OptimisedI've spent the last three years helping small businesses put AI to work, so I can say this with some confidence: the gap between what the advertising promises and what actually happens in a working business is enormous. Some AI tools genuinely deliver for a five-person firm in 2026. Others, heavily marketed though they are, mostly turn into unnecessary expense. It's worth being clear about which is which.
The short answer is this. AI today does four things reliably. It helps you draft your writing — emails, proposals, product descriptions, social posts. It answers routine customer questions around the clock. It takes some of the weight off bookkeeping and admin. And it speeds up marketing content considerably. For most small businesses, all of that costs somewhere between £16 and £115 a month.
But it's worth understanding the other side too. AI won't run your business for you, it won't replace your experience or your judgement, and it certainly won't produce miraculous revenue growth just because you've bought a tool. What follows is what these technologies are genuinely useful for, what they cost, and what results you can expect according to independent research rather than the vendors' marketing.
Where to start? What the real numbers show
If you try to work out how many small businesses are actually using AI, you'll quickly notice the sources disagree wildly. Before believing any headline figure, it's worth understanding why.
One of the more reliable studies is a Goldman Sachs survey conducted in early 2026, covering more than 1,200 small businesses across every US state. It found 76% of firms already using AI in one form or another.
A separate QuickBooks-commissioned study of more than 2,200 businesses found 68% using AI regularly, and 28% daily.
Meanwhile a US Chamber of Commerce study, looking specifically at generative AI, put the figure at 58%.
At first glance the numbers look contradictory, but in fact they answer different questions. Some measure any AI use, others only generative AI, others use a different sample of respondents. Whatever the differences, they all point the same way — AI is rapidly becoming an everyday small business tool.
There's one number the technology companies rarely put in their advertising, though.
In that same Goldman Sachs survey, only 14% of firms said AI was fully integrated into their core business processes. The US Federal Reserve applied an even stricter test, counting only cases where AI is genuinely used day to day rather than merely trialled. As of April 2026, that figure was about 18%.
So the real picture looks like this:
- most small businesses have tried AI at least once;
- a large share use it for one task or another;
- but only a small minority have genuinely reworked how they operate.
That last group gets by far the most out of it. The good news is that joining them usually takes more consistency rather than more money.
There's one more important finding.
Researchers at MIT's NANDA initiative analysed AI projects at large companies and found that 95% of generative AI pilots produced no measurable financial return.
Worth noting that the study looked at large organisations, not small businesses. Even so, the central conclusion applies to everyone: the teams that did well picked one clear problem and worked at it consistently.
The lesson is simple.
AI tends to fail when a company adopts it because "everyone's doing it now". It succeeds when it's pointed at one specific, frequently repeated, easily measured task.
Keep that in mind for the rest of this article.
The best £16 you'll spend this year
If you make only one move towards AI this year, make it a paid subscription to a general AI assistant.
It's probably the least impressive-sounding advice in this article, and it's also the advice that most often pays back the most.
By a general AI assistant I mean tools like ChatGPT, Claude or Gemini. You give it a task, and it helps you draft the text, analyse the information, or work through the options.
By 2026 their prices have all but converged.
- ChatGPT Plus — around £16 a month;
- Claude Pro — around £16 a month (or roughly £14 billed annually);
- Google AI Pro — around £16 a month.
Cheaper plans have appeared too — ChatGPT Go at about £6, or Google AI Plus at around £4 a month. For business use, though, the cheapest ChatGPT plan isn't the best choice, because it shows adverts.
There are also £75 and even £150 subscriptions, but in your first year most small businesses simply don't need them.
Start with the £16 plan.
What to use it for
Not for impressive experiments. For the everyday work.
For example:
- writing a proposal for a client;
- drafting the email you keep putting off;
- turning scattered voice notes into a clear quote;
- reducing a 30-page supplier contract to the points that matter;
- writing a job advert;
- rewriting your website's service descriptions so they don't read like 2014;
- preparing for a difficult conversation with a customer.
According to Business.com research cited by Booth Associates, the average small business using AI saves about 5.6 working hours a week, and owners and managers save more than 7.
From my own practice, those figures hold up reasonably well for people whose work involves a lot of reading and writing. If your work is mostly physical, the time saved will be considerably smaller.
Which tool should you pick?
If this is your first subscription, the specific choice matters far less than the habit of using it.
I usually recommend a simple method.
Take the free versions of two different tools. Then run five real tasks from your working week through both, and see which one's output needed less fixing.
