I've spent most of the last few weeks in Q2 forecast conversations, and the same moment keeps repeating.

We spend most of the meeting tearing apart someone else's plan. Is the sales capacity real? Does the hiring ramp match the pipeline? What happens to cash if the vendor slips a quarter? Every number gets a pressure test. Then we get to the finance team's own budget, and there's no discussion. It's the one plan in the building nobody pressure-tests.

This quarter, before the finance budget goes anywhere, I want to do the one thing we'd never let another team skip. Give it a real stress test.

More and more of the finance teams I talk to are keeping it simple with one assumption: AI makes finance more efficient, so the team can do more with less, so keep the budget flat for now. Let me think that through.

I don't know if I buy it. Not because the efficiency isn't real, but because of what happens to the work once the efficiency arrives.

The last 20% is where the time goes

The testing and learning, especially with LLMs, is the cheap part. Say you want a new close process, or an automated solution for revenue recognition. You can stand something up in an afternoon that works well enough to look like magic in the meeting. Then you try to rely on it. The last 20%, the permissions and the reliability and getting the data clean enough that you'd put your name on the output, is where the time goes. And that's one process. A lot of finance work is becoming systems work now, and systems work has a different cost curve than building a spreadsheet. Whatever timeline is in your head, plan for it to be two to five times longer.

The work you free up fills right back in

The regular work doesn't go anywhere either. Automation took some of the manual load off, and we got faster at the close and faster at the reports. Then the operational work expanded to fill the space we cleared. The reward for getting efficient was more tactical work. Budget as if the speed bought you free time, and you're planning for a year that won't show up.

Your next finance hire looks different, and costs more

When you do hire, it's going to look different and probably cost more. The classic FP&A skills still matter: forecasting, variance, judgment, the ability to tell the story. But the next hires need systems fluency and a feel for data and process design on top of all that. You're no longer hiring a finance person. You're hiring someone who's part finance and part engineering, usually senior and expensive, in a market where every other team wants the same person. Look at the bench you have. Can anyone on it lay the data foundations for what's coming, or are you about to ask your best analyst to do it on nights and weekends?

Tool sprawl is its own line item

There's also the spend on tools to support the new way. Every team in the company is buying software and standing up agents, and finance is usually the one asked to make sense of where the money went. Meanwhile you've built your own stack, a few tools and a handful of half-finished agents, all cutting at the same data. Before you renew any of it, the honest question is whether one foundation you trust would do more than twenty things that each do a little. Most of the time, it would.

What a flat budget really assumes

A flat finance budget assumes the team will build all of this between the close and the board deck, with no new resources. That isn't a plan. It's how you get burnout, automations that get used once and then collect dust, and a team that reaches the next quarter's close right where it started, only now with more tools and less patience.

Finance needs room to build now, so the foundations are there to make next year easier. Skip that, and you'll spend the year doing the old work at the old speed and calling it progress.

We spend our careers making other teams' plans survive contact with reality. There's a question worth sitting with before this forecast goes anywhere. When did we last turn that same scrutiny on our own?