2027 budget v13.7 final (final).xlsx

“Maybe what was useless was how budgets were being built”

Every 90 days I run our AI finance accelerator.

It helps finance pros implement AI workflow they can use in their business. And one of our graduates Eduardo had this to say about budgeting.

Chances are, that right now it's budgeting time for you.

And you are probably on to V13.7 of an Excel spreadsheet.

But this is the old way.

AI has made the process a lot easier for you. You might not have experienced this yet, and you may have your doubts. But, with the right framework, it will help you build budgets faster, and make sure your budgets really come true.

To help you achieve this, I have for you a 2-part mini-series on How to use AI for Budgeting.

Part 1 – Today: How to build your budgets with AI

Part 2 – Monday: How to find budget savings with AI

3-4 days wasted

In my previous company, we had every month, 3-4 days of just booking entries into SAP.

It involved downloading all of the GL and then working in spreadsheets to combine all of the data.

We were always behind, and spent more time processing data than we did generating insight, or making recommendations about likely scenarios.

And I see this everywhere.

Kimberley from my community was really struggling with her Sage data. She was using Power Query and Power BI to form budgets for 20+ locations. But the problem was things like location names, account names, and dates were all in different formats, so she had to waste time rebuilding everything all the time (if this made you smile because you relate to it, this is exactly how I felt when I heard her when she told me this).

And Katie – also in the community – had a manual process that was 8 steps and took about 195 minutes. Steps 3 to 7 (merge, variances, top 10, commentary, formatting) took roughly 170 of those minutes.

All of this taking time away from what they really wanted.

You want to be leading budget conversations, not lagging behind with incomplete data.

And to use AI to do this effectively you need 3 things:

1. A fast way to process your data
2. The right AI instructions to help you build your budgets
3. A way to present them, and get feedback quickly

For Kimberley, she walked Claude through the transformation steps for this year's data and saved the logic as a Claude skill file.

And for Katie, the first thing it does is a data check, so a wrong budget version or a changed GL format gets caught before any commentary is written.

Now they have a much better way of working that saves them hours every month with their budgeting.

Here's how you can achieve this too.

How to build a budget with AI

Step 0 – Challenge Your Requirements

This is the step I think most people are missing.

Because every year, you take last year's budget, update the numbers, change a few assumptions, and do the same process all again.

In my experience, what made my budgets much better was first thinking about the final outcome.

It's a bit like Elon Musk's five-step process, which starts by challenging the requirements.

A lot of the time, you're budgeting things that don't need to be budgeted.

Maybe you're going into too much detail. Maybe the business has changed. Or maybe new regulations and assumptions mean parts of last year's budget are already obsolete.

So before you even look at your data, ask yourself these:

• What is the final outcome, and why? What are you trying to achieve with this budget?
• What is the current situation? Where is your business today?
• What is straightforward? What can you reuse from last year without changing anything?
• What is different? What has changed in the business, your assumptions or the way you need to budget?

And this is where you can also start working with AI. Brainstorm with AI, and make a plan with AI.

Explain to AI where last year you had a lot of difficulties, and where you want help improving.

You can even ask AI to interview you first, so it understands your business before making recommendations.

Like this, you have a clear plan before you spend hours processing data.

Step 1 – Processing Your Data

As we discussed in my previous Bots & Agents series, AI can now work autonomously to carry out tasks.

And a good use case for this is cleaning data and consolidating budget versions so your data is more consistent.

It also now matters less where your data is.

Connect your AI to your Google Drive or Outlook, your emails, or any other application that contains your data (provided your company allows it), and ask things like this. Just make sure to use each one individually instead of at the same time:

• "Search my emails for last year's shared budget versions and download the files."
• "Take all of the budget versions from my Google Drive folder and merge them into one Excel with the following calculated columns…"
• "Take these messy exports from my ERP system, and produce a budget vs actuals report that combines all 12 locations."
• "Look at the messy columns across the files that I've given you, and map them consistently to these new columns using the following logic…"

An instruction like the last one is what helped Kimberley remove several hours from the 10 hours it previously took her to map her data.

But more importantly, next year she won't start from scratch.

She will just describe what's changed (new columns, new accounts), then run her Claude Skill again.

As Kimberley put it:

Just make sure to ask AI to include an audit tab, and that you're instructing AI to use Python in the background when appropriate to produce formulas when generating Excel files (you don't need to know how Python works).

Like this, you can keep the calculations dynamically linked and check that the totals reconcile.

But you still need to verify the results.

Rule: Every number must be calculated, not generated. And when you ask for your audit tab, make sure that it includes how it’s carried out its calculations. Like this, you can check them.

Step 2 – Building Your Budget Model

Now we are moving on to how you can use AI to build your budget model.

Everyone's data and budgeting process is slightly different.

So before you build anything, ask AI what the best model for your budget is.

And this is where the prompting frameworks I teach can help.

CSI

Use it to start any prompt.

• Context: Who you are and the setting. The background the AI needs.
• Specific: The exact problem. So the AI does not have to guess.
• Instruction: The task you want done.

Start by uploading your prepared data and providing your business context, then use CSI to ask AI for different ways to build your model.

Example:

This was a method we taught in our AI Finance Club 'Budgets and Blueprints' Masterclass, where the output for one team came back with these 3 options:

1. Lender-first, covenant-driven. Build the whole model backward from proving the bank’s covenants every quarter.
2. Project-by-project, bottom-up.
3. AI-native interactive dashboard.

Like this, instead of immediately building a model based on last year's spreadsheet, you can compare different approaches and decide which one works best for your business.

If you're using ChatGPT, you can use the 'branching' facility to split your chat into a conversation for each proposed approach so you can compare.