ABOUT 2 MONTHS AGO • 3 MIN READ

Claude released Opus 5. This is how Finance and FP&A teams should use it.

profile

AI for Finance

Practical Applications of AI in Finance, Python and machine learning for FP&A

Hi Reader ,

Every few months a new model drops and the internet tells you it "changes everything." Most of the time, it doesn't, not for the actual work Finance and FP&A teams do every day.

But I believe the new generation of Claude 5 Models: Fable 5, Sonnet 5 and now Opus 5 are a bit different.

I spent the last few days testing Opus 5 against real financial modeling, scenario planning, and reporting workflows, and I want to walk you through exactly what I found, including the prompts I used, so you can run the same experiments yourself.

Attached to this email in the bottom is my full Claude Opus 5 for Finance & FP&A Guide — every prompt, dataset, and experiment from this newsletter, plus a curated list of the best resources for learning Claude and a 6-week adoption roadmap built specifically for finance teams.

Let's get into it.

Idea #1: Build an entire interactive simulator — not a static model

The single most impressive thing I got Opus 5 to do was build a fully interactive P&L simulator as a working web app — sliders for revenue growth, margins, headcount, opex — that recalculates a full 12-24 month P&L instantly and lets you compare scenarios side by side. Not a description of a model. An actual working one.

Here's the prompt I used:

"Build me an interactive P&L simulator as a single-page web app (HTML/JS, charts included) for scenario planning. It should let the user adjust key drivers — revenue growth by segment, gross margin %, headcount and comp, opex categories, one-time items — via sliders, and instantly recalculate a full P&L across a 12-24 month horizon, with a trendline chart and a toggle to compare 2-3 saved scenarios. Include sensitivity indicators showing which driver moves EBITDA the most. State your assumptions before writing code, then validate your calculations tie out before showing me."

A quick aside, because this comes up every time I show this to someone: once you've built something like this, the natural next question is "okay, but where does this live?" If you ever want to host a simulator like this on your own website — to showcase it to a client, embed it in a proposal, or just make it feel like a real product instead of a one-off file — Webflow is one of the best options out there for getting something like this live quickly, without needing a dev team to stand up hosting and infrastructure just to share a tool. Worth a look if "cool internal demo" is something you'd rather turn into "polished client-facing asset."

You can even connect your agents directly to Webflow. With Webflow's MCP support, tools like Claude can prototype and publish pages straight into your site — while staying on-brand and governed, not just a one-off export you have to rebuild by hand. Handy if you want the simulator you just built to go from "prompt" to "live client-facing page" without a manual handoff in between.


Idea #2: Create a live command center, not a static dashboard

Beyond one-off models, Opus 5 is strong enough at long-horizon agentic work that you can ask it to build something closer to a real dashboard — a command center that pulls in data and updates as your assumptions change, rather than a snapshot you have to manually refresh every week.

One example I did was this one:

Prompt is in the Excel guide below.

Idea #3: Visualize data in ways a slide deck never could

This is where Opus 5's stronger visual generation really shows. Instead of a static waterfall chart, I had it build an interactive cash conversion cycle explainer — a visual you can actually step through and adjust live, styled more like a boardroom exhibit than a textbook diagram.

What impressed me most, overall

  • Polished decks and reports built from a rough, half-formed brief — not a generic template, an actually structured output.
  • Complex data analysis and financial models, worked through end to end, not just described in prose.
  • Research across dense material — filings, reports, long documents — that comes back genuinely ready to use, not a summary you still have to rework yourself.

Get the full Opus 5 for Finance and FP&A Guide

Everything above — plus every prompt, the datasets I used to test them, and a 6-week adoption roadmap for rolling this out across a Finance or FP&A team — is in the attached guide.

Claude Opus 5 for Finance Guide.xlsx

If you try any of these, hit reply and let me know what you build. I read every one.

— Christian Martinez

AI for Finance

Practical Applications of AI in Finance, Python and machine learning for FP&A