Nicolas Boucher Online
Most people write prompts the same way they write Google searches.
A few keywords. Maybe a sentence. Then frustration when the output is vague.
The truth is, the quality of your AI output is almost entirely determined by the quality of your input. And there are frameworks — tested, structured, repeatable — that change everything.
I have compiled 100 prompts and tips across four categories. These cover the frameworks I use in my own work, the data prompting techniques that save hours, the hacks that most finance pros never discover, and the finance-specific workflows that turn AI into a permanent part of your close process.
Read them in order or jump to what matters for you right now.
A framework is not a template. It is a thinking structure.
When you give Claude a framework, you are not just formatting your prompt. You are telling it how to think about your problem. The output quality jumps immediately.
The one I use most — and the one I teach first in every training — is CSI+FBI. It gives your AI the context, the task, and the expected output all in one prompt. But there are 24 others here, and the right one depends on what you are trying to do.
Start With CSI+FBI
Most finance prompts fail because they are too vague. "Analyze this file" is not a prompt. It is a hope.
CSI+FBI forces you to give your AI everything it needs upfront:
→ C: Context — who you are, what this is for
→ S: Specific — exactly what you need, with numbers and constraints
→ I: Instruction — the actual task, step by step
→ F: Format — how the output should look
→ B: Blueprint — an example of a good output
→ I: Identity — the persona Claude should adopt
When to Use the Other Frameworks
Not every prompt needs CSI+FBI. Here is a quick guide to choosing:
AIDA (Attention, Interest, Desire, Action) — Use when writing stakeholder communications, board memos, or investor updates that need to move someone to act.
RISEN (Role, Instructions, Steps, End Goal, Narrowing) — Use for complex multi-step tasks where you need Claude to follow a precise sequence.
SNAKE (for Python code) — Specify file type, Notebook environment, Action needed, Key data columns, Structure of output. Use every time you ask Claude to write automation code.
SMART — Use when setting AI tasks that need measurable outputs. Forces precision on scope and deadline.
Top 25 Data Prompt Tips
Data prompting is where most finance professionals have the biggest untapped opportunity.
They already have the data. They just do not know how to ask for what they need.
These 25 tips cover everything from generating Excel formulas on demand to detecting anomalies in a dataset to building a full financial dashboard outline — all from a well-structured prompt.
The Rule That Changes Data Prompting
Before you paste any data into your AI tool, do three things:
→ 1: Tell your AI what the data represents — what each column means, what the time period covers
→ 2: Tell your AI what you want to do with it — summarise, analyse, flag, forecast
→ 3: Tell your AI what format you want the output in — table, bullet points, Python code, Excel formula
Without these three, you get a generic response. With them — whether you are using Claude, ChatGPT, Gemini, or Copilot — you get analysis you can put in front of the board.
Top 25 Prompt Hacks
Frameworks get you started. Hacks make you dangerous
These 25 techniques are the difference between using Claude as a search engine and using it as a thinking partner. Most finance professionals never discover them because they are not obvious — they come from experimentation.
The one I use every single day: chain-of-thought reasoning. Telling Claude to think step-by-step before giving an answer dramatically improves accuracy on complex financial questions.
The Three Hacks That Change Everything
1. Chain-of-thought reasoning. Add “Think step-by-step” to any complex prompt. Claude will reason through the problem before giving you the answer. For financial analysis, this catches errors that instant responses miss.
2. Chunking. Never give Claude a 50-page report and ask for a summary. Break it into sections. Process each one. Then ask for a synthesis at the end. You get more accurate output and a cleaner audit trail.
3. Socratic questioning. Ask Claude “What would make this forecast wrong?” or “What assumptions is this analysis relying on?” The answers are often more valuable than the original output.
Top 25 Finance-Process Tips
This is where the rubber meets the road.
The first three sections gave you frameworks and techniques. This section gives you the actual workflows — specific prompts for specific finance tasks that you can copy, adapt, and use this week.
Every tip here is a real use case. Board pack prep. Variance commentary. Dunning letters. Closing checklists. Runway calculations. These are the tasks that eat your week, and every one of them can be accelerated with the right prompt.
The One Thing to Remember
Most finance professionals are using about 10 prompts out of 100.
That gap is not a technology problem. It is a knowledge problem.
Pick one section this week. Try three prompts you have never used. See what happens to your close time, your reporting speed, or the quality of your variance commentary.
That is how you become an AI Finance Pro.
Not by knowing all 100 prompts.
By actually using them.
https://tinyurl.com/y857yvcc
