You Do Not Need to Be a Genius With Excel Anymore. AI Does That Now.
AI tools have fundamentally changed what a financial analyst or modeller can accomplish alone. Tasks that previously required hours of manual work --- building model structure, auditing formulas, running scenario analysis, generating investor narratives --- can now be done in minutes with the right prompts. The analyst who uses AI does not just work faster. They work at a level of breadth and consistency that was previously only possible with a team. What AI cannot replace is the judgment about which numbers matter and why.
Author: Yanni Papoutsi · Fractional VP of Finance and Strategy for early-stage startups · Author, *Raise Ready*
Published: 2025-03-08 · Last updated: 2025-03-08
Reading time: \~8 min
How Is AI Changing the Job of the Financial Analyst?
The financial analyst's job is changing in the same direction as every knowledge work role: the mechanical parts are being automated, and the judgment parts are becoming more valuable. For analysts and modellers specifically, this means the gap between what a solo operator can produce and what a full team previously produced is closing quickly. Key facts at a glance:
The Tasks Where AI Actually Helps (And Is Not Just a Gimmick)
There is a lot of noise about AI in finance. Here is what is actually useful right now, based on applying these tools to real models across multiple companies:
Formula auditing and error checking.
Ask an AI to review a financial model and identify potential circular references, broken links, or logic errors. This used to take a modeller hours of careful tab-by-tab work. With the right prompt and a pasted section of model structure, it takes minutes. The AI is not always right, but it surfaces candidates for review faster than manual auditing.
Scenario and sensitivity analysis construction.
Building a clean scenario toggle that switches between conservative, base, and aggressive assumptions across a full model is fiddly and error-prone when done manually. AI can generate the structural logic, the named ranges, the switch formulas, and the summary tab in a fraction of the time. The analyst still needs to define the scenarios. The mechanics become much faster.
Translating model outputs into investor narratives.
One of the most time-consuming tasks in fundraising prep is writing the commentary that explains the numbers: why the model makes the assumptions it does, what the key sensitivities are, how the business dynamics drive the forecast. AI can generate a strong first draft of this commentary from a structured model input. The analyst edits and adds judgment. The blank page problem goes away.
Market research and comparable analysis.
Building a comparable company set for benchmarking unit economics or valuation multiples previously required hours of manual data collection. AI tools with web access can produce a first-pass comparable set in minutes, which the analyst then refines and validates.
Due diligence preparation.
Generating a list of 30 investor diligence questions mapped to specific model tabs, then drafting the answer framework for each one: this is a task AI does well and that previously required experienced senior finance time. A founder using AI for this arrives at investor meetings better prepared than one who did not.
Key insight: The value of AI in financial analysis is not about replacing the analyst. It is about eliminating the ceiling on what one analyst can do alone. The limiting factor shifts from hours in the day to the quality of judgment about which questions to ask.
How to Actually Use AI in a Financial Model Workflow
The difference between using AI and using AI well is prompt quality. Here is the workflow that produces consistent results:
Step 1: Structure the model first, use AI second.
AI is a powerful collaborator once the model structure is defined. It is less useful when asked to design the structure from scratch, because it does not know your business. Define your drivers, your revenue logic, your cohort approach. Then use AI to build the mechanics faster. Step 2: Use AI for formula generation, not for assumption setting. Tell AI: "Here is my revenue model structure. Generate the Excel formulas for a monthly cohort calculation that outputs NRR by period." Do not tell AI: "What should my churn assumption be?" The first question has a mechanical answer. The second requires business judgment that AI cannot have.
Step 3: Use AI to audit what you built.
After building a section of the model, paste the logic into an AI tool and ask: "Are there any logical errors in this revenue calculation? What edge cases might break this?" This is faster and often more thorough than self-auditing.
Step 4: Use AI for the investor narrative.
Once the model is final, use AI to generate the first draft of the assumptions commentary: "Here are my key model assumptions and the rationale behind each. Generate an investor-facing explanation of this model's logic and key sensitivities in 500 words." Edit heavily. Do not publish first drafts.
What AI Cannot Do (And Where the Judgment Still Lives)
There is a specific failure mode in using AI for financial analysis that is worth naming directly: AI produces outputs that look correct and are wrong. A formula that is syntactically valid but logically backwards. A market sizing calculation that uses the right structure but misunderstands what the numbers represent.
This is not a reason to avoid AI. It is a reason to apply judgment to everything it produces.
The things that still require human judgment:
Which assumptions drive the | Requires knowing the business from business | the inside
Whether a forecast is credible to | Requires knowing this specific investors | investor's frame
When to challenge the CEO's vision Requires relationship and context in the model
How to defend a number under | Requires owning the reasoning, not pressure | just the output
Identifying what an acquirer will Requires M&A pattern recognition focus on
Specific AI Tools That Are Useful for Finance Work Right Now
This is not a sponsored list. These are tools that are actually useful based on direct use:
Claude and ChatGPT: For narrative generation, formula logic, model structure review, and diligence question preparation. The quality of output depends almost entirely on the quality of the prompt. Perplexity: For market research and comparable data. Better than a standard search for structured financial queries because it aggregates and cites sources.
Notion AI / Coda AI: For turning model commentary and board update drafts into polished documents faster. Useful for the reporting cycle. Excel Copilot (Microsoft 365): For formula generation and data cleaning directly in the spreadsheet environment. Still early but improving quickly.
The analyst who has built a personal library of prompts that work is operating at a structurally different level than one who is prompting from scratch each time. Treat prompt development as a skill worth investing in.
The Question Worth Asking
The most useful framing for a financial analyst or founder thinking about AI is not "will AI replace me?" It is: "What could I do if I had 40 extra hours this week?"
Because that is roughly what the mechanical automation of model tasks creates. The question is whether those hours go toward more analysis, more investor preparation, more business understanding, or back into operational noise.
The analysts and finance leaders who use this time to go deeper on the judgment work are the ones who become harder to replace. The ones who use AI to do the same work faster without going deeper are only delaying the question.
Frequently Asked Questions
Can AI build a financial model from scratch?
AI can generate a template financial model structure from a description of the business. The quality degrades quickly on the assumptions, which require actual business knowledge. The better use case is AI building the mechanical structure while the analyst defines the assumptions and logic.
Is it safe to share financial model data with AI tools?
Enterprise versions of most AI tools (Claude, ChatGPT Teams/Enterprise, Microsoft Copilot) offer data privacy protections that make them appropriate for confidential financial work. Free consumer versions of these tools typically train on inputs. Check the data policy before sharing sensitive information.
How long does it take to get good at prompting AI for finance work?
A few weeks of consistent practice produces significant improvement. The most effective prompts are specific, provide context, define the output format, and give an example of what good looks like. Building a personal prompt library as you go compounds value over time.
Will AI replace financial analysts?
The mechanical parts of the role are already being automated. The judgment parts are becoming more valuable because there is more capacity to act on good judgment when the mechanical work is handled faster. Analysts who build AI into their workflow are better positioned than those who do not. Analysts who only do mechanical work and do not develop judgment are at higher risk.
Get the complete guide with all 16 chapters, exercises, and model templates.
Get Raise Ready - $9.99