AI Training Case Study: How We Upskilled a Finance Team on AI

The most common question we get about AI training is not "which tool?" but "will it actually stick?". Most corporate AI training does not, because it is a one-off demo with no practice attached. This is how we ran a programme that did stick, for a finance team that needed to move faster on the work that was eating their week.

The starting point

The client is the finance team of a fast-growing US healthcare company. Capable people, plenty of AI curiosity, but the usage was ad hoc: a bit of ChatGPT here, a Copilot prompt there, and no shared method. Meanwhile the work that consumed their time was exactly the work AI is good at supporting: reviewing board packs, running the same checks that took the best part of a day, writing variance commentary, and pulling together go-to-market and market research for finance decisions.

The goal was not "teach them ChatGPT". It was to turn scattered experimentation into a reliable, governed capability the whole team shared, tied to their real tasks.

Why modular, not a single day

You can run this as one to three intensive days or as shorter modules. For a junior-to-mid team we recommend the modular route: four live sessions of around three hours each. The reason is simple. People retain a skill they practise, not a skill they watch. Spacing the sessions lets participants apply each module to their own work in the days between, hit real friction, and bring genuine questions back to the next session. A single long day feels productive and fades within a week. The spaced format is what turns a workshop into a change in behaviour.

The programme, module by module

Here is how the finance programme runs. Every module is hands-on from the first hour.

Module 1: Prompting that works

Briefing AI the way you brief your best analyst. We teach the RACEF framework (Role, Action, Context, Examples, Format) and an output contract that makes answers finance-grade, every figure tied back to source, and how to iterate toward a reliable result. People are writing real prompts against real tasks inside the first hour.

Module 2: The tools and what each is for

Claude and ChatGPT side by side, plus Projects and Custom GPTs. When to reach for which, and how to analyse, summarise and compare large document sets securely. Finance teams do not need every tool; they need to know which tool earns its place for which job.

Module 3: Your finance workflows, built live

This is where it gets specific. Variance analysis and commentary, reporting and board packs, models, headcount, and go-to-market finance. Participants build reusable prompts and templates against their own tasks, so they leave the session with work already done, not just notes on how they might do it later.

Module 4: Reusable assistants, automation and governance

Building Claude Projects and Custom GPTs, a shared team knowledge base, and automating routine tasks into simple workflows. This module also includes a proper session on data privacy, confidential information and AI governance for a regulated finance function, because capability without guardrails is a liability. Everyone finishes with a prompt library, reusable assistants and templates built around their real work, and a clear map of where AI fits day to day.

What changed for the team

The point of the programme was capability, not a vanity metric, but the shape of the change was clear. Board-pack review, previously a careful manual read, became a structured first pass the team could trust and refine. Routine checks that took the best part of a day were compressed dramatically once the prompts and assistants were built for them. Variance commentary moved from a blank page to a reviewed first draft. And go-to-market and market research that used to mean hours of tab-hopping became a repeatable, sourced workflow.

Beyond finance: legal and other teams

Finance is where most of our training lands, but the same spine runs for other document-heavy, judgement-led functions. The legal programme follows the identical structure retuned to legal workflows: contract analysis and comparison, large-document review, drafting support, and the confidentiality, privilege and governance considerations specific to legal work, with the case studies and exercises tailored to that team. If your function reads, reviews, drafts and decides, the method fits.

If you want this for your team, our AI workshops for teams run in person or virtually, the self-paced AI for finance course covers the same frameworks for individuals, and AI consulting is there when you want us to build the workflows alongside the training.

Frequently asked questions