Question for you:
How many times when you ask AI to analyze something, have you got the exact answer you wanted?

Something you could show to the board, or use in a meeting without spending an hour fixing it first?

For most finance teams, the answer is never.

Meanwhile, you see teams at other companies that are completely ‘AI Native’ doing things that you did not think were possible.

This is because most of the time you are asking AI questions like you're talking to Google.

"Analyze this file." "Help me with variance analysis." "What's wrong with this forecast?"

This is more like making a wish, than giving an accurate instruction.

And when AI gives you a back generic output that doesn't understand your business (or what you wanted) you blame the tool.

Right now, I see a big difference between two types of finance pro.

On one side: Finance pros are using AI to deliver insights in hours that used to take days. They're spotting things before anyone else sees them. They're showing up to meetings with answers instead of just a lot of data.

On the other side: Finance pros are still "trying" AI. Still cleaning up messy outputs. Still spending more time fixing AI mistakes than they spend just doing it themselves.

The difference between these two groups is not the tool they're using.

It's how they're using it.

And this is what I will show you today.

Tool vs Technique

Most teams spend too much time debating which AI tool is best.

So they test them all, and ask each one the same question. "Analyze this variance report."

AI gives back three paragraphs about how "revenue increased due to market conditions and operational efficiencies, while expenses rose due to inflationary pressures and strategic investments."

This sounds very impressive. But this means nothing.

This is because AI doesn’t really know what you are asking. So it does its best to make something up that might be useful (but is often not).

So now you have to fix this response, and after 30 minutes of work, you realize you could have just written it yourself in 15 minutes.

So, you blame the tool, try a different one, and get the same results. Repeat.

All the time you are doing this, your peer in another department just delivered a complete variance analysis in 20 minutes. Accurate and board ready, with insights you didn't think of.

This is because AI Native finance pros don’t care about the tool, they focus on the technique.

Just like a trained race car driver can go faster in a slow car than you can in a fast car.

AI Native finance pros can achieve more with Gemini in Instant Mode in 20 minutes, than you can with a ChatGPT Enterprise in 2 days.

The one skill that keeps them ahead?

Prompting.

​

​

What is a Prompting Technique?

It is a structured approach or methodology used to instruct an AI model.

It’s more than just asking questions. It’s about designing your input to guide the AI towards generating a more precise, relevant, and high-quality output.

Instead of just saying "Help me with this…", a prompting technique gives specific steps to achieve a specific goal.

A vague prompt → vague answer.

A precise, well-structured prompt → clearer, more useful result.

Here are 3 prompting techniques that will turn you into an AI Finance Native in no time.

​

​

1. CSI

Important - Please only upload sensitive company data to business or enterprise AI accounts with the relevant security measures.

This framework you can apply to all of your prompts. Super easy to use, and will improve your outputs straight away.

• Context: Providing enough context so that the LLM can understand the background of what the task is.
• Specific: Clearly define the exact problem to make sure that that AI does not have to guess.
• Instruction: Tell the model what to do or the ‘task’ you want it to execute.

Sample Prompt:

“Context: I am the CFO of a mid-sized company based in Europe.

Specific: I need to prepare a cost reduction plan to present to our board next week.

Instruction: Generate a structured plan with actionable ideas.”

​

​

2. CSI + FBI

This framework builds on the previous, and is great if you want to produce more consistent outputs.

• Context: Providing enough context so that the LLM can understand the background of what the task is.
• Specific: Clearly define the exact task or focus to eliminate ambiguity.
• Instruction: Tell the model what to do, often including desired format or output style.
• Format: Defines the desired structure or layout of the output.
• Blueprint: Guides the tone, style, and quality by including examples or stylistic cues.
• Identity: Assigns a role or persona to the AI to align its responses with a specific perspective or voice.

Sample Prompt:

Context: I am the CFO of a mid-sized company based in Europe.

“Specific: I need to prepare a cost reduction plan to present to our board next week.

Instruction: Generate a structured plan with actionable ideas.

Format: Provide the output in a table with columns for: Initiative, Effort, and Risk Level.

Blueprint: Focus specifically on areas like cloud infrastructure, marketing speed, and headcount optimization. Ensure suggestions are realistic for a 200-employee company.

Identity: Act as a senior McKinsey consultant who has helped tech firms through cost transformations programs.”

​

​

3. Chunking

This technique is important if you’re working on a heavy or complex task.

It will break down complex information or problems into smaller, more manageable chunks.

With bigger tasks, there is more risk that the AI will not be able to cope, so working through it step-by-step will give you more control.

Example Scenario: Preparing a board-ready financial analysis

In this example, the data is complex. So, instead of giving it all to the AI at once, apply chunking.

Prompt 1:

You are an expert financial analyst. I’m preparing an acquisition assessment and will provide the information in several chunks. For each chunk, acknowledge receipt and summarize key points without giving recommendations yet. Wait for further instructions.

Prompt 2:

Target Company Financials (Past 3 Years).

Please summarize the trends in revenue, EBITDA, cash flow, and margins. Do not proceed beyond the summary.

→ AI returns a focused summary.

Prompt 3:

Operating Assumptions for Forecast.

Summarize the growth assumptions, cost structure, capital requirements, and working-capital needs.

→ AI returns a focused summary.

Prompt 4:

Key Risks and Integration Considerations.

Extract and summarize only the operational, regulatory, and financial risks, along with estimated integration costs.

→ AI returns a focused summary.

Prompt 5:

Synergy Opportunities and Strategic Rationale.

Provide a summary of synergy categories, expected savings, and strategic fit.

→ AI returns a focused summary.

Prompt 6:

Now produce a concise board-ready summary analyzing:

1. Financial health,
2. Valuation implications,
3. Risk-return profile, and
4. Go/no-go considerations

Use all previously provided summaries but do not repeat them verbatim.

​

​

The Bottom Line

Finance pros focusing on technique over tools are pulling ahead. They’re delivering analysis faster, showing up to meetings more prepared and becoming the go-to person for strategic questions instead of just data requests.

So, either you can spend another six months testing different tools and getting average results. Or, you can spend one week mastering one of these techniques and start seeing the difference straight away.

1. CSI for all prompts
2. CSI + FBI for more consistent results
3. Chunking for complex tasks you need to work through step-by-step

Pick a technique and use it for every AI task you do this week. See what changes.

Which will you use right now?

Best,

Your AI Finance Expert,

Nicolas

Behind the Scenes

It was amazing to spend time with members of our AI Finance Club face to face in Amsterdam earlier this year, as well as our resident AI in Finance expert Christian Martinez.

You can learn more about the AI Finance Club here.

P.S. - What did you think of this edition? Hit reply and let me know if you're planning to try this for your team (I read all replies).

P.P.S. - I’ve tackled this in how to use ChatGPT for finance in 2025. For more tips, check out my video. ​Watch it now!​