AI for SMEs: where do you start and what does it deliver?

AI for SMEs is not a tool you switch on, but work you organise differently. It is software that takes over or prepares language-heavy and repetitive thinking work: sorting email, reading documents, drafting quotes and answering questions from your own information. For a business of ten to 150 people, the gain rarely comes from a stand-alone chatbot. It comes from hours you get back on processes that are still done by hand every day.

You read everywhere that AI changes everything. In your own business you mostly see a few colleagues using ChatGPT while the work carries on as before. The tools exist and they're affordable. The question is which process you tackle, how you measure whether it works and what you agree around it.

This article is about where to start: which process first, what it delivers in hours, what it costs and which rules apply. Want to know first what AI can do in your business? Then read what can AI do for my business, with five places where AI saves time almost everywhere and a 2-minute AI check. Applications by department are in AI applications for business, and the step-by-step rollout plan is in AI implementation for SMEs.

Which work suits AI in an SME?

AI is good at everyday work involving language and documents: reading, categorising, summarising, drafting and finding information. It's less good at calculations, decisions with serious consequences and work where the data isn't digital.

In a small or mid-sized business that comes down to five kinds of work:

• Reading and extracting: pulling data from invoices, delivery notes, job sheets and forms, so nobody retypes it.
• Sorting and routing: categorising incoming email, tickets and requests by topic and urgency and getting them to the right person.
• Preparing: draft replies, quotes from your own text blocks, meeting notes and reports.
• Finding: answering questions from your own manuals, procedures and files, with the source attached.
• Checking: flagging missing data, unusual amounts and duplicate records before anyone starts on them.

Throughout, a person still decides where it matters. AI prepares, flags and sets things up.

What's the difference between using AI and AI in your processes?

An employee who has ChatGPT rewrite an email is using AI. You only notice it in the hours once AI has a fixed place in a process that runs every day. It helps to separate three levels. They differ a lot in cost, lead time and risk.

Level 1: AI as a tool

An employee uses an AI tool alongside their work: drafting, summarising, translating, analysing an export. No integrations, no IT project. It goes wrong when there's no training and no house rules. Then half the team stops after a few weeks and the rest paste in data that shouldn't be there.

Level 2: configured applications

A recurring task gets a fixed form. A shared assistant that writes quote text in your house style, or one that categorises customer questions. This goes wrong when you scope it too broadly. One task done well works. "Everything to do with customer contact" doesn't.

Level 3: AI in your processes, with integrations

AI becomes a link in a chain of systems. Incoming invoices are read, checked and prepared in the accounts. Email is categorised and routed. Leads are enriched in the CRM. This is where most of the gain is, and where it most often goes wrong on data quality and exceptions. How that works is covered in business process automation.

Most businesses do well to do all three, in this order. Level 1 builds confidence and produces the examples you use to choose levels 2 and 3.

What can AI do by type of business?

Different sectors tend to start with different applications. A few examples:

• Professional services: draft quotes and proposals, meeting notes, making internal knowledge searchable.
• Wholesale and retail: enriching and translating product data, categorising customer questions, processing orders from email and PDF.
• Logistics and transport: document checks, customer status questions, preparing schedules.
• Manufacturing and engineering: structuring job sheets and reports, making maintenance history searchable.
• Construction and installation: preparing estimates, processing job sheets, searching specifications. See also work orders and planning at an installation company.
• Administrative services: invoice and receipt processing, preparing checks, completing files.
• Healthcare and social care: supporting record-keeping and admin, with extra care because special category personal data is involved.

A fuller overview by department, noting where a person stays in charge for each application, is in AI applications for business.

Where should a small business start with AI?

With one process that happens often, mostly runs the same way and whose outcome a person can check quickly. Not with a tool, and not with a whole department.

Four questions help you choose. If a process scores on all four, it's a good first candidate.

