---
"title": "What should an AI training program for a small firm include?"
"description": "The parts of an AI training program that actually change how a
  small firm works, from a shared baseline and role-based sessions to practice
  on real work, standards, support and measurement."
"section": "AI Training"
"sidebarTitle": "AI training program basics"
"published": "2026-10-20T13:12:48.092Z"
"updated": "2026-10-20T13:12:48.092Z"
---

## What should an AI training program for a small firm include?

It depends on where your team is starting. If most people have barely used AI, start with a short shared baseline session, then move quickly to practice on each team's real work. If some people already use it daily, skip the basics and focus on role-specific workflows, shared standards and safe data handling. Either way, a program that works includes five things: a baseline, role-based practice on real tasks, written standards, follow-up support for a few weeks, and a simple way to measure whether the work changed.

## Who this is for

Founders, COOs and heads of people at firms of roughly 10 to 200 people who want their team using AI well, and who don't have a learning and development department to design a program for them.

## The short version

- Train on the work people actually do, not on generic examples.
- Keep sessions short and spread out, with practice in between.
- Write down what good looks like, so the training outlasts the session.

## Why small firms need a different approach

Large companies can run long courses, build internal academies and accept that adoption will take a year. A small firm usually can't. Every hour in training is an hour away from clients, and there is rarely someone whose full-time job is to follow up afterward.

That changes the design. A small firm's program should be compact, built around a handful of high-value tasks, and designed so that people can keep improving on their own once the formal sessions end.

## The five parts of a program that works

**1. A shared baseline**

Everyone, from partners to support staff, should leave the first session understanding the same few things:

- What the firm's approved tools are and what they're good and bad at.
- What data can and can't go into them, under your [acceptable use policy](/docs/write-an-ai-acceptable-use-policy).
- How to give clear context and instructions, and how to check output before relying on it.
- Where to go with questions.

One practical hour is often enough for this. The goal is confidence and safe habits, not expertise.

**2. Role-based practice on real work**

This is where most of the value sits. Group people by the work they do, then practice on the tasks that fill their week. For example:

- **Client-facing staff:** first drafts of emails, meeting summaries, proposals and follow-ups.
- **Analysts:** research synthesis, document review and turning notes into structured output.
- **Operations and finance:** recurring reports, reconciliations, process documentation and internal requests.
- **Leaders:** preparing for meetings, reviewing drafts faster and deciding where AI should and shouldn't be used.

Bring real, approved material to the session. People learn far more from improving their own Tuesday report than from a demo.

**3. Written standards**

Training fades unless it is captured somewhere. Keep a short, living set of materials:

- Shared prompts or configured assistants for the firm's most common tasks.
- Examples of good output for each.
- A checklist for reviewing AI-assisted work before it goes to a client.

Store them where people already work, and name an owner who keeps them current.

**4. Follow-up support**

Most of the learning happens in the weeks after a session, when people try things on live work and get stuck. Build in support for that period:

- A short check-in a week or two after each session.
- A channel or contact for questions.
- A few internal champions who are a step ahead and can help colleagues.

This is the main difference between a one-off workshop and a program. Our page on [workshops vs ongoing coaching](/docs/ai-workshop-vs-ongoing-coaching) covers that choice in more detail.

**5. A simple measure of change**

Decide before you start how you'll know it worked. Usage numbers alone can mislead. Better signals are changes in the work: time to produce a first draft, turnaround on recurring tasks, number of revision rounds, or how many people use the shared prompts each week. See [how to measure whether AI training worked](/docs/measure-ai-training-results).

## A sample structure for the first two months

| When | What | Who |
|---|---|---|
| Week 1 | Baseline session and tool setup | Everyone |
| Weeks 2 to 3 | Role-based sessions on real tasks | Each team |
| Week 4 | Check-in, collect what worked, publish first standards | Champions and team leads |
| Weeks 5 to 6 | Second round on harder or more frequent tasks | Each team |
| Week 8 | Measure against the baseline and decide next steps | Leadership |

Adjust the pace to your calendar. The order matters more than the dates: baseline, practice, capture, practice again, measure.

## What to leave out

- **Long tool tours.** Feature-by-feature walkthroughs are forgotten quickly. Teach features when a task needs them.
- **Theory for its own sake.** Most people don't need to know how models are trained to use them well at work.
- **Mandatory hours with no follow-up.** Attendance without practice changes very little.

## Common mistakes

- **Training leaders last.** If partners and managers don't use AI themselves, they won't encourage or review AI-assisted work well. Include them early.
- **Using generic examples.** If the exercises don't look like the firm's work, people won't connect them to their week.
- **No policy before training.** People hesitate when they don't know what's allowed. Settle the basics first.
- **Treating it as a one-time event.** Tools and models change. Plan a light refresh every few months.
- **Measuring logins instead of work.** Check whether tasks got faster or better, not just whether accounts were used.

## When you do not need us

If you have someone with time to design sessions around each team's real work, keep standards current and follow up for a few weeks, you can run this program yourself with the structure above. Outside help is most useful when nobody has that time, when the team has stalled after an earlier attempt, or when you want sessions built around your workflows from day one.

## How Ruby helps

- **AI training:** hands-on sessions for each role, built on your real work, plus the follow-up that turns a session into a habit.
- **AI strategy:** choosing which tasks to focus on first and setting up standards, policy and measurement.
- **AI engineering:** building assistants or workflows when training shows a task deserves more than a shared prompt.

A good first step is [one practical hour of AI training](https://rubyadvisory.com/book-a-call) for your team.

## Related pages

- [How to train non-technical employees to use AI](/docs/train-non-technical-teams-on-ai)
- [Why your team isn't using the AI tools you paid for](/docs/why-ai-adoption-stalls)
- [How to measure whether AI training worked](/docs/measure-ai-training-results)
- [One-off AI workshop vs ongoing AI coaching](/docs/ai-workshop-vs-ongoing-coaching)