If most of your work happens in Gmail, Google Docs and Google Sheets, Gemini is worth choosing, since it's already built into those apps and comes with 2 TB of cloud storage.
If you write a lot — reports, proposals, long documents — I'd usually suggest Claude.
If a broad ecosystem of plugins, integrations and other tools matters most, ChatGPT is the better pick.
The truth is that all three are good enough today. After a few weeks of use, your ability to describe a task precisely will affect the results far more than the platform you chose.
One important warning
AI sometimes states incorrect information with total confidence.
In the technology world this is called hallucination — the system presents wrong facts as though they were entirely settled.
So never send a client AI-generated:
- calculations,
- legal statements,
- medical information,
- financial recommendations
without checking them yourself.
Treat AI-written text as though it came from an extremely fast, never-tired, but occasionally careless intern.
Customer questions — the first place AI genuinely pays for itself
Ask any small business owner where most of their week goes and you'll usually get the same answer: answering the same questions over and over.
What are your opening hours? Do you have this in stock? Where's my order? Can I move my appointment?
This is where AI creates the most value today. The tasks repeat constantly, the answers are usually clear-cut, and the stream of questions never stops.
The tool for this is an AI chatbot. It goes on your website or connects to Instagram, Facebook Messenger, WhatsApp and other channels. The assistant understands the customer's question, finds the relevant information in your business data, and answers without anyone stepping in.
Crucially, current systems bear almost no resemblance to the old chatbots that made you press "Choose 1", "Choose 2", "Choose 3". The new generation understands natural language and holds an ordinary conversation.
According to Tidio's own research, these agents now resolve 55–70% of customer enquiries independently. Other figures they cite suggest handling one enquiry with AI costs about £1–£3, against £6–£9 for a person.
Naturally, statistics published by the platforms themselves deserve some scepticism. But even halving those numbers leaves the trend obvious — automated customer support can already save a small business a serious amount of time and money.
What does it cost?
Here I want to help you avoid one of the most common mistakes: overpaying.
Not all AI chatbots price the same way.
One of the clearest comparisons on the market shows Tidio's basic plan starting at around £22 a month. Its Lyro AI agent add-on starts at roughly £30 a month, which includes 50 automated conversations. Larger plans scale up to around £220 for 500 conversations.
In practice, most small online shops pay about £75–£115 a month for the whole system.
If your customer volume is still modest, Crisp is worth a look. It offers one of the more generous free plans on the market, which makes it well suited to first experiments.
Why Intercom isn't right for everyone
There's one platform I usually tell small businesses to steer clear of: Intercom.
The problem isn't the technology. Quite the opposite — its Fin AI is regarded as one of the most capable on the market.
The problem is the pricing.
On top of a monthly per-seat fee of roughly £22 to £100, you pay close to £1 for every conversation the system successfully resolves.
In other words, the more customers you serve, the faster the bill grows.
In one worked example, a five-person team with around a thousand AI-resolved conversations a month would pay about £4,500 for the system.
That's an enterprise budget.
So for online shops, local service businesses and most start-ups, platforms with a flat monthly fee — Tidio, for instance — are the far more sensible choice.
Intercom generally earns its price only at fast-growing SaaS companies with considerably more complex support operations.
Two rules that decide whether a chatbot is any use
From experience, there are two settings that account for almost all of the success.
- Always leave a route to a human
The customer needs an obvious button or phrase that passes the conversation to a person.
There's no faster way to annoy someone than trapping them in a chatbot loop for ten minutes.
If the AI doesn't understand the situation, a human takes over.
- Feed it your business information before you launch
Don't let the chatbot improvise.
It needs to know:
- delivery times;
- your returns policy;
- the area you cover;
- opening hours;
- prices;
- your frequently asked questions.
Without that information, the AI will start… inventing it.
And an invented returns policy can very quickly become one you're obliged to honour, at your expense.
AI that answers the phone is quietly becoming one of the most useful tools going
Another technology gaining ground quickly is the AI phone assistant.
It answers your business calls, speaks naturally with the caller, answers questions, books appointments, reschedules them, and passes the call to a person when needed.
For appointment-based businesses — dental practices, salons, garages, heating and plumbing firms, clinics — this can be a more important investment than a website chatbot.
A missed call often means a lost customer.
One analysis of ten small businesses describes a two-dentist practice where no-shows were costing £600 to £900 a day, and the receptionist spent roughly two hours a day just confirming appointments by phone.