• Volume: does the task happen often enough to earn back the setup?
• Repetition: does it mostly run the same way, or is every case different?
• Data: is the input already in a system, or only in someone's head?
• Checking: can a person quickly see whether the outcome is right?

That last one is skipped most often and causes the most disappointment. If you hear "we're already doing something with AI" internally, ask one question back: which process now runs without anyone touching it? If the answer is "none", the gain is still on the table.

To score candidates process by process, use the selection model in which business processes suit AI.

What does AI deliver, in hours?

Work it out per process: how often it happens, how many minutes it takes each time and which share of that disappears. That gives you a number of hours per week to set against the costs.

A fictional worked example, to show the calculation. An admin team processes forty purchase invoices a day, and each one takes four minutes: open it, retype it, match it to the order. That's two hours and forty minutes a day, over thirteen hours a week. If a workflow takes over the reading and matching and one minute of checking per invoice remains, that frees up about ten hours a week.

Two things to keep it honest. An hour freed up is only a saving if the work really disappears. Often it's capacity: the same team can handle more. That's valuable too, but it's something else. And use your own figures, not the ones from an example. The manual work calculator turns hours into an amount, and the business case for AI shows how to work out the payback period.

What does AI cost a small business?

Licences are rarely the big item. The rollout is: configuring an application, integrations with your existing systems, your own team's hours and upkeep afterwards.

The more systems involved, the more it costs. An assistant that writes text is a different thing from a workflow that posts invoices into your accounts. An indication per type of project is in what an AI solution costs.

One lesson almost always holds: businesses that spend months comparing subscription prices lose more to delay than they save on licences.

Which rules apply to AI in SMEs?

The GDPR and the EU AI Act, small businesses included. As a user of AI, it's mainly about not using prohibited applications, helping your people learn to work with AI and being open with people who deal with your AI.

The AI Act (Regulation (EU) 2024/1689) asks organisations that use AI to take measures that support their staff's AI literacy. Since the amendment by Regulation (EU) 2026/1744 you don't have to guarantee a set level per person, but you do need to be able to show which measures you take. The GDPR still applies whenever personal data is involved.

In practice that means one AI policy of a few pages: which tools are allowed, which data may go into them and who is responsible for what. See writing an AI policy, AI literacy and the AI Act and GDPR: what do you need to do?.

Do it yourself or bring in help?

Doing it yourself makes sense at levels 1 and 2, if you have someone with a technical bent and no hard deadline. Bringing in help is wiser for integrations with your ERP or CRM, sensitive data, time pressure or when you don't know which process should come first.

Either way, keep ownership in-house. A partner can build and speed things up, but you judge whether the result is good enough. You don't need your own IT department for that; see implementing AI without an IT department. What to look for in a partner is covered in choosing an AI implementation partner.

What goes wrong most often?

Hardly ever the technology. It goes wrong on choosing the process, ownership and measurement.

• Starting with the tool. You end up with licences nobody uses and a board that concludes after six months that AI does nothing.
• Starting too big. Pick a task, not a department.
• No owner. Board sponsorship is not the same as someone who steers it every week.
• Underestimating data quality. Much of the work is in cleaning up and unlocking data, not in AI.
• Not measuring. Recording one number up front takes an hour and saves months of debate.

More on why pilots stall in why AI pilots get stuck.

When is AI not the answer?

When a process really needs a fixed rule, when the volume is too low or when the process itself isn't right yet. Then a setting in your software, an integration or cleaning up the process first is the better step.

We say so when that's the case. Six situations where you're better off doing something else are in when AI is not a good idea. How to clean up a process first is covered in process optimisation.

Next step

Pick one process, give it an owner and record how it runs today. If you have several candidates and want to know which pays off most, the AI Opportunity Audit lines them up: what the current work costs, which three opportunities deliver the most and what we wouldn't do. If you already know which process it will be, we build it as part of AI automation. How to approach that step by step is in the step-by-step plan for implementing AI.