AI Training for Teams
Workshops run on your team’s actual work — their quotes, their emails, their reports. Not generic prompt tips that everyone enjoys and nobody uses.
Everyone Enjoyed the Workshop. Nobody Changed Their Monday.
The standard failure mode of AI training, and it is almost entirely a content problem.
The generic AI workshop follows a familiar script. Someone explains what a large language model is. There is a slide about how it predicts the next token. Everyone writes a haiku about the company. There is a section on prompt engineering with acronyms in it. People leave entertained, mildly optimistic, and completely unchanged, because at no point did anyone touch a task they will actually do next week.
We build the entire session on your team’s real artefacts. Before the workshop we collect the genuine article — the quotes they write, the emails they answer, the reports they assemble, the enquiries they field — and the exercises run on those. This changes the day completely. People are not imagining how AI might apply to their job; they are watching it succeed and fail on the exact document that was annoying them yesterday.
It also produces the single most valuable thirty seconds of the workshop: the moment someone watches the model produce an answer about their own material that is fluent, confident and wrong. You cannot manufacture that with an invented example, because nobody in the room has the expertise to catch the error. On their own work, everybody catches it instantly — and that experience calibrates people faster than any amount of warning about hallucinations.
And we split the audience, because a CFO and an admin team do not need the same session. Executives need ninety minutes on what to believe, what to fund, and what to ask a vendor. Frontline staff need a half day on their tasks, where the hours actually are. Managers need to spot when AI is creating rework rather than removing it. Technical staff need evaluation and failure modes. Putting all of them in one room and calling it AI training wastes everyone’s day equally.
Sessions by Audience
Different people need different things. These are separate sessions, not one deck with slides skipped.
Executive Briefing
90 minutes · leadership team
What to believe, what to fund, and what to ask a vendor. No hands-on lab, no token-prediction slide. Aimed squarely at the decisions your executives are actually being asked to make this quarter.
- What the technology can and cannot do right now
- How to read a vendor AI claim critically
- Where the money goes and where it disappears
- The board answer, and how to defend it
Role-Based Team Workshop
Half day · up to ~12 people
The core session. Built on your team’s real work, capped at twelve so everyone gets hands-on time. Separate versions for finance, operations, sales, customer service and admin.
- Exercises built on your actual documents
- Where AI wins big — and where it fails plausibly
- Each person leaves with 2–3 committed tasks
- Tailored per function, not one generic deck
Safe-Use Training
60 minutes · repeatable for new starters
What may and may not go into which tools, in plain English with examples from your business. We hand you the materials so you can run it yourself for every new starter.
- Built on your acceptable-use policy
- Real examples, not abstract categories
- Materials handed over for internal reuse
- Designed to be repeated, not to be a one-off
Technical Session
Half day · developers & system owners
For the people who will build or maintain. Evaluation harnesses, integration patterns, failure modes, cost control — the things that separate a demo from something that survives production.
- Building evaluation sets from your own data
- Integration patterns and their trade-offs
- Failure modes and how to detect them early
- Cost and latency control in practice
Judgement, Not Tricks
The prompting techniques that mattered two years ago mostly do not now. What does not go out of date is knowing when to trust the thing.
Which tasks are worth handing over
The most useful skill and the least taught. Some tasks AI does better than your best person. Some it does worse than your worst. Telling them apart quickly is worth more than any prompting technique.
Spotting confidently wrong answers
The failure mode that actually costs money. Not obvious nonsense — fluent, plausible, subtly incorrect output that a busy person accepts. We drill this on your own material, where people can catch it.
What must never leave the business
Customer personal information, health information, anything under NDA, anything you would not email a competitor. Concrete categories with real examples, not a warning slide.
When to stop fighting the tool
Knowing when you are three prompts deep on something you could have done manually in four minutes. Sunk-cost recognition is a genuine, teachable skill and almost nobody teaches it.
Verifying without re-doing the work
If checking the output takes as long as producing it, you have gained nothing. Practical verification strategies that are cheaper than the task itself.
Reading a vendor claim
Every product you use has an AI feature now. How to tell the genuinely useful from the marketing veneer — and what question makes a vendor uncomfortable in the right way.
How Training Works
Remote across Australia in your local hours, in person around Melbourne. The follow-up is the part most providers skip.