An AI phone assistant automated both jobs.
Across all ten businesses studied, average AI spending was about £90 a month, against an average recorded benefit of roughly £3,100 a month.
One more detail worth noting: not one of these businesses made anyone redundant.
Quite the opposite.
The time that repetitive calls used to swallow went into serving customers, seeing patients and growing the business.
Naturally, examples like these should be read as best-case scenarios rather than promises. Companies publish their successes far more readily than their failures.
But the trend matches what I see in practice.
Businesses that always answer the phone win more customers than businesses that let calls go to voicemail.
And AI can now do that at three in the morning.
Marketing content without an agency
Most small businesses meet AI through marketing, and the research bears that out. A 2026 SBE Council survey found 82% of small employers investing in AI, with content and marketing among the areas where the benefit shows up quickest.
The reason is straightforward: you see the results immediately.
Where a social post used to take 45 minutes, it now takes ten. A month's communications plan can often be put together in an afternoon.
Which tools do you actually need?
The set I recommend is very simple.
The first is a general AI assistant at around £16 a month. It helps you produce:
- social post drafts;
- newsletters;
- ad copy variations;
- product descriptions;
- website copy.
The second is a design platform. For most small businesses, that's Canva.
Canva Pro costs around £11 a month for one user. It gives you Magic Studio: image generation, background removal, automatically resizing a design for different social platforms, and other AI features.
If you produce visual content regularly, that subscription often pays for itself in the first week.
One more benefit many people missed — in 2026 Canva included the professional Affinity design suite at no extra cost. As a result, a single subscription now covers work that previously needed considerably more expensive Adobe software.
But there's an important limit here
Since this is a guide without the gloss, here's what the advertising tends to leave out.
AI-generated content is not a finished product.
It's a first draft.
Publish it entirely unedited and your social feed will look like the tens of thousands of other businesses doing exactly the same thing this year.
That kind of text is usually:
- too generic;
- too polished;
- devoid of any actual opinion;
- instantly forgettable.
The businesses winning today use AI to speed up the work, not to produce the finished article.
They let AI build the structure, then add what no software can produce:
- photos from yesterday's job;
- genuine customer feedback;
- their own experience;
- the owner's opinion;
- specific examples from the working day.
That's what makes content feel real.
In research published by CO by US Chamber of Commerce, business owners themselves said their most important competitive advantage remains a human voice and a personal style.
Hard to disagree.
Let AI do the mechanical work. Keep the personality for yourself.
A warning about "magic" writing tools
The market is full of platforms promising to produce hundreds of SEO articles at the press of a button and lift your site to the top of Google.
Treat those promises with real caution.
Search engines are getting steadily stricter about low-quality, mass-generated content.
Over the past year I've watched more than one small business flood its site with barely edited AI articles and lose a chunk of its organic traffic.
Ten good articles a human has reviewed beat a hundred generated ones.
Quality has mattered more than quantity for a long time now.
Bookkeeping, invoices and admin
This is probably the least interesting part of the article.
It's also one of the most profitable.
The MIT study mentioned earlier revealed an interesting pattern.
Most companies direct their largest AI investments at sales and marketing.
But the biggest financial return usually appears somewhere else entirely — in automating administrative processes.
For a small business, that means bookkeeping.
If you use QuickBooks, you're probably already paying for AI features
Plenty of QuickBooks Online users never realise their subscription already includes AI capabilities.
The system can:
- categorise bank transactions automatically;
- match receipts to payments;
- flag suspicious anomalies;
- chase clients over unpaid invoices.
According to QuickBooks' own data, businesses using Intuit Assist's automatic invoice reminders get paid an average of five days sooner, and the number of overdue invoices settled on time rises by around 10%.
The company also claims AI saves up to 12 working hours a month on accounting tasks.
Since that's vendor-supplied statistics, treat it cautiously.
Still, one thing shows up in practice constantly.
AI has no problem sending a third or fourth reminder to a client.
People find that awkward.
Which is precisely why automated reminders often get the money back faster.
You don't need to change accounting systems
Xero, FreeAgent, FreshBooks and other popular platforms offer much the same features today.
So the main advice isn't to switch systems.
It's to switch on the features you're already paying for.
The essentials:
- connect your business bank accounts;
- connect your payment cards;
- let transactions import automatically.
Then set aside one working session to correct the mistakes the AI makes.
The system learns from your corrections.
The longer you use it, the more accurate its suggestions become.
What not to automate blindly
There's one line I wouldn't cross.
