If you run a small or mid-sized business, you've likely had the AI conversation with yourself more than once. The pressure is real, and it isn't just vendor noise: your competitors are already experimenting, and peers mention it like they've got it sorted. Meanwhile you, sitting where the decisions land, have neither the hours to go deep nor the appetite to commit to a platform you don't yet understand. The instinct to wait until it all makes sense is the one move that quietly costs you ground.
The core observation we keep coming back to is this: someone needs the time and the mandate to explore the options for your business, and directors typically have neither. So hunting for which AI solution to go for becomes a stressful endeavour, something you're trying to do in the margins while keeping the show on the road and evolving it with the technology at the same time.
The first move solves exactly that, and it costs almost nothing. It isn't a purchase. It's a person.
Why a champion beats a consultant (and beats doing it yourself)
The instinct is to either bring in outside help or lead it yourself.
Bringing someone in is expensive, and slower than it looks. A consultant or new hire arrives knowing AI (and even that to varying degrees, we've found) but not knowing you: your customers, your quirks, the spreadsheet that secretly runs the business, the reason a "simple" process is actually three exceptions in a trenchcoat. They spend their first month learning what your own people already know in their bones, and you pay premium rates for that ramp-up.
Doing it yourself rarely works either. Not because you couldn't, but because you can't carve a continuous, curious block of attention out of a director's week. AI adoption rewards tinkering: trying the thing, watching it fail, trying it differently. That's how all real learning is done, and it's the first kind of work to get squeezed out when the calendar fills. Done properly, this stage isn't just a procurement exercise; your organisation is learning by doing.
So pick a champion from inside. Someone who already understands the business, is naturally curious, and is a little impatient with how things are done today. They don't need to be technical. They need to be interested, trusted, and given permission.
What "giving them permission" actually means
A champion with no time and no mandate is just an employee with a side-quest. To make this real:
- Give them hours, not just encouragement. A protected half-day a week is enough to start. Without it, this quietly becomes the thing they'll get to "when it's calmer." It is never calmer.
- Give them a real problem, not a sandbox. "Go learn AI" produces nothing. "See if you can halve the time we spend writing quotes" produces a champion who comes back with something you can use.
- Give them room to fail. The first few experiments will be mediocre. That's the cost of finding the good ones. Make it explicitly safe to report "that didn't work."
Start with what's free or nearly free
You don't need to buy a platform to begin. The barrier to useful AI is currently very low. Point your champion at the freely available tools first:
- ChatGPT (OpenAI): the most widely used general assistant, and the easiest place to start. Good at: drafting, summarising, brainstorming, explaining things in plain language, working through a messy email or document. Less good at: knowing anything specific about your business unless you tell it, and it will occasionally state something wrong with complete confidence. There's a usable free tier; the Plus plan is around £19/month.
- Claude (Anthropic): a close peer to ChatGPT. Users often say it writes more naturally and handles long documents more carefully. Good at: thoughtful drafting, analysis, and working through large amounts of text at once. Less good at: a smaller ecosystem of third-party add-ons than ChatGPT. There's a free tier; the Pro plan is around £16/month (Anthropic bills in US dollars, so the exact figure moves with the exchange rate).
- Google Gemini: most useful if your business already runs on Google Workspace. The free Gemini app can chat, summarise, and draft, but it doesn't reach into your Gmail, Docs, or Drive; that in-app integration needs a paid Workspace plan, where Gemini is included in the Business tiers (from around £14 per user/month). Good at: working with content you already keep in Google, and quick research. Less good at: the genuinely useful in-app features sit behind that paid plan, not the free app.
- Microsoft Copilot: two different things share this name, and the difference matters. The free Copilot is a general chatbot. The version that works inside Word, Excel, Outlook, and Teams and across your company data is Microsoft 365 Copilot, a paid per-seat add-on (roughly £18–25 per user per month) that sits on top of a qualifying Microsoft 365 licence; you can't buy it on its own. Good at: bringing AI into the Office apps your team already uses, once it's licensed. Less good at: starting cheaply, and connecting to anything beyond Microsoft 365. Pulling in your other tools (a CRM, Confluence, Box and the like) needs connectors configured by someone with administrator access, not something your champion can quietly set up alone. Treat it as a later step, not a first experiment.
- NotebookLM (Google, free with a Google account): a quieter tool that punches above its weight. You upload your own documents and ask questions across them, and every answer cites the exact passage it drew from. Good at: "what does our own policy actually say about this," summarising a stack of reports, and staying grounded in the sources you give it. Less good at: anything outside the documents you upload; it deliberately won't draw on the open web, and the free tier caps you at around 50 sources per notebook.
- Perplexity: a research tool rather than a chat assistant. It searches the live web and answers with numbered, clickable citations on every answer. Good at: current questions that need real sources you can verify. Less good at: longer drafting and back-and-forth work, where ChatGPT or Claude are stronger. Users often say its citations aren't always reliable, so spot-check that a linked source actually says what's claimed. There's a free tier; Pro is around £16/month.
Prices are correct as of May 2026 and shown in pounds. Several of these tools bill in US dollars, so the sterling figure moves with the exchange rate, and UK VAT may be added at checkout. Prices change often, so check each provider's own pricing page before committing.
When a specific task starts to look promising, let them spend small money on low-cost trials of more specialised tools: a meeting-notes taker, a transcription tool, something niche to your industry. Most are a few pounds a month and cancellable. The budget for this stage should be the price of a couple of subscriptions, not a project.
What you're really building
Beyond a few automated tasks, this stage builds evidence and judgement inside your own walls. After a few months you'll have someone who can tell you, in your language, what AI is genuinely good at in your business, where it falls over, and what's worth paying for. That person is worth more than any vendor's demo.
It also surfaces the next problem, which is a good problem to have. Free tools are ideal for individuals experimenting. They start to strain the moment you want a whole team using AI consistently, when it needs to know your business, follow your rules, keep your data in the right place, and not depend on one enthusiast quietly doing clever things in a chat window. That's when a proper adoption platform starts to earn its place, and you'll approach that decision as an informed buyer rather than a hopeful one.
But that's the next step. Don't rush to it. This one is a person, a few free tools, a real problem, and permission to experiment. Get it right and everything after is easier, and you'll know what you're buying, and why.