.avif)
.png)
Where It Goes Wrong
One all-company session for everyone at once. Generic examples nobody recognizes. A product demonstration in place of practice. Nothing written down afterward. Nobody owning it once the trainer has left.
AI Aviators is Dragon Horse's hands-on training program. Your people practice on their own live work, in the tools you already pay for, until they can do it without us. Not a webinar and not a certificate. Capability that stays.
Watching someone use AI well is not the same as being able to. The distance between the two is a few weeks of supervised practice on real work, and almost nobody schedules it.
Most AI training is a demonstration. Someone shows the tool, everyone nods, and three weeks later usage has drifted back to where it was. Aviators sessions run the other way round: your people bring the work they were going to do that week, and they do not leave until it is finished.
Dragon Horse calls this being AI Aviators. AI is an instrument, and instruments do not fly themselves. Our job is to make your people good at flying yours.
Aviators is delivered as working sessions rather than lectures, and we run them by function. A marketing team and a finance team need almost nothing in common from AI, so the useful example for one is dead weight for the other.
Every program moves through the same four stages, run against your team's real tasks rather than a generic curriculum. Which of your tasks are worth practicing on. Supervised repetition until people are fluent. A written standard so everyone works the same way. And a named owner who keeps it running once we have gone.
People leave a session enthusiastic and return to a full inbox. Without practice on their own work and a standard to follow, the new habit loses to the old one inside a fortnight, and the tool takes the blame.
One all-company session for everyone at once. Generic examples nobody recognizes. A product demonstration in place of practice. Nothing written down afterward. Nobody owning it once the trainer has left.
Small groups by function. Their own live tasks. Enough supervised repetitions that it stops feeling risky. A one-page standard they helped write. And a named owner who runs the next session without us.
We do not bring a curriculum or a platform. We bring the format; your team brings the work it was going to do that week anyway.
We pick the tasks worth training on and name the ones that should stay entirely human. Practicing on the wrong task is how a program loses credibility in week one.
Small groups, live work, your own tools. We sit with people while they do the task, correct the approach in the moment, and keep going until it is repeatable without us.
The prompts, checks and review points that worked become a one-page standard your team authored and will therefore follow, covering what data is allowed and what always gets human sign-off.
A named owner inside your business runs the next session. We return a quarter later and measure what people are actually still doing, rather than what they said they would.
A week of your team's real work reviewed function by function, with the tasks worth training on picked, sized and put in order before anyone books a session.
Sessions run in whatever you already pay for, whether that is Copilot, ChatGPT, Gemini or Claude. We sell no software and take no commission from any vendor.
How to ask, how to check the answer, and when to stop and do it yourself. Verification is the half most training skips and the half that decides whether the output ever gets used.
Everything that worked in the room written up as short playbooks in your team's own language, with your own examples, held by you rather than by us.
One or two people per function taught to run the next session on their own, so the program carries on after the engagement ends rather than needing us booked again.
We come back a quarter later, look at what is genuinely being used, and repair the parts that lapsed. Agreed at the start of the engagement, not sold to you afterwards.
The only honest measure of training is what your team can still do a quarter after the trainer has gone.
By the end of a session people have finished real work with AI, rather than agreeing in principle to try it later.
One written way of working per function, so quality stops depending on which individual happened to attend.
People are taught what never goes into a model and what always gets reviewed as part of the task itself, rather than in a separate policy nobody reads.
Playbooks and trained champions held by your team, so the capability does not walk out of the building with a consultant.
A finance team and a marketing team need almost nothing in common from AI. Sessions are sized to a single function so the examples are ones people recognize and the practice is on work they are accountable for.
We sell no platform and take no commission, so nothing in the training is shaped by what we would like you to buy. Dragon Horse runs AI across a working agency every day, which means the practice comes from doing the work rather than from reading about it.
Where a few people are getting a lot out of it, most are not, and nobody has written down what the good ones are doing differently.
Where licenses were bought, access was switched on, and everyone was left to work it out in their own time.
Where client confidentiality or a regulator means people need to be taught the guardrails at the same moment they are taught the technique.
Consulting decides what is worth doing and sets the policy around it. Aviators makes your people able to do it. Most clients run a short consulting engagement first, then train the teams who will actually be doing the work.
Usually yes, and that is the most common starting point. Having a license is not the same as being good with it, and the distance between an average user and a fluent one is very large.
Frequently, and we name those tasks before the training rather than after. Some work should stay human because the judgment is the product, and some because the risk is not worth the saving.
The people doing the work, not only the people responsible for it. Sessions run by function, and eight to twelve people is the size where everyone still gets hands on their own task.
Tell us what your team is trying to do with AI and we will scope a training program against your actual work. You do not need to buy any software to start.
Not ready to talk? Get a free marketing audit →