Most AI training sold to deal teams is a demo with a lunch. Someone shows a model summarizing a public 10-K, the room nods, and on Monday everyone goes back to the way they worked on Friday. The firm paid for a day and got a vocabulary.
A workshop is worth running only if the team leaves with things that work on their material. Here is what that looks like, and how to check for it before you book one.
It is built on your documents, not the vendor’s
If the facilitator arrives with their own examples, the day is a lecture. The useful version starts with a call a week before, where you pick the two or three workflows that cost the most hours and send over the sample documents: the CIM you drafted last month, the diligence request list you send every time, the weekly pipeline update you rebuild by hand.
The reason is simple. A prompt that works on a public filing tells you nothing about whether it works on your firm’s half-finished Word document with tracked changes and a table pasted from Excel. Your material is where it either holds up or falls apart, and you want to find out in the room, not alone the next week.
Everyone builds something
The right room is six to twelve people: the associates and analysts who actually do the work, plus one senior person who decides what gets adopted. Each person takes a real task from their week and leaves with the prompt, template, or checklist that makes it faster. Not a slide about it. The thing itself, tested on a second and third example until it stops surprising them.
Watch for the opposite pattern. If the agenda has more than fifteen minutes of “what is a large language model,” the day is about the facilitator’s comfort, not your team’s output.
One small tool exists by the end
Prompts and templates are the floor. The ceiling of a single day is one small working tool, scoped in advance: a check that compares a draft against the term sheet and lists the mismatches, a first-draft generator for the recurring memo, a data pull that turns a messy export into a clean table. Something that runs on a live example by the end of the afternoon.
This is where a workshop earns its fee. The tool is useful on its own, and it also answers the question the partners actually have: is there a bigger build worth doing? Sometimes the honest answer is no, and a good facilitator says so. Often the answer is yes, and now it can be scoped on evidence instead of a pitch.
Nothing new has to be bought
A workshop should be training on the tools the firm already has. Most advisory teams already pay for a team plan on Claude or ChatGPT. The work is making those tools do real work on real documents, inside the firm’s own accounts, so nothing confidential is uploaded anywhere new. If a vendor’s workshop requires their platform, it is a sales demo with a longer runtime.
The checklist
Before you book, ask:
- Will you work on our documents, inside our accounts?
- Does every attendee build something, or watch?
- What is the one tool we will have at the end, and how is it scoped?
- Do you tell us if there is nothing bigger worth building?
- Do we need to buy anything first?
If the answers are yes, yes, a specific thing, yes, and no, you will get your day back several times over in the first month. If not, keep the day and the money.
This is how we run ours. The format is described on the AI workshop page, and the first call is free.