Hiring an AI Consultant vs Training Your Team vs Buying Software
An honest comparison of three ways UK SMEs can adopt AI, what actually drives the cost of each one, and how to work out which path fits your business.
Every business owner I speak to asks some version of the same question: “What’s the right way to do this AI thing?”
They’re drowning in options. Consultants promising transformation. Software vendors with slick demos. Internal teams eager to “learn AI” but unsure where to start. And the stakes feel high because they are. PwC’s 2026 CEO Survey found that 56% of companies report seeing no meaningful benefit from their AI investment. Most businesses are spending money on AI. Most aren’t getting results.
The right path depends on your situation. But the wrong path is almost always the same: half-committing to multiple approaches and doing none well.
Here’s how to think about each option honestly: what it costs, where it works, and where it falls apart.
Option 1: Buy Software Off-the-Shelf
What it looks like: Subscribe to ChatGPT Enterprise, Copilot, Jasper, or whatever SaaS tool promises to solve your problem. Give your team logins. Hope for the best.
What it costs: £20–£50 per user per month for most AI SaaS tools. For a 50-person company, that’s £12,000–£30,000 per year before you factor in the time cost of people figuring out what to do with it.
When it works: Your AI needs are genuinely standard. Drafting emails, summarising documents, basic content generation, meeting transcription. Your team is already comfortable trying new tools without hand-holding.
When it fails: Almost everything else. Generic tools solve generic problems. Your competitive advantage comes from doing things differently. Off-the-shelf AI does things the same way for everyone.
The real cost isn’t the subscription. It’s the wasted potential. The British Chambers of Commerce found that 54% of UK SMEs are now using AI, but only 1 in 10 are turning it into real productivity gains. Most of that gap is businesses buying tools they never properly deploy.
Gartner predicted 30% of generative AI projects would be abandoned after proof of concept by end of 2025, and that includes enterprise companies with dedicated teams. For SMEs without that support structure, the abandonment rate is almost certainly higher.
The honest verdict: Fine for simple, well-defined needs. Terrible as your entire AI strategy.
Option 2: Train Your Internal Team
What it looks like: Send people on courses. Buy learning subscriptions. Maybe hire someone with “AI” in their job title. Build internal capability over time.
What it costs: Training is priced per person and per day, so the total moves with how many people you send and for how long. A public course is the cheapest per head and the least specific to your business. A programme built around your own workflows costs more and is the only kind that reliably survives contact with the day job.
The invoice is the part you can see. For a 50-person firm sending ten people through a proper programme, the weeks those ten spend away from their work are the rest of the cost, and nobody puts them in a budget.
When it works: You have the luxury of time. You have a genuine learning culture, not just lip service to “upskilling.” Someone internal can translate training into real projects. And you’re prepared to invest in ongoing application, not just a one-off course.
When it fails: Training without immediate application. Learning science is clear on this: up to 90% of new skills are lost within a year if not applied in practice. The training industry calls it “scrap learning,” and it’s endemic.
I’ve seen teams complete expensive AI programmes, feel inspired for a fortnight, then go back to exactly how they worked before. Not because the training was bad. Because there was no bridge between learning and doing. No one had redesigned their workflow. No one had given them permission, or time, to actually change how they operated.
techUK found that lack of expertise is the number one barrier to AI adoption for UK businesses. Training sounds like the obvious fix. But training that doesn’t connect to real work isn’t a fix. It’s a cost.
The honest verdict: Essential for long-term capability. Ineffective without a plan to apply what’s learned immediately.
Option 3: Hire an AI Consultant
What it looks like: Bring in someone who’s done this before. Pay for their expertise, their mistakes-already-made, their pattern-matching from dozens of implementations.
What it costs: Consultants sell days. The total is days multiplied by a rate, and the rate moves with how senior the person actually doing the work is, which is not always the person who sold it to you.
Ask for the three numbers behind any proposal: how many people, for how many days, at what rate. Multiply them and see whether you get the total on the front page. If it does not reconcile, ask what the difference is, because there may well be a reasonable answer and you want to hear it before you sign rather than after.
What moves the day count is how much of your business has to be understood before anything can start, and whether the engagement ends with a document or with something running. A first engagement worth having usually runs to a few weeks of somebody’s time. Anything much shorter is a conversation, which is fine as long as you are not paying for it as though it were a project.
When it works: You need to move fast. The stakes are high enough that getting it wrong matters. You want someone accountable for results, not just recommendations. And you recognise that AI implementation is about organisational change as much as technology. McKinsey’s 2025 State of AI survey found that high-performing organisations are 2.8 times more likely to invest in both external expertise and internal capability building simultaneously.
