// in this post
If you're reading this, you've probably already decided AI training or consulting is worth paying for. This is the part that matters more: choosing well. A bad engagement doesn't just waste the budget, it teaches your team that "AI initiatives" are something that happens to them rather than something that works, which makes the next attempt harder too.
This is written to be useful even if you end up hiring someone else.
The stage question, first
Before choosing a provider, check whether you need one yet. If what you actually need is individual skill-building on a handful of common tasks, that's often available for free, and we've mapped what's actually free in the UK right now rather than the "free trial, paid unlock" versions that aren't. Paid help tends to earn its cost at a different stage: when the problem is organisational rather than individual, how work is structured, who owns which decisions, whether the benefit spreads past the people who sat in the room.
What to ask before you hire anyone
What would you do before recommending anything? A provider worth paying scopes your actual situation first, your workflows, your tools, what's already been tried, before proposing a plan. If the answer to "what would we do" is the same regardless of what you've just told them about your business, that's a package, not a service.
How do you measure whether this worked? "People attended" and "people were satisfied" are activity measures. A good answer names an outcome tied to your actual work: a task getting faster, an error rate dropping, a specific workflow changing. If the measurement is about the training itself rather than what happens after it, the provider hasn't thought past the workshop.
What happens when the engagement ends? This is the question that separates a good provider from one building a dependency. A good answer describes what your team can now do without them: a documented workflow, a policy your people own, a skill that's actually transferred. If the honest answer is "you'll probably need us again," that's worth knowing before you start, not after.
Does this cover how AI gets used safely, or only how it gets used? Tool features without any governance, data handling, or acceptable-use content is training for a problem you don't fully have yet. What it takes to build AI capability goes into why training on its own, without the organisational design around it, tends not to produce lasting change. A provider who's only ever pitching the training half of that is selling you half the answer.
The red flags, plainly
A fixed package pitched before they've asked a single question about your business. No smaller first step, just the whole programme or nothing, with no way to test the relationship before committing the full budget. Training that's entirely feature demonstrations (prompting tips, button tours) with nothing about how the work or the decisions around it actually change. Vague outcomes: "AI-powered transformation" rather than a specific thing that gets better. And the absence of the ending question above, a provider who hasn't thought about their own exit hasn't thought about your independence.
What a good engagement structure looks like
A scoped, fixed-cost first step that produces something concrete on its own, not just a relationship-building exercise before the real (and much larger) proposal. A clear view, before you commit further, of what the rest would involve and what it would cost, even if the exact figure depends on what that first step finds. Training that sits inside a plan for how work changes, not instead of one. And a defined point where the engagement ends and your team owns what's left, whether or not you ever hire them again.
This is a normal, well-established structure among providers who are confident their work holds up: a small, fixed-scope diagnostic first, then a proposal for the rest, scoped to what's actually needed. If a provider can't describe their own version of that first step, ask why.
Where to go from here
If you're ready to see what this looks like in practice, our own Discovery is built exactly this way: a fixed two-week, fixed-scope first step, before anything larger is proposed. Get in touch if you'd rather talk it through before deciding anything.
FAQ
What questions should I ask an AI training provider before hiring them?
Ask what they do before recommending anything (a proper provider scopes your actual situation first, not a generic package). Ask how they measure success beyond attendance or satisfaction scores. Ask what happens to the skills and systems after the engagement ends. Ask whether they cover governance and safe use, not just tool features. If any answer is vague, that's the answer.
What are red flags when choosing an AI consultant or trainer?
A fixed package pitched before they've asked about your workflows. No first, smaller step, just the full programme or nothing. Training that's entirely about tool features (prompting, buttons, demos) with no mention of how work or decisions actually change. No plan for what happens when the engagement ends, which usually means the plan is for you to need them again.
Do I actually need a paid AI training provider, or can I do this myself?
If your organisation is small, the need is narrow, and someone internal has the time and standing to drive it, you may not. The free-training route can work well for individual skill-building. Where paid help tends to earn its cost is exactly where free courses stop: redesigning how your organisation's work actually happens, not just teaching individuals to use a tool.
Is AI training the same as AI adoption consulting?
No, and the difference matters. Training builds individual tool skill. Adoption consulting addresses the organisational design around it, workflows, quality checks, decision rights, so the individual skill actually compounds into results. A lot of what's sold as one is really the other, worth clarifying before you sign anything.