TL;DR: Budget owners increasingly take finance's reports and ask AI their real questions. Abacum CEO Julio Martínez argues FP&A teams should stop competing to answer faster and instead open the governed model to the business, retire reports that only exist to be pasted, and measure success by fewer questions.
The report has become an attachment
A controller at one of our customers recently told us what their most common request has become. It used to be "can you explain this number?" Now it's "can you send me the report, so I can give it to Claude?"
The budget owner has already decided how they want their answer. They're going to ask an AI. Finance is just the step where they first pull the data, and the report finance spent hours building ends up as an attachment to that prompt.
The end of finance as the human API
I've described finance as the human API between the plan and the rest of the company. Every question routes through your team, and everybody has a quick question. What contractors cost us last quarter. Why marketing is over. Whether there's room for one more hire in March. That model was slow, but it worked, because finance was the only place the numbers lived. That's no longer true.
I see most FP&A leaders instinctively answer faster. Give the team an AI assistant, tighten turnaround times, and build a better request queue. I think that's optimizing the wrong thing. Answering questions is the part of the job finance should be trying to lose.
Most ad hoc requests are lookups
Business partnering is supposed to be the strategic part of FP&A. But "What did we spend on software in Q3?" isn't partnership. It's a lookup with a calendar invite attached. Every one your team handles is an hour not spent on the questions nobody has asked yet: the deal quietly slipping, the hiring plan no longer matching the roadmap, the channel eating budget faster than planned. By the time those show up as a variance on a report, it's too late to change the outcome.
If finance stays the query desk, the business will route around it anyway. Budget owners will paste exports into whatever AI they like, work from last month's numbers, and build their own version of the truth. Finance loses the time and the control at once.
The software side has moved faster than most finance teams. Andreessen Horowitz wrote this month that the new wave of AI-native finance tools "does the work itself." In most companies I talk to, people still spend their time answering everyone else's questions.
What to change this quarter
Open the model to everyone. Give budget owners direct, permissioned access to the governed numbers through the AI tools they already use. If the data can answer a question, it shouldn't need a person in FP&A to answer it.
Retire the report that only exists to be pasted. If a report's main job is to be dropped into an AI chat, the report is the wrong product. Publish the definitions and the logic once, properly, and let people ask what they need.
Measure fewer questions, not faster answers. Most finance teams track turnaround time. Track volume instead, and set a goal to reduce it. A falling number means the model is doing its job.
Spend the hours on what nobody asked. The time you get back belongs to the performance loop: watching the operating signals early enough to change a decision. Our guide Systematizing decision making as a CFO covers weekly reviews and how to decline requests with evidence.
Where the FP&A job is heading
Finance wins by owning the definitions, the logic and the context that every other team's AI reads from. That's a bigger job than answering questions, and a much harder one to replace. We made a related argument about accountability for AI-produced numbers.
How many questions did your team answer last month that a budget owner could have answered themselves, if you'd let them?
How Abacum helps
Abacum is an AI-native FP&A platform that lets finance open its model to the business without losing control of the numbers. Abacum connects via MCP, so budget owners and executives can ask questions in Claude, Gemini or ChatGPT with the same governance guardrails intact, working from the company's governed numbers instead of a pasted export.
Abacum Intelligence turns plans and performance into clear, shareable explanations, and every AI-generated result is traceable: right-click any cell to see exactly which data sources, formulas and assumptions drove it. Underneath, your data stays connected to the systems it already lives in. For how the model is built, see behind the model.
