AI Adoption FAQs
Answers to the questions I hear most from solopreneurs, small firms and corporate teams adopting AI — covering Microsoft Copilot, Claude and ChatGPT.
“We’ve got Copilot licences and nobody’s using them”
1. We’ve rolled out Copilot but hardly anyone uses it - is that normal?
Completely normal, and it isn’t everyone being lazy. Having a licence is not adoption - people are likely just waiting for permission, a reason, and a safe way to have a first go. The most effective way to move forward is helping each person use it on one real task on their to do list.
2. Where should our team start with Copilot?
Start with the work, not the technology. Pick one task someone already does regularly (a status report, meeting prep, a first draft) and apply Copilot to that. A general instruction to “Learn AI” is not a great starting point; “get help with this one thing” works.
3. Do staff need training to use AI, or can they just pick it up?
A bit of both. Anyone can open Copilot, Claude or ChatGPT and type, but without a little help most people get generic results, decide it doesn’t work, and eventually stop. Short, hands-on sessions where people work on their own tasks is more effective than a generic webinar and they leave having finished something.
4. How do I get sceptical or nervous staff to try AI?
Start with a frustration they already have, not with the tool. Nobody needs to become “an AI person” - they need one small win on their own work. It also helps to practise on realistic fake data first, so nothing feels risky while they learn.
5. What does an AI adoption workshop involve?
It’s hands-on, not a lecture and it leans on peer-to-peer learning. People work in small groups on real tasks from their own to-do list and leave with something finished and ideas on how to keep going. I design each one around your firm’s actual work, not a generic syllabus.
“The outputs aren’t good enough to use”
6. Why does Copilot give us such generic answers?
Because it doesn’t know who you are. Without context it writes for the person it assumes you are and what you need, which tends to produce very average, bland outputs. Tell it your role, your firm and what good looks like (a past example works well), and the output changes dramatically.
7. How do I get AI output to sound like me, not a robot?
Show it how you sound. Create a short tone-of-voice guide from your own writing. Giving that to the tool with each task is often the single biggest quality jump. I build these with clients regularly - it’s an hour well spent.
8. Why does Copilot keep pulling old or wrong documents?
Copilot is hooked into your files, emails and calendar, so it has a lot to draw on - but without any guidance it can rummage in the back of the digital cupboard. It doesn’t know which version is the gold standard unless you tell it, so old templates and outdated guidance surface. Be direct: use this file, follow this method, ignore everything else.
9. We tried AI once and the result was poor - is it worth another go?
Yes. One poor result is likely to mean the ask was too broad, or there was not enough context - not that the tool can’t do the job. Change one thing (eg narrow the task, add an example) and try again before you write it off.
10. Why are AI answers such overwhelming walls of text, and how do I stop that?
AI tools like Copilot, Claude or ChatGPT very much want to please you, so they will give you a ‘full monty’ answer which can be very overwhelming and actually unhelpful. There are ways to choreograph the answer to make sure you are in the driving seat. For example, you can ask for it in small chunks, tell it to wait before continuing, or ask for three bullet points instead of an essay.
11. How do I stop AI-drafted work feeling like it isn’t mine?
Treat it as a first draft, because that’s what it is. Tone, trimming and judgement stay with you. The test I use: would you comfortably put your name on it? If not, it is good to get into the habit of going through a feedback loop - for example being very specific about why it doesn’t work, giving a better example and reshaping it until you are happy to stand behind it. If you take a moment to ask it to reflect on the before and after, Copilot can identify patterns and propose writing guidelines for you to apply next time.
“Can we trust it?”
12. How do I know whether I can trust what AI produces?
Check it the way you’d check a capable junior colleague’s work. Ask where it got its information, and verify anything before it leaves your desk. Asking for sources - and checking them - is an example of responsible use.
13. Who is accountable when AI-assisted work goes to a client?
You are - and that doesn’t change with the tool. Clients pay for your judgement and your name on the work. It doesn’t really matter how the first draft was produced - the key thing is that the quality of the work meets your standards and your clients’ expectations.
14. Is it safe to put client data into Copilot, ChatGPT or Claude?
It depends on the tool and how it’s set up - enterprise versions with the right settings are a different animal from free consumer tools. The rule of thumb I give teams: if you’re not sure, don’t. It is also good practice to let clients know that you use AI tools in your work - a line in your terms of business means nobody is surprised later. In training I use realistic fake data, so people can practise properly without any risk.
15. What should we tell staff they can and can’t paste into AI tools?
Give them easy to understand guidance rather than an extensive policy that nobody reads. Something like: no client names, no live financials, no personal data - unless the tool has been approved by the firm for exactly that.
“Where is AI worth the effort?”
16. Which parts of our business should we use AI on first?
The low-risk, structured, recurring work, so that the effort you put into setting AI up pays back every time the task repeats. A genuine one-off task that’s quick to do may not be worth prompting at all. Working out that triage is exactly what a discovery session is for.
17. Is AI really saving small firms time, or is it hype?
For small firms the payback can be remarkable - but it isn’t automatic. The disappointing headlines mostly come from big organisations, where saved time gets absorbed into more meetings and busy work. In a small business the saving is yours to keep. I’ve sat in sessions with clients where we were honestly giggling at how much we’d got done.
18. When are AI agents worth building - and when aren’t they?
Agents suit narrow, repeated, well-defined work: status reporting, meeting prep, standard outlines. They are not a replacement for real thinking, and they’re not worth building for one-off tasks. Agents are best deployed where consistency matters – there is no need to put agents everywhere.
19. Why is using AI taking me longer than doing the job myself?
That’s the learning curve, and it’s normal. There’s an upfront time investment before any saving, and it pays off where the task repeats or the output gets reused. If it’s a genuine one-off that’s easy manually, doing it yourself is a perfectly sensible decision.
20. Microsoft keeps releasing AI features that cost extra (like Copilot Cowork) - should we switch them on?
For the moment I would recommend holding off until you understand the costs. Features like Cowork are charged on credits on top of your subscription, usage is hard to predict, and there’s little warning when you’ve run out. Check with IT that they aren’t auto-provisioned - most teams can get a long way on the flat-cost subscription they already have.
“What would working with you look like?”
21. What does an AI discovery session involve, and what do we get at the end?
A discovery session is a guided conversation that takes a deliberately wide view of your firm - where AI could improve productivity, quality or consistency, and where it is unlikely to. I then take it away, do the thinking and the research, and come back with prioritised recommendations. You get a clear, ordered view of where to start and why.
22. How much does AI adoption support cost for a small firm?
For a solopreneur or micro business, a discovery session and AI implementation report costs £950 + VAT. A bespoke hands-on workflow-based training session with up to 7 people starts at £3,250 + VAT. Subsequent ongoing support is tailored to your needs from £750 + VAT a month (figures correct as at April 2026)
23. How long does it take to see results from AI adoption work?
Some of it lands the same day - people regularly leave a workshop buzzing, having finished a real piece of work. Habits across a team take longer; a quarter is a fair expectation for adoption you can see. The speed depends less on the tool and more on whether people start to use it regularly on work that repeats.