Collect Real Work
Before the session we gather genuine artefacts from the team — real quotes, emails, reports, enquiries. The exercises get built on those, which is the whole reason the training sticks.
Run the Session
Hands-on, capped at about twelve people so everyone gets time on the tools. Wins and failures both demonstrated, on material the room knows well enough to judge.
Commit to Tasks
Everyone leaves having written down two or three specific tasks they will use AI for. Not intentions — named tasks, in their own words, that they will do again next week.
Follow Up & Report Back
Several weeks later we find out what stuck and what did not. The "why not" answers are usually the most valuable output, because they surface the real blockers. You get that feedback straight.
Related Services
Training works best with a policy behind it and something real to apply it to.
AI Governance Consulting
Safe-use training needs a policy to teach. If you do not have one your staff can understand, start here.
GovernanceAI Implementation
A system nobody understands gets worked around. Training is part of every implementation we deliver.
ImplementationFractional AI Officer
Building judgement internally is the fastest way to reduce how much external AI leadership you need to buy.
Fractional leadershipFrequently Asked Questions
What managers ask before booking AI training for their teams.
No, and we would argue prompt engineering as a standalone discipline is already fading. The tricks that mattered in 2023 — elaborate role-play preambles, threatening the model, tipping it — mostly do not matter now, and the ones that do will not matter next year as models get better at inferring intent. What does not go out of date is judgement: knowing which tasks are worth handing to AI at all, recognising when an answer is confidently wrong, understanding what should never leave your business, and knowing when to stop fighting a tool and just do the work. We teach that. The prompting techniques come along as a by-product, because you cannot practise judgement without touching the tools.
Yes, and it is the entire reason the training sticks. Generic AI workshops fail in a predictable way: everyone enjoys the session, nobody changes their Monday. So before the workshop we collect real artefacts from the team — actual quotes, actual emails, actual reports, actual customer enquiries — and build the exercises on those. People leave having used AI on a task they will do again next week, having seen where it worked and where it produced something plausible and wrong on their own material. That last experience is the most valuable thirty seconds of the day, and you cannot manufacture it with a made-up example.
Definitely not the same training, and that is where most programmes go wrong. Executives need ninety minutes on what to believe, what to fund and what questions to ask a vendor — not a hands-on lab. Frontline and admin staff need a half day on their actual tasks, because that is where the hours are and where the enthusiasm is highest. Managers need to know how to evaluate whether the AI in their team is helping or quietly creating rework. Technical staff need something entirely different again — evaluation, integration and failure modes. We run these as separate sessions with different content and different lengths. Putting your CFO and your admin team in the same AI workshop wastes both of their time.
By taking both seriously, because both are correct about something. The sceptics have usually watched AI confidently produce nonsense and concluded it cannot be trusted — and they are right, for the tasks they tried. The enthusiasts have had genuine wins and want to automate everything — and they are right that the wins are real. The training deliberately shows both: exercises where AI is dramatically better than the manual process, and exercises where it fails in a subtle, plausible way that a busy person would miss. Sceptics tend to leave with two tasks they will actually use it for. Enthusiasts leave with a much better sense of where it will burn them. That convergence is the actual outcome we are aiming at.
Executive briefing is 90 minutes. Role-based team workshops are a half day, capped at about 12 people because beyond that nobody gets hands-on time. Safe-use training is a 60-minute session designed to be repeatable for new starters, and we hand you the materials so you can run it yourself. Everything runs remotely across Australia and we schedule in your local business hours — including genuinely early starts for Perth teams rather than pretending AWST does not exist. We also run in-person sessions around Melbourne, where we are based, and will travel interstate for multi-session programmes where it makes sense.
We build the measurement in, because "everyone enjoyed it" is not an outcome. Each participant leaves the workshop having committed to two or three specific tasks they will use AI for, written down, in their own words. We follow up several weeks later to find out what stuck, what did not, and why — and the "why not" answers are frequently the most useful thing the whole exercise produces, because they surface the real blockers: a tool nobody has licensed, a policy nobody understands, a process that turns out to be different from how it was described. That feedback goes back to you. If the honest answer is that nothing changed, you should know that, and you should probably not book the follow-up session.
Train Them on Their Own Work
Tell us what your team actually does and we will tell you which session fits — and whether training is the right spend at all.