Table of Contents
Your Data May Be AI-Ready. Is Your Planning Process?
An AI model flags that revenue is tracking well below forecast for the next two quarters. The number looks plausible, and the logic holds up when you trace it back.
Do you put it into the plan?
A lot of AI in financial planning stalls at exactly that moment. The appetite is there: in GrowCFO’s AI in Planning, Budgeting and Forecasting Survey 2026, 81% named reducing manual effort as AI’s biggest opportunity in planning. What’s holding them back is control – just 1% say AI is central to how they plan today.
81%
1%
Finance wants the help, but it hasn’t handed AI the plan. Dan Wells, GrowCFO’s founder, sums up the mood in the report’s foreword: finance leaders are “unwilling to trade control for automation.”
Treat that as a design requirement. The most useful form of AI in financial planning is AI-assisted planning, where models propose forecasts and flag anomalies while finance stays accountable for what goes into the plan. Making it work depends on a governed path from an approved decision into the plan itself, with the version control and audit trail finance needs to put its name behind an AI-assisted number.
AI-Readiness Starts with Clean Data and Ends with a Governed Decision
Most AI-readiness conversations focus on the input side, and for good reason. Clean, connected data gives a model something reliable to work from, so fixing the data foundation usually comes first.
Trusted data doesn’t settle what goes into the plan, though. A revised forecast or a suggested assumption is still a proposal until someone accepts it. Finance may know about a contract the model hasn’t seen, or need to understand what the change does to cash and headcount before anyone signs off.
For planning, human-in-the-loop has to become human-in-the-model.
That means staying in the process: challenging an assumption, adjusting it, testing the consequences, and approving the result inside the planning process itself. When the survey asked what would make finance leaders comfortable with AI-generated forecasts, 46% chose human approval before changes are adopted, and 44% chose a full audit trail.
Take that revenue miss in our introduction as an example. The sales lead knows a delayed renewal is likely to close next month. So finance
- Keeps part of the AI’s adjustment
- Checks the effect on cash
- Notes the renewal as the reason
- Sends the revised forecast for approval
Every one of those steps leaves a trace in the model, which is what makes the final number defensible.
AI’s most useful job in all of this is clearing the manual assembly work that keeps finance from exercising judgment, the same work that makes continuous planning still feel manual for so many teams. But applying that judgment only matters if finance can stand behind the result.
Finance Can Only Stand Behind What It Can Govern
What makes an AI-assisted number safe to act on is everything that governs the decision behind it.
The approved decision has to stay tied to the model and the right version, with ownership and permissions intact. It needs the commentary that explains why it moved, so the reasoning stays with the number, and it has to pass through review and approval with a change history behind it. Because it’s forward-looking plan data, your actuals stay exactly as they are.
As AI proposes more of the numbers, the governed process around each decision gets more valuable. An approval trail used to be good hygiene, back when a person typed in every forecast number. Once a model proposes the number, that trail becomes the control model, and it’s the evidence a CFO relies on in front of the board. If a director asks why the Q3 forecast moved between versions, finance should be able to show who changed it and the reason they gave.
Nearly four in five respondents (78.4%) said that they would present an AI-generated forecast to the board IF finance reviewed it first or the inputs and audit trail were transparent.
https://acterys.com/blog/ai-feedback-loop-planning/And because each decision is captured with the human’s reasoning attached, the model has something trustworthy to learn from over time. So human-in-the-model does double duty: it keeps finance in control, and it helps the AI get better on finance’s terms.
78.4%
Where Finance Already Works Matters: Power BI and Excel
Excel ranked as finance’s primary reporting tool, as you’d expect, with Power BI coming in second.
A little more surprising, 1 in 10 respondents named using their BI tool for planning as the improvement they most want. While that’s a somewhat small group, it sits at the leading edge of a gap many finance teams feel between where the business analyzes performance and where finance plans it.
When finance spots a variance or an AI-flagged change in a dashboard or report, the response shouldn’t depend on exporting data to a separate tool and re-keying it later. Every export creates another copy of the plan, and each copy is another place for the commentary and approval history to drift away from the number. The decision should be reviewed, adjusted, approved and committed in the same governed model finance already analyzes in.
Think of it as one model with two front doors: Power BI for the people who live in reports, Excel for everyone who’d rather work in a spreadsheet.
Excel stays central to that picture. With 86.1% of respondents still running spreadsheet-based or spreadsheet-led planning, the practical goal is to connect Excel to the governed model, keeping the flexibility finance values while adding the version control a shared plan needs.
1 in 10
What Should CFOs Ask When AI Proposes a Number?
Start with visibility. In the survey, 55% of finance leaders said clear sight of source data would make them more comfortable with AI forecasts, and 50% said the same of transparent assumptions.
- Can finance see the source data and assumptions behind what the AI proposed?
- Is it possible to adjust the proposal, or can you only accept or reject it?
- Will you see the financial impact before the plan changes?
- Does the change go through the same review and approval as any other planning decision?
- Once it’s approved, does the value land in governed planning data with its owner and full change history intact?
If any answer involves exporting to a spreadsheet and fixing it there, the weak point sits in the process around the AI, and a better model won’t fix it.
Where AI in Financial Planning Pays Off
AI will help finance spot changes earlier and weigh responses faster. The payoff comes when an approved decision reaches the plan without breaking the governance around it, so finance can put its name to an AI-assisted forecast and defend it.
Acterys is built around that idea, adding governed planning to Power BI and Excel. Finance can adjust an AI-supported forecast, test it against scenarios, route the change for approval and commit it with a full audit history. Teams keep working in the Microsoft tools they already use, and every decision stays connected to the data behind it.
The full GrowCFO report — a study of 273 finance leaders, sponsored by Acterys — is worth reading as a benchmark. Only 20.1% of respondents are satisfied with their current planning process, and the report breaks adoption and investment down by company size and by where teams are on AI, so you can see where yours sits.
The real test is whether your process can let AI in without giving up control, and the survey is a good place to start.