Letting AI file anything with HMRC on its own is a bad idea.
Categorising transactions isn't dangerous.
If the system gets one wrong, you correct it.
Tax returns are another matter entirely.
Mistakes there can be very expensive.
So don't dispense with your accountant.
Make sure they receive tidier, better-organised data instead.
Admin — another area where AI saves time
The same principle applies to other everyday tasks.
Most video meeting platforms — Zoom, Google Meet, Microsoft Teams — now have AI assistants built in.
They automatically:
- transcribe the meeting;
- produce a summary;
- pull out the key decisions;
- build an action list.
It sounds minor.
But in a business with more than a few people, these tools all but eliminate the eternal question:
"Hang on, what did we actually agree?"
It's one of those features that saves a few minutes daily and adds up to dozens of hours a year.
What AI still can't do in 2026
It's important to be clear about where AI's capabilities end. Software vendors rarely discuss this, but for a business it's among the most important topics.
First, AI cannot rescue a business model that doesn't work.
If your business is already succeeding, AI can help you work faster, more efficiently and at lower cost. But if the underlying idea doesn't work, technology won't change that — it will simply let you make the same mistakes faster.
The other thing AI won't replace is human relationships.
It won't build trust with customers, strengthen long-term partnerships, or make the decisions your business's future depends on.
For instance:
- which client you're better off letting go;
- when to hold firm against pressure to cut your price;
- which compromises you absolutely will not make.
Those decisions need experience, instinct and accountability.
AI handles the repetitive 80% of the work very well, but the remaining 20% still needs a person. That framing describes the real state of the technology better than anything else.
What about "agentic" AI?
The term agentic AI turns up in marketing more and more often.
It sounds impressive, but it's worth understanding what it actually means.
In plain terms, it's a system that carries out several linked steps in sequence.
For example:
- receives a customer enquiry;
- drafts a reply;
- updates the record in the CRM;
- creates a reminder;
- schedules the next contact.
Systems like this already exist and are improving quickly. They're expected to spread across finance, HR, scheduling and other areas over the next few years.
But there's an important caveat.
I've yet to see a system that can run for long without human oversight.
The problem is simple.
If the system makes a mistake at step one, every subsequent step compounds it.
So automate the process by all means, but build in checkpoints where a person reviews the output regularly.
Especially where money, contracts or customer data are involved.
Don't forget data protection
Another topic that often gets skipped: privacy.
Everything you put into an AI system remains your responsibility.
Customer data is subject to the same law as before — UK GDPR and the Data Protection Act 2018 — regardless of whether a person or an AI is processing it.
So a few simple rules are worth following.
If you handle sensitive information:
- use business AI plans that state clearly that your data isn't used to train models;
- don't upload complete client files to personal or free accounts;
- write a short internal policy setting out what staff may and may not put into an AI tool.
None of that costs anything, and it can save you from some very expensive mistakes.
Why most small business AI projects fail quietly
Over a few years I've come to recognise a recurring pattern.
It's almost predictable.
Over a weekend, an owner enthusiastically signs up for four or five different AI tools.
In the first week, each gets tried once or twice.
The results aren't impressive, because nobody mentioned that describing tasks well is itself a skill you have to learn.
The conclusion follows: "AI is overrated."
A few months later the subscriptions are cancelled.
End of story.
The research says the same
This isn't just a personal impression.
The Goldman Sachs survey found that:
- 45% of small businesses cite a lack of technical knowledge as the main barrier;
- 47% say the hardest part is simply choosing the right tools.
Meanwhile, US Small Business Administration data suggests roughly half of AI-using firms have invested in neither staff training nor reworking their processes.
In other words, the tools were bought.
The way of working never changed.
It's also striking that 82% of the smallest firms not yet using AI believe the technology simply isn't relevant to their business.
In some cases that may be true.
If there are two of you, you're not short of customers, and you spend your days on site, AI genuinely won't be a priority.
But in most cases the problem isn't the technology. It's the expectations.
Your staff are already using AI, even if you don't know it
Here's an aspect owners often ignore.
If you employ more than a handful of people, some of them are almost certainly already using ChatGPT, Claude or similar.
Usually from their personal accounts.
MIT researchers named this shadow AI.
Staff use apps to get their work done without the company having officially approved anything.
You can fight it.
It's far smarter to make use of it.
Choose one official tool.
Buy the business licences.
That way customer data stays out of personal accounts.
Then run a one-hour training session.
