How a newcomer with no prior process built a budgeting and forecasting function from scratch by using AI, nothing else.

THE SETUP
Day one. You have been handed the budgeting and forecasting brief for an organisation. Nobody has
given you a guide, a template or a predecessor to call. Most people in this position would freeze, copy
last year’s numbers and hope for the best. You are going to do something different. You are going to
open AI and actually figure it out. What follows is what that looks like in practice and why the result is
better than anything built the old way.

Everyone who has budgeted before starts by opening last year’s file. You do not have one. So, you do the
one thing the experienced professional almost never does: you ask what the budget is actually for.

Not what it should contain or what format it should be in. But what should it help the business decide?

Open AI. Describe the company in plain language like what it sells, how it makes money, what it spends on, who calls the shots. Then ask: what does a budget for this business need to do? What questions should it be able to answer? What decisions will it drive?

The output gives you something no template ever could i.e. clarity on the destination before you have touched a single number. Now you now know what the good looks like. Now every decision you make from this point is tested against that.

The experienced team sends a template and submissions come back in eight different formats. Someone budgets headcount one way while someone else puts it in another. The same expense is coded in three different ways. The next two weeks are spent untangling the inputs, not analysing them.

You do not send a template. You design the structure first.

Describe the organisation to AI: the departments, the cost categories, the reporting lines and the planning horizon. Ask it to design what submissions should look like, what each part of business needs to provide, in what format and why. Ask it to flag the most common inconsistencies in budgets like yours and build safeguards against them into the structure before anything arrives.

Every submission fits and nothing needs reformatting. The data arrives ready to use because it was
designed to. While the experienced team takes three weeks to collect and clean, you receive in two days
and that too already structured.

“The template was designed for last year’s company. You designed for this one.”

Here is the sharpest difference between the traditional approach and yours.

The experienced analyst builds a model which takes a week. It has forty tabs with assumptions buried in
formulas. When the CFO asks what happens if revenue drops 10%, the analyst says: give me until Thursday.

You are not building a model. You are asking questions.

Feed your inputs into AI like the revenue assumptions, the cost base and the headcount plan. Then ask:
where is this budget most exposed? Which three assumptions would hurt most if they were wrong? What
does a realistic downside look like and what would we need to do about it?

AI answers in minutes. You are in the room when the CFO asks the question. The answer is ready. Not just
because you built a better model but because you started with the question.

A forecast is not a prediction. It is a live view of where the business is heading given what you know today. Most organisations treat it as a quarterly reconciliation exercise. By the time it’s finished it describes a business that has already moved on.

You approach it differently because you have never learned to do it the slow way.

As the month unfolds, you feed AI the incoming actuals and ask it to compare them against the budget. Not to produce a variance table that any system can do but to tell you what the variances mean. Which ones signal a real change in the business? Which are the timings? Which need a decision before month end?

The forecast update takes an afternoon, not a week. It tells leadership what is happening and what they need to do about it. That is the difference between a forecast and a decision tool.

“The variance report describes what happened.
The forecast tells you what to do next.”

The traditional board pack is a data delivery exercise. Thirty slides with every variance explained and every cost line present. It takes two days to produce and arrives the morning of the meeting. The board reads the executive summary. Then someone asks a question that the pack does not answer.

You have never produced a board pack before, so you do not know it is supposed to be that way.

You ask instead: what does the board need to decide this month? Then you describe that to AI. The financial position, what is moving and why, the decision on the table. AI drafts the output around that decision and not the dataset. Not every number, just the ones with a story. Not every variance, just the ones that needs
a response.

You walk in with something the board can act on. They will not know you have been doing this for six days. That is the whole point.

The traditional budgeting process was built around genuine constraints when data was hard to gather, analysis took days and producing a presentable output required a team. AI removes every one of those constraints.

The newcomer’s advantage is not intelligence or experience but the absence of habit. They never learned to spend three weeks on just data cleanup because nobody told them they were supposed to.

They never learned to wait until Thursday for a scenario because nobody showed them the forty tab
model. They just asked AI and got the answer. That’s the smart way of approaching it.

Six days. No template. Better budget.