next year. She said it like it was good news, and people clapped. I am the one everyone sends the messy cases to, the vendor disputes and the payments that do not match anything, and I honestly cannot tell if that makes me safe or just the last one out the door. Planning starts again in January. What should I be doing between now and then?
>
> *(an Accounts Payable Lead at a mid-size logistics company)*


The real question is not whether the tool will work. Some of it will work, and the easy matches will go first. The question is whether the people planning the cut can see the part of your job the tool cannot do, and whether they know what that part saves the company. Treat that as a [Leverage](index.html?pillar=Leverage) problem. Last week's blind spot, [The Cost Center Label](index.html?blog=the-open-floor&slug=the-cost-center-label), explained why AI savings targets land first on teams the company books as cost. Your team is one of them, so you have to make your case in finance language before January.

You can check three things before planning starts.

- **Split your work into two piles.** One pile is what the tool can count, like invoices matched per day and time spent per invoice. The other pile is what it cannot count, like the vendor disputes you settle, the duplicate payments you catch, and the calls you make when nothing on the screen lines up. The target was built on the first pile. Your case lives in the second pile.
- **Put the second pile in money.** Pull three or four real cases from the last quarter. Write down what each one would have cost if nobody had caught it, such as a paid duplicate, a late fee, or a supplier who stopped shipping. A finance director reads those numbers the same way she reads the savings target.
- **Ask to help set up the tool.** Offer to own the exception rules, which means the list of cases the tool should hand back to a person. The person who decides what the tool cannot handle is much harder to cut than the person waiting to see what it can handle.

Start with one thing this week. Write a single page with your three best cases and what each one saved, and send it to your manager before the planning cycle starts. If your manager reads it and the target does not move, that tells you how the company sees your team. Use the same page to start talking to companies that pay for that kind of judgment, while you still have a job and fresh examples to show.