Ask whoever on the team knows AI best to show three colleagues three ways they use it in their daily work.
Practical sessions like that usually deliver better results than buying the most expensive plan.
And staff frequently spot opportunities a manager would never have thought of.
The person raising invoices every day knows exactly which parts of the process eat the most time.
How to avoid failing
The answer is surprisingly simple.
One tool.
One task.
One month.
One metric.
That's it.
Pick the job that annoys you most and recurs most often.
Give it one AI tool.
Use it daily for at least a month.
Even if the first week isn't impressive.
By week four, your ability to describe a task properly will be completely different.
At the end of the month, assess one clear metric.
For instance:
- how many hours you saved;
- how much faster you reply to customers;
- how many invoices are paid on time;
- how much the admin load has dropped.
Only then decide whether to carry on.
Interestingly, 93% of AI-using small businesses in the Goldman Sachs study reported a positive effect.
The difference between them and the disappointed group usually wasn't the choice of technology.
It was consistency.
Your first month: a plan that works
If you want a plan rather than theory, here's what I usually give small business owners.
It's simple, inexpensive, and aimed at a quick result.
Week one
Subscribe to one general AI assistant at around £16 a month.
From that day, use it for every piece of business writing without exception.
Writing an email to a client?
Use the AI.
Putting together a proposal?
Use the AI.
Writing a job advert or a newsletter?
Again, use the AI.
At the same time, start noting how much time you save.
You don't need elaborate spreadsheets — a note on your phone will do.
The point is to have real data when you decide, a month later, whether the subscription earned its keep.
Week two
If you regularly produce content for social media or advertising, get Canva Pro.
It's around £11 a month for one user.
If you barely do any visual communication, skip this step.
The aim isn't to accumulate tools.
The aim is to pay only for what actually gets used.
Week three
Switch on the AI features in your accounting software.
If you use QuickBooks, Xero, FreeAgent or another modern system, most of them are probably already included in your current subscription.
Connect:
- your bank accounts;
- your payment cards;
- automatic transaction imports.
This usually costs nothing extra.
Week four
If a large part of your day goes on answering customer questions, try an AI chatbot.
You don't need an expensive plan straight away.
Start with the free version.
Tidio, for instance, lets you trial a limited number of AI conversations at no cost.
Watch how the system handles real customers.
Only once you're satisfied it answers accurately and reliably is it worth moving to a paid subscription.
What does all this cost?
Follow this plan and your total monthly spend will usually land between £16 and roughly £55.
That's less than most offices spend on coffee.
In return, recovering more than five working hours a week is entirely realistic, provided you use AI consistently.
At the end of the month, do one important thing.
Cancel everything that didn't earn its place.
That habit is what stops a business from quietly paying for dozens of subscriptions nobody uses any more.
Four questions before buying any new AI tool
The market produces a new "revolutionary" solution every week.
To avoid getting lost in the promises, ask yourself four questions each time.
- Does this solve a real problem?
Does it automate work I genuinely do every week?
Or only work I could theoretically do?
If the problem doesn't exist, it doesn't need automating.
- Can I trial it without spending much?
A good product almost always lets you start free or for a token amount.
If a company demands a large commitment before you even know whether it suits your business, be careful.
- Will the pricing stay predictable?
Some tools charge a flat monthly fee.
Others charge per request, per conversation or per action.
At first glance that can look cheap.
But as the business grows, that model can become a problem very quickly.
Which is exactly why Intercom came up earlier.
- What if this tool disappeared tomorrow?
Would it be easy to move to a competitor?
Or would you rebuild everything from scratch?
The more dependent you become on a single vendor, the greater the risk later.
So favour solutions you could replace without serious loss.
To finish
If I had to sum up the state of AI in small business in 2026 in one sentence, it would be this:
The technology has finally reached the level where it does what a sensible business owner expects of it — reliably, quickly, and at an affordable price.
The biggest obstacle today is no longer the technology.
It's available to almost anyone.
The real challenge is using it consistently.
The businesses getting the most out of it aren't the ones investing thousands in complex systems.
Usually it's the ones using one or two well-chosen tools with discipline, every day.
A £16 subscription used regularly can create more value than a £1,500 platform forgotten after a fortnight.
So start with one task.
Pick the one that takes the most of your time each week.
Find one suitable AI tool.
Use it daily for at least a month.
After thirty days, assess the result honestly.
If the benefit is obvious, carry on.
If it isn't, look for something else.
That steady, measured approach is what gets a small business the most out of AI today.