Designing an AI Training Program by Role

1. One Generic Course Teaches No One Their Job

The default move is to book one AI training, put the whole firm in a room, and call it covered. A vendor walks through prompts and features for ninety minutes, everyone nods, and nothing changes the following Monday.

It fails because a firm is not one audience. A managing partner who needs to decide what is safe and where the firm should bet has almost nothing in common, training-wise, with a second-year associate who will run the tool eight times a day. Teach to the middle and you bore the partner, lose the analyst, and leave both without the one thing their own job required.

The fix is not more training. It is training designed by role, so each seat learns the part of AI that touches the work they actually own. That is more effort to build and far more likely to land, because people change what they do when the lesson is about their Tuesday, not about AI in general.

2. Train the Workflow, Not the Tool

The most common mistake is teaching the tool: here are the features, here is the prompt box, here are some tips. People leave able to describe the tool and unable to use it on Monday, because the work was never in the room.

Train the workflow instead. Not "here is how Claude works" but "here is how you screen this week's CIM against our box, draft the first cut of this memo in our format, or spread this borrower." The lesson is built around a job the person already does, so the skill has somewhere to live the moment they walk out. A feature is forgettable. A workflow they ran on real work this morning is not.

This is also why role matters so much. The workflow is the unit of training, and each role owns different workflows. Design from the work each seat does and the curriculum almost writes itself, which is the same discipline behind the whole approach to training a PE team.

3. The Roles and What Each Owns

A firm has six broad roles for the purpose of AI training. Each one learns a different thing, to a different depth, in a different format. The table is the program in miniature.

The rest of this guide takes the rows in turn. Notice the depth is inverted from the org chart: the most junior seats get the longest, most hands-on training, and the most senior get the shortest. That is not a slight. It is who touches the work.

4. Analysts and Associates: the Heavy Users

Analysts and associates are where AI earns its keep, because they do the assembly work AI compresses: reading the same CIM everyone reads, spreading a borrower by hand, building the first draft of a memo, working through a data room. They will use the tool more than anyone else in the firm, so they need the deepest training by a wide margin.

Theirs is not a lecture. It is a hands-on intensive on live work, where they screen a real CIM, draft a real memo section, and spread a real set of financials, with a trainer in the room when it goes sideways. And it teaches the habit that matters most at this seat: verification. An analyst whose name is on the output has to know where the tool is strong, where it is confidently wrong, and how to check fast. That instinct is the whole skill, and it only forms through reps on real material.

This is the population an Associate Onboarding is built for, and the format that fits a deal team is a Deal Team Intensive on the firm's actual pipeline. Train this seat well and the rest of the firm sees the wins first here.

5. VPs and Principals: Review and Delegate

VPs and principals sit in the middle, and their AI skill is different in kind. They will not be the heaviest hands-on users. Their job is to review AI-assisted work and to delegate it well, which means they set the standard the juniors work to.

So they learn two things. First, how to read an AI-assisted memo or model with the right kind of suspicion: what to trust on sight, what to check every time, where the tool tends to be confidently wrong. A reviewer who cannot tell good AI-assisted work from plausible-looking nonsense is a hole in the firm's quality control. Second, how to delegate to a junior who is using the tool: what to ask for, what standard to hold, how to make sure the analyst is verifying rather than pasting. Get this seat right and it multiplies the analyst training. Get it wrong and a careful analyst's work gets waved through by a reviewer who never learned where to look.

The format is a working session on their own review and delegation, not a generic overview. They have less time than the analysts and a more specific need, which is exactly the gap a Partner and IC Review is shaped to close.

6. Partners and IC: Decide and Govern

Partners and the investment committee will not be daily users, and pretending otherwise wastes the scarcest time in the building. Their training is short and aimed at the decisions only they make.

They need three things. A grounded sense of what AI can and cannot actually do, so they neither dismiss it nor overtrust it. A view on where the firm should bet, because the strategy and the budget are theirs to set. And the governance call: what is safe to put through which tool, and what the firm will tell an examiner or an LP who asks. They also need to know how to read a memo they know was built with AI, because soon every memo will be, and the IC has to keep judging the argument rather than the polish.

The right format is a short, decision-focused Executive Briefing, ninety minutes, not a half-day workshop. The deeper version of what this seat owns, and how an IC stays sharp as AI-assisted memos become the norm, is in the AI-ready investment committee.

7. IR and Back Office: the Quiet Wins

The deal team gets the attention, but two other groups quietly save a lot of hours, and a role-based program is the only one that remembers them.

Investor relations. IR drafts LP updates, answers the same DDQ questions for the ninth time, and summarizes long documents under deadline. A tool that drafts from a library of past answers and your house language turns a day into an hour, with the firm-specific rule that a person reads every word before it reaches an LP. Their training is a focused workshop on those exact workflows, not a general session.

Operating partners. Their AI need is unusual, because the value is not only at the fund. It is inside the portfolio companies they work with, most of which have no technology bench. They learn where AI creates value in an operating business and how to carry a playbook in, which is a different skill from using the tool at the desk. The full version of that job is training portfolio company teams on AI.

Back-office functions round it out: finance, compliance, and administration all have document-heavy tasks that AI handles well, and each deserves a short, practical session aimed at its own work rather than a seat in the all-firm course they will not remember.

8. Sequencing the Program

Order matters as much as content, because the program needs momentum and a firm has limited attention to spend. Run the roles in the wrong order and the energy drains before the wins arrive.

Start with the analysts and associates. They use the tool most, they produce the visible wins fastest, and a deal team that is genuinely saving hours becomes the proof the rest of the firm can see. Then train the VPs and principals, so the review and delegation layer is ready to hold the standard as the juniors lean on the tool. Brief the partners and IC early in parallel, because they set the budget and the governance, and a short briefing up front buys the air cover the whole program needs. IR, operating partners, and back office follow as their workflows come into focus.

The principle is the same one that governs any rollout: narrow and deep beats broad and shallow, and a win that travels beats a launch that does not. One group trained well, saving real hours, pulls the next group in by demonstration. The reverse, a thin session for everyone at once, gives the whole firm permission to ignore it.

9. Making It Stick

A program is not the training. It is what happens in the weeks after, and most of that depends on two things the calendar invite does not contain.

The first is a champion: one person who owns the effort after the trainer leaves, fields the questions, fixes the prompts, and keeps the wins visible. Training without a champion fades by the second week, because the moment someone hits friction with no one to ask, they go back to the old way. The second is measurement. You do not need a dashboard, you need an honest read on whether each role is genuinely doing the work the new way now, without being reminded. The off switch is the simplest test: if you turned the tool off, would this team be upset. If yes, it stuck. If no, the session was theater.

So budget for both from the start. The champion question is large enough to be its own decision, covered in building an internal AI champion, and the honest read on whether any of it worked is measuring AI adoption. Tie the program to both, and the cost question answers itself: the real comparison is laid out in how much AI training costs.

10. Where to Start

Map your roles before you book anything. Write the three columns for your own firm: each seat, the workflow it owns, the format that fits its time. The moment you do, the single all-firm session stops looking like a plan and starts looking like a placeholder.

Then start with the seat that uses the tool most, train it on real work, and let that win pull the next group in. A program built this way is slower to schedule and far harder to waste, because every session is aimed at a job someone already does.

If you want the whole firm trained by role, one function at a time, with a champion and a way to measure it, that is exactly what a Guided Launch is built to run. The full menu of formats by seat, from the ninety-minute briefing to the deal-team intensive, is on the training page.