Last month's financial review started well.

You walked the team carefully through the numbers.

Revenue was up 8%, COGS variance was out by $127K, and operating expenses were above budget.

Your team had verified everything. It was all accurate, and the presentation looked amazing.

But knew what was coming next - Because there’s a BIG FP&A skills gap in your team.

Your CEO asks: "This is all great. But what should we do about it?"

You have no answer.

So you make something up, and continue to worry about the value you are delivering until you are asked this same question again.

You want to improve your quality of analysis. But there are just not enough hours in the day.

Today I’m going to explain how you can upgrade your team’s financial analysis (without needing to hire a data scientist).

So next time your CEO asks, you are much better prepared.

All Description, No Prescription

Right now, during your financial reviews, you are describing what happened.

You’re explaining why it happened, but offering no prediction of what happens next, or a prescription for what to do.

You are showing a rear view, instead of creating a map forward.

Most finance teams have mastered descriptive analysis (what happened) and do some diagnostic analysis (why it happened).

But they stop there, just before the strategic value starts.

This is a big problem, because this way your team are just reporters, not advisors. And, without advisors, the board loses faith (and you miss an opportunity).

Predictive and prescriptive analysis is hard because it requires taking a position.

"Revenue grew 8%" is safe, because it's factual.

"Revenue will likely slow to 4% next month" leaves you vulnerable, as it's a prediction you’re making that can be wrong.

But that's exactly what makes you valuable.

Becoming Indispensable

There are 4 types of analysis your team needs to master to become indispensable.

Type 1 - Descriptive: What happened? This is where most teams are.

You're reporting the facts:

Example: Our revenue hit $1.8M. Our COGS increased by $127K. Our headcount grew by 3 positions.

Type 2 - Diagnostic: Why did it happen? This is where good teams go.

You're explaining the drivers:

Example: Our Enterprise segment added $95K of revenue from 3 new logos. Our COGS variance was driven by supplier price increases and production inefficiencies.

Type 3 - Predictive: What will happen? This is where strategic value starts.

You're forecasting based on patterns:

Example: Based on our current pipeline and seasonal patterns, we expect 4-6% growth next month, with some downside risk if we struggle with Q4 renewals.

Type 4 - Prescriptive: What should we do? This is where you become indispensable.

You're recommending action:

Example: We need to increase Q4 leads pipeline on enterprise by 10% to recover the $2M gap we have with our targets. We can do this by adding additional promotional budget with social media and B2B influencers. We propose an increase of $450k based on the previous leads generation campaign of the last 3 quarters.

–

If you don't force AI to do it, it will probably not do a predictive and prescriptive analysis for you. Why? Because the models have much more examples of descriptive and diagnostic analysis than predictive and prescriptive ones.
​
This is why you have to explicitly demand it in your prompt structure.

If you don't specify all 4 types in your prompt, AI will stay mostly at descriptive and some diagnostic. It will give you 3-4 standard observations and call it done.

But when you push it. When you explicitly ask for 3 descriptive, 3 diagnostic, 2-3 predictive, and 2-3 prescriptive, you force it to do complete analysis.

Here’s how to achieve this.

​

​

The 4-Analysis Prompt Framework

As you will see below, asking for the 4 analysis types is a trick to force the AI to do better quality work (otherwise it will just be lazy).

This is because when you ask: "analyze this" it will default to the most common (average) results. Whereas, when you ask for 4 types of analysis, the AI has to go beyond the standard questions/responses it would normally give you.

Step 1

As a learning exercise, try doing some high-level analysis using individual analysis types. You can either use some of your company data (providing you are using a secure AI tool) or find publicly available data (e.g company financial statements).

Upload the data and ask.

Step 2

When the team are more competent with what the different types of analysis look like, you can build the following prompt into all of your AI analysis.

When doing analysis, make sure you are using a reasoning or ‘thinking’ model:

• ChatGPT Business / Enterprise: Select 'Thinking mode (not 'Instant')
• Claude Enterprise: Select ‘Extended Thinking‘
• Gemini: Select 'Thinking with 3 Pro' (not "Fast")
• Copilot: Use the 'Analyst' agent (not standard Copilot)

Thinking mode forces the AI to act more agentically, thinking through its task instead of generating instant outputs. This way, it reviews its own work before giving you output. For financial analysis, this dramatically improves quality.

The instant models are fast but shallow. Thinking mode takes 1-3 minutes but delivers deeper analysis.

​

​

The Bottom Line

You don’t want to be the ‘finance data reporter’.

You want to be the strategic partner that gets respect (that people actually listen to).

When you consistently deliver predictive and prescriptive analysis, you’re no longer seen as the numbers person.

You’re seen as the business partner who happens to use numbers.

And the best part is, you're not learning advanced statistics or hiring data scientists.

You're just prompting AI differently, demanding complete analysis instead of accepting basic observations.

Your Move

Take your next variance report or monthly analysis. Before you run it through AI like you normally would, add step 2 to your prompt.

That's it. One change to something you’re already doing.

See what happens. Compare the depth of insight you get, and then ask yourself: “Would this change how leadership sees my work?”

If the answer is yes, you've just found a way to get further ahead.

Best,

Your AI Finance Expert,

Nicolas

P.S. - What did you think of this edition? Hit reply and let me know (I read all replies).

P.P.S. - My new video is out! I got blown away by the number one keynote speaker on innovation and AI: Shawn Kanungo. He built live, in front of me, an app which can help many CFOs for their presentation.

Watch it now here and share it around to impress your friends and colleagues.

Behind the Scenes

This year, I had the chance to give over 50 trainings (both in-person and remote) on AI in Finance. Looking ahead to 2026, I'm excited to continue teaching and plan to travel more and organize meetup events. If you and your company want a training, feel free to reach out.