When it fails: You hire badly. The UK AI consulting market has a genuine problem: too many generalists selling strategy decks and disappearing. No implementation. No accountability. No measurable outcome.
A good consultant should pay for themselves in the first project. If they can’t articulate exactly how they’ll deliver return within six months, in terms you can measure. Keep looking.
The honest verdict: The fastest path to real results if you choose well. The most expensive waste of money if you don’t.
The Option Nobody Talks About: Hybrid
Here’s what actually works for most businesses I see, and it’s backed by the data. McKinsey found that high performers are 2.8 times more likely to invest in both external expertise and internal capability. Not one or the other. Both.
Start with a consultant for the first 90 days. Get the foundation right. Identify the highest-value problem. Build one or two working systems. Establish what “good” looks like in your specific context.
Train your team on those specific systems. Not generic AI training. Training on the tools you’ve actually built, the processes you’ve actually changed, the workflows that now look different. This is how you beat scrap learning. People learn by doing real work, not sitting in classrooms.
Buy software only where it genuinely fits. Don’t start with the tool. Start with the problem, build the solution, then evaluate whether off-the-shelf software can handle part of it. Usually it can handle some. Rarely it can handle all. The consultant helps you see the difference.
The 90-day model maps to how change actually works: quick win, capability transfer, then expansion. It’s not revolutionary. It’s just practical.
How to Decide: Three Questions
1. What’s the cost of getting this wrong?
Low risk (experimenting, non-critical processes) → Start with software. Trial things. Learn what’s useful.
Medium risk (moderate business impact) → Training plus structured experimentation. Give your team tools and a clear framework for testing.
High risk (significant investment, regulatory exposure, competitive urgency) → Get expert help. The cost of a consultant is almost always less than the cost of a failed AI project.
2. How fast do we need results?
Exploring → Training path. Build understanding first.
This quarter → Consultant. Buy speed and certainty.
No urgency → Software trials. Low cost, low commitment.
3. What’s our internal capability?
Strong tech culture, self-starters → Software plus targeted training. They’ll figure out the rest.
Mixed capability → Consultant-led engagement with training baked in. Transfer the knowledge as you go.
Limited confidence → Consultant essential. You need someone to show, not just tell.
A Worked Example
A 50-person professional services firm. £5M turnover. No meaningful AI capability yet. Leadership knows they need to move but isn’t sure where.
Month 1–3: Consultant engagement Audit current workflows. Identify the single highest-ROI opportunity: perhaps proposal generation, or client onboarding, or knowledge management. Build a working system. Measure the impact.
Month 4–6: Team training on the live system Train the people who’ll use it daily. Not “AI fundamentals.” Practical training: how to use this system, how to spot when it’s wrong, how to improve it. Document everything so the knowledge doesn’t walk out the door.
Month 7–12: Expand to adjacent processes Now you have internal capability and confidence. Apply the same approach to the next highest-value problem. The consultant becomes an occasional advisor, not a hands-on implementer. Software purchases are targeted, not speculative.
Almost all of the first-year cost sits in months one to six, because that is where the outside help is. Whether it was worth spending comes down to a question you can answer before you commit: which team, and which hours of their week, is the first project supposed to give back? If you cannot name them, the arithmetic will not work at any price.
The Three Mistakes to Avoid
1. Buying tools before defining problems. Software vendors are very good at making you believe their tool solves your problem. They’re less good at asking whether you’ve properly defined the problem. Start with the problem. Always.
2. Training without a plan to apply. A training budget without an implementation plan is a donations budget. Every pound you spend on training should have a corresponding plan for how that skill gets used within 30 days.
3. Hiring a consultant who advises but doesn’t build. Strategy without implementation is an expensive opinion. Insist on working systems, not slide decks. Insist on measurable outcomes, not frameworks. If they can’t show you something working within the first month, they’re the wrong consultant.
The Bottom Line
There’s no shame in any of these paths. Businesses succeed with all three, and fail with all three.
The failure mode isn’t choosing the wrong option. It’s choosing one that doesn’t match your reality, or spreading too thin across all of them.
Software without adoption is wasted money. Training without application is wasted time. Consultants without accountability are wasted opportunity.
Pick the path that fits your timeline, your risk tolerance, and your team’s actual capability. Then commit to it properly.
If you are stuck between two of these, the answer is usually in how the work actually runs today rather than in the options themselves. Going and looking at that is working out what to do, and it ends with a plan rather than a recommendation to buy something.
Jonathan Gill is the founder of Squared Lemons. He reads proposals like the ones in this article and says plainly whether they will fix the problem, and builds the first version when the answer is that something needs building.