Is my business ready for AI?
That is the whole test. It has nothing to do with the size of the business, the sector, or whether anyone on the team is technical. A sole trader plumber with a WhatsApp full of quote requests is more ready than a twenty-person firm where nobody can explain how a job gets priced.
The reason readiness gets overcomplicated is that most advice is written by people selling the tools. The uncomfortable part is that the preparation work, writing down how things are done, is unglamorous and unavoidable.
What does "ready" actually mean in practice?
Take a real example. A letting agent sends the same tenancy renewal reminder dozens of times a month, with slight variations for property and date. The task repeats, the inputs sit in a spreadsheet, and the office manager knows instantly whether a draft reads right. That is ready.
Now take the same agent's decision about which properties to take on. It happens irregularly, the reasoning is instinct built over fifteen years, and nobody has written it down. That is not ready, and no amount of software changes it.
What are the strongest signs you are ready?
Look for these in your own business:
- Somebody spends more than two hours a week retyping information that already exists somewhere else.
- You have a document, email template or checklist that people copy and edit each time.
- Your enquiries arrive in a consistent format, even a messy one, such as a web form or an inbox.
- You can describe the job to a new starter in under ten minutes.
- The output is drafted by a person and reviewed by a person, so a wrong draft is caught before it reaches a customer.
That last point matters most. Reversible tasks are the right place to start, which is why chasing late invoices and general paperwork drafting are where most UK small firms get their first real win.
What are the signs you are not ready?
The head-only problem is the most common in small firms. Pricing logic, supplier preferences and the reason a particular client gets special treatment all live in the owner's memory. A tool cannot see any of it. The fix is to write the process down, which pays for itself even if you never buy anything.
The broken-process problem is subtler. If quotes go out late because you are unsure what to charge, speed is not the constraint. Automating the sending will not help.
The definition problem shows up as arguments about output quality with no reference point. If two people in the business disagree about whether a customer email is good, no tool will settle it.
How many UK small businesses are actually using AI?
The gap between those numbers is instructive. Surveys that count any use, including features built into tools people already pay for, land higher. Surveys that ask about deliberate AI use land lower. If you have ever used a summarise button in your email client, you are already inside one of those figures without having made a decision.
The ONS also finds the information and communication sector well ahead of the average, and the smallest firms well behind it. Sector and size matter more than most owners expect.
What should the first AI task be?
ONS research puts large language models at the top of the list of AI technologies UK businesses actually use, ahead of visual content tools. Text work is where the mainstream is, and it is where the tooling is most mature.
Practical first tasks, roughly in order of how quickly they pay back:
- Summarising calls and site visits. Low risk, immediate time saving, no customer sees the raw output. Meeting notes are the standard starting point for a reason.
- First-draft replies to repeat enquiries. You still read and send them.
- Quote and proposal drafting from a template you already use.
- Chasing overdue invoices on a schedule, with your wording.
Notice what is missing. Nothing on that list makes a decision. The human stays in the loop at the point where being wrong would cost money.
How much should it cost to start?
Those figures come from Microsoft's own UK pricing page, where annual commitment sits at the lower end of that band and monthly billing at the higher end. Assistant subscriptions from other providers sit in a broadly similar range for a single business user. Check the current price yourself before you budget, because these plans change often.
The larger cost is not the licence. It is the hour or two someone spends writing down how the task is done today, plus a few weeks of checking output before trusting it. Budget that time honestly. Owners who skip it are the ones who conclude after a month that the tools do not work.
There is also a real trade-off to name. Annual billing is cheaper per month and worse for testing. For a first task, pay monthly and accept the higher rate until you know whether it sticks.
Will this mean fewer staff?
That finding sits in the BCC's 2026 report with Atos, produced with its Insights Unit and the University of Essex. Read alongside the ONS finding that adoption is shallow, it describes something recognisable: firms are removing admin from people who were already at capacity, not restructuring teams.
Be careful about promising staff either outcome. Saying nothing will change is a claim you cannot back. Saying jobs are at risk creates resistance that will quietly sabotage the rollout. The honest version is that the first tasks are admin tasks, and the people doing them will be involved in checking the output.
