AI has made financial analysis available to everyone. Unfortunately, it hasn’t made everyone good at it.
The sales director has discovered Claude.
By lunchtime, he’s built a revenue forecast, identified $400,000 in potential savings, and produced a dashboard predicting a 22% improvement in operating profit.
The CEO is impressed.
FP&A is concerned.
Mostly because the revenue forecast includes deals that haven’t closed, the savings assume vacant positions have no operational consequences, and the 22% improvement relies on a definition of operating profit nobody in accounting recognizes.
But the dashboard looks fantastic.
And somewhere in the building, a finance analyst is preparing to spend the afternoon explaining why none of it works.
Welcome to the new FP&A problem.
Everyone Can Produce Financial Analysis. Who’s Responsible for Getting It Right?
For years, finance departments have encouraged business leaders to become more financially literate.
Understand your numbers. Challenge assumptions. Take ownership of performance.
Now AI tools like Claude can analyze spreadsheets, build financial models, explain variances, and generate executive-ready reports in minutes.
That’s genuinely useful.
Sales shouldn’t need to wait three days for finance to calculate pipeline conversion rates. Operations should be able to explore staffing scenarios without submitting a formal request to FP&A.
But something interesting happens when financial analysis becomes this accessible.
People start confusing the ability to produce an analysis with the ability to evaluate one.
AI has made financial answers cheaper. It hasn’t made financial judgment automatic.
And when a questionable answer arrives with a polished chart and a confident explanation, it’s surprisingly easy to mistake presentation quality for analytical credibility.
The $300,000 Mistake That Looks Like a Brilliant Strategy
Consider a hypothetical $20 million company.
The sales team uses AI to identify its least profitable customers.
The analysis finds a group generating $2 million in annual revenue at a 15% contribution margin.
The recommendation?
Eliminate those customers and focus on higher-margin accounts.
Sounds reasonable.
Until FP&A asks one inconvenient question.
“What happens to the costs when those customers leave?”
Silence.
The company would lose $2 million in revenue and $300,000 in contribution toward fixed operating costs.
If those costs remain and the company doesn’t replace the business, operating profit falls by $300,000.
The AI may have calculated the margin correctly.
The recommendation is still financially damaging under those assumptions.
That’s the danger.
A calculation can be accurate while the conclusion is completely wrong.
And the person presenting it may have no idea what they missed.
Finance Is Becoming the AI Fact-Checking Department
Here’s the irony.
AI was supposed to free finance teams from repetitive reporting so they could spend more time advising the business.
Instead, some teams are discovering a new responsibility: reviewing financial analysis produced by everyone else.
Marketing has an AI-generated customer acquisition forecast.
Sales has a revenue model.
Operations has a staffing scenario.
The CEO has asked Claude to evaluate all three.
Unfortunately, each model uses different assumptions.
Marketing expects higher conversion rates. Sales assumes stronger pricing. Operations plans staffing around lower demand.
Individually, the models might look reasonable.
Together, they describe three different companies.
And now FP&A gets to reconcile them.
For a company with 30 finance professionals, that’s frustrating.
For a lower-middle-market business with one FP&A manager and two analysts, it can consume a meaningful portion of the team’s capacity.
The company hasn’t eliminated financial planning work.
It’s multiplied the number of financial plans that need supervision.
The efficiency gains are apparently being stored somewhere very secure.
The Real Risk Isn’t AI Hallucinating
Everyone worries about AI inventing numbers.
Fair enough. That’s a problem.
But there’s a subtler risk that can be harder to catch.
AI can produce a mathematically correct analysis using the wrong business assumptions.
It can calculate growth from an outdated forecast.
It can combine bookings with recognized revenue.
It can treat a budget target as an expected outcome.
It can recommend eliminating a low-margin product without understanding how that product supports the rest of the business.
None of those mistakes requires a fabricated number.
They require missing context.
And context is often scattered across conversations, contracts, operating constraints, customer relationships, and decisions that never made it into the spreadsheet.
The most dangerous financial analysis isn’t always the one with incorrect math. Sometimes it’s the one with correct math and an incomplete understanding of the business.
Is AI Actually Devaluing FP&A?
Here’s where the conversation gets uncomfortable.
In some respects, yes.
AI is reducing the value of work that previously required specialized technical effort.
Basic variance commentary? Easier.
First-pass financial models? Faster.
Dashboard creation? Increasingly accessible.
Routine reporting? Becoming less dependent on someone manually assembling it.
If an FP&A team’s value proposition is primarily producing spreadsheets and PowerPoint presentations, there’s a genuine problem.
But that doesn’t mean the function itself is becoming less valuable.
It means the source of its value is changing.
A finance professional who understands contribution margin, operating leverage, cash flow, commercial incentives, and organizational constraints brings something different to the conversation.
They can recognize when the model’s recommendation would damage the business.
They can challenge assumptions that look reasonable individually but collapse when combined.
They can explain why a favorable expense variance might actually represent a profitability problem.
And they can tell management something it doesn’t particularly want to hear.
AI can help with all of that.
But giving someone Claude doesn’t automatically give them the judgment to do it well.
The Answer Isn’t to Take Claude Away
Finance shouldn’t respond by demanding exclusive ownership of every financial analysis.
That would be a spectacular way to become the department everyone works around.
Instead, companies need financial guardrails that allow employees to use AI without creating competing versions of reality.
Start with a few practical rules:
- One financial language. Revenue, EBITDA, gross margin, churn, headcount, and forecast assumptions need consistent definitions.
- Show your work. Material analyses should identify data sources, assumptions, reporting periods, and calculations.
- Review based on consequences. A quick departmental sensitivity analysis doesn’t need the same scrutiny as a recommendation to eliminate a product line.
- Assign accountability. Business leaders own their operating assumptions. Finance challenges material financial conclusions. Management owns the decisions.
And perhaps most importantly, teach business leaders how to question AI-generated analysis.
What assumptions drive the result?
Which costs actually change?
What information is missing?
What happens if the recommendation is wrong?
Those questions are considerably more valuable than learning another clever prompt.
FP&A’s Future Was Never About Owning the Spreadsheet
There’s a temptation to view decentralized financial analysis as a threat to finance’s authority.
Some of that concern is justified.
Poor analysis can influence real decisions, and small FP&A teams can quickly become overwhelmed by the verification burden.
But the bigger opportunity is to redefine what finance contributes.
Let departments explore their numbers.
Let AI accelerate analysis.
Let business leaders become more financially capable.
Then make FP&A responsible for something worth protecting: the quality and consistency of the financial thinking behind important decisions.
Because the future of FP&A isn’t being the only department that can produce a financial model.
It’s being the department that knows when the model is wrong.
And when the sales director announces that Claude has discovered a way to increase operating profit by 22%?
Wonderful.
Ask to see the assumptions.
Preferably before the CEO approves the bonus.
When everyone has a financial model, someone still needs to make the numbers agree. The Schlott Company’s FP&A consulting services help businesses strengthen forecasting, reporting, and financial decision-making without creating another layer of unnecessary bureaucracy.