How do I know whether it worked?
Pick the measure that a customer would notice. Time to first response on an enquiry is a good one for service businesses. Average days from job completion to invoice sent is a good one for trades.
Then set a stopping rule in advance. Something like: if quote turnaround has not improved by four weeks, cancel the licence and go back. Written down beforehand, that rule is easy to follow. Decided afterwards, it never gets applied, and small firms end up carrying subscriptions nobody uses.
What if the answer is "not yet"?
Concretely, over four weeks: write down the five tasks that eat the most time, describe each one as if training a new starter, find where the source information lives for each, and note who would check the output. At the end you will either have an obvious first candidate or a clear reason why none of them qualify yet.
The ONS finding that a large share of UK businesses report no barriers at all to adoption is worth sitting with. For many firms, nothing external is stopping them. What is missing is a specific job for the tool to do.
If you want to see what that looks like applied to a particular trade rather than in general, the sector guides on the Claude for small business hub work through the same readiness test with concrete examples.
What is the one thing to avoid?
Everything above reduces to that. Readiness is not a technology question. It is a question of whether you can describe your own work clearly enough for something else to help with it.
Signs Your Small Business Is Ready for AI (and Signs It Is Not) — FAQ
Is my business ready for AI if I have never used it before?
Probably yes, if you have one repetitive written task that happens most weeks and follows a rough pattern. Readiness is not about technical skill. It is about having a task worth automating and a way to check the output. Most owners start with something small and reversible, such as drafting replies to enquiries or summarising notes after a call, and keep a human check on everything that leaves the business. Office for National Statistics research on AI in UK businesses puts large language models at the top of the list of technologies firms actually use, so text work is where most begin. If you have no task like that, wait. Buying a licence before you have a job for it is the most common way small firms waste money on this.
What are the clearest signs my business is not ready?
Three signs matter more than the rest. First, your data lives only in your head or in a shoebox of receipts, so there is nothing for a tool to work from. Second, nobody can say what good output looks like, which means you cannot check the result and cannot tell whether it helped. Third, the underlying process is broken rather than slow, in which case automating it just produces the wrong answer faster. A fourth warning sign is cash pressure. If a monthly licence would hurt, fix the cash position first. None of these are permanent blockers. They are just work that has to happen before software helps, and doing that work usually improves the business on its own.
How much should a UK small business expect to spend to start?
Start at the price of a licence or two rather than a project. Microsoft's UK pricing page for Microsoft 365 Copilot puts the business add-on in the region of £16 to £20 per user per month, depending on whether you commit annually or pay monthly. Assistant subscriptions from other providers sit in a broadly similar band for one business user. That is the realistic floor for one person testing one task. Budget for the time cost too, because someone has to write down how the task is done today and check the output for the first few weeks. If a tool cannot justify roughly that monthly figure in saved hours within a month, stop paying for it. Check current pricing directly, since these plans change often.
Will adopting AI mean cutting staff?
The evidence so far says no for most small firms. In its 2026 research with Atos, where the sample was overwhelmingly SMEs, the British Chambers of Commerce found the large majority of SMEs using AI reported no impact on workforce size over the past year, and most said job roles had remained unchanged. That matches what shallow adoption looks like in practice. Firms are using one or two tools to take admin off people who were already stretched, not restructuring teams. Treat headcount reduction as an outcome you have not earned yet. The realistic first-year gain is fewer evenings spent on paperwork and faster response times to customers, which is a better thing to promise your team anyway.
Should I wait until the tools get better?
Waiting for better tools is usually a way of avoiding the harder work, which is writing down how your business actually runs. That work does not expire. A clear description of how you quote, invoice, chase and hand over keeps its value whichever tool you eventually use. Office for National Statistics research suggests the average adopting business uses only one or two AI technologies, little changed since 2023, so most firms that adopted early are not far ahead of you. Adoption has widened rather than deepened, which is good news if you feel behind. Spend the waiting period documenting processes rather than reading product announcements, and you will pick a tool faster and judge it better when you do.
