---
title: "AI for Small Businesses: Where to Start (No Data Team)"
description: "No data team, no problem. Here is where SMEs actually start with AI: one process, real numbers, no hype."
canonical: "https://www.whatsnext-ai.com/blog/ai-for-small-business"
published: "2026-08-25T00:00:00.000Z"
updated: "2026-09-15T10:55:27.408Z"
---

# AI for Small Businesses: Where to Start (No Data Team)

No data team, no problem. Here is where SMEs actually start with AI: one process, real numbers, no hype.

![Brightly colored small retail store shelves displaying mugs, tumblers, socks and bucket hats on rainbow pegboard](https://fgzcpjbyiakhjifaciaj.supabase.co/storage/v1/object/public/media/xiao-dacunha-qvbqs6dh7ga-unsplash.jpg)

Do not start with a big AI plan. Pick one process that costs time or blocks revenue, test whether AI genuinely helps there, and build from that win. You need no data team, just one clear process and existing data.

Search for how a small business should start with AI and you get two kinds of answers. One describes the problem ("SMEs are falling behind, the AI divide is widening") without telling you what to do on Monday. The other tries to sell you a licensing engagement or an "AI Council" retainer before you have solved a single problem.

Neither answers the question you actually have: where do I begin, when I have no data team and no data scientist on staff?

This piece answers it. If you want the wider plan around that first step, read [how to build an AI strategy in 5 steps](https://www.whatsnext-ai.com/blog/how-to-build-an-ai-strategy). No abstract maturity models, just a concrete first step, backed by real numbers from our own production builds. And honest about where the cheap route lets you down.

### "No data team" is exactly the right place to start

The biggest misconception in small business is that AI only becomes relevant once your data is in order. A warehouse, a BI tool, maybe someone who understands it. By the time all that exists, the budget is gone and so is the appetite.

In practice it works the other way around. You do not need a data team to begin. You need one well-defined process where the data you already have is good enough. And you have that data more often than you think: in a spreadsheet someone already fills in, in your webshop, in your inbox, in the folders on your OneDrive.

Take [Multi Ratio](https://www.whatsnext-ai.com/cases/how-multi-ratio-cut-time-to-online-from-two-weeks-to-a-day), a nearly 75-year-old Dutch specialist in office furniture and warehouse fit-out, 25 employees. No data team. But a Magento webshop with around 8,000 products, much of it rotating second-hand and refurbished stock. Every new product needed a full product page before it could sell, and that took one to two weeks. Not because the writing took two weeks, but because it depended on three part-timers who each had to be available in turn.

Their data was not "in order" in the technical sense. But the inventory manager already filled out a standardized Excel spec sheet for every product. Consistent and structured enough to serve as reliable input for a system. That was the starting point. No data team, no migration, no new platform. One existing process with usable input.

### Step one: choose the process, not the tool

Most small businesses start from the wrong question. They ask "which AI tool should we get?" when the question is "which process costs us the most time or money right now?"

Look for work that meets three criteria:

- It happens often and follows a recognisable pattern, not once a quarter.
- It costs measurable time, or it holds up revenue until it is finished.
- The input already exists somewhere in a reasonably consistent form.

Concretely, these are the processes small business owners name most often: answering the same customer questions over and over, writing quotes and standard letters, summarising long documents, and re-typing data from one system into another. These are not glamorous AI projects. They are the places where time leaks away.

At Multi Ratio the process was obvious: writing the product page. Description, USPs, standard terms, SEO text and metadata, in the brand's tone of voice. That was the work split across three people. Remove that work, and the reason for the chain disappears.

### Step two: be honest about the data you have

You do not need to make your data perfect. You do need to know whether the data you already produce at the start of the process is usable as input.

At Multi Ratio the test was simple: is the Excel spec sheet the inventory manager already fills out consistent enough to run a system on? The answer was yes. The team was already delivering clean data at the start of the chain, but no one downstream was reusing it structurally.

That is a check you can run yourself, a simple [AI-readiness](https://www.whatsnext-ai.com/glossary/ai-readiness) check. Before spending a single euro on technology: does the input for this process already exist somewhere, and is it reasonably structured? If yes, you have enough to start. If no, your first project is not AI, it is capturing that input in a structured way, and that is usually a matter of a good form, not a data team.

### Why the cheap route gives you a demo, not an asset

Ask an AI model itself how a small business should start, and the answer is predictable: grab a no-code tool like Zapier or Make, click something together, done. That is true, up to a point. For a Friday-afternoon demo it is fine.

But there is a difference between something that works in a demo and something that keeps your business running for years. No-code connections break the moment a screen or a template changes. They live in an account you do not own. And as soon as it genuinely involves your processes and your data, you are locked into a subscription that stops the day you stop paying. And mind your data: sensitive customer or business information does not belong in a free tool you do not control.

That is why we build in code, not low-code. Not on principle, but because the difference is precisely what you are left with. Multi Ratio's AI content pipeline runs inside their own Magento, on their own tone of voice. If What's Next disappeared tomorrow, Multi Ratio would still have a working content engine, not a dead subscription.

That is the trade-off you should make at your very first project. A cheap no-code setup is a fine way to test whether AI matters for a process. But once the answer is yes, you want to build it as an asset, not rent it. To understand that trade-off in depth, read [build versus buy](https://www.whatsnext-ai.com/blog/buy-vs-build-why-custom-ai-is-better-than-saas-subscriptions).

### A human deliberately keeps the last word

A fair fear in small business: if AI takes over, who catches the mistakes? The answer is that AI should not take over. It should do the boring part and leave the judgment to you.

At Multi Ratio the product manager reviews every generated version in a dashboard, adjusts or regenerates it, and only then publishes to Magento. Refurbished stock differs item by item, so the judgment stays with a person. What has changed is that this one person no longer has to wait for two others.

That is what a first AI project should look like. Not a system that publishes unchecked, but a system that does the groundwork and leaves the decision to you.

### What this looks like at scale

Doing one process well sounds small. The effect is not. Look at what happens when you extend this pattern across very different companies.

At Multi Ratio a new product went from one to two weeks to online within a day, 14 times faster, and one person now runs the whole process instead of three. The rotating second-hand stock, once the slowest to reach the webshop, is now the fastest.

At CirFood, in food service and catering, the number of tenders submitted went from 4 to 45 a year with the same team, where a tender used to take 10 days and now takes minutes.

And at Backstage IT, an IT nearshoring firm, each SDR reclaims 16 hours a week, and 40% of the pipeline now comes from opportunities AI surfaces.

Three different sectors, three different processes. The same pattern every time: one clear process, existing data, and a system the company keeps in its own hands.

### How to take the first step, concretely

To sum up, without a data team and without a big plan:

- Choose one process that recurs often, costs measurable time, and whose input already exists.
- Check whether that input is consistent enough to build on. If not, capture the input in a structured way first.
- Test small whether AI genuinely helps this process, in a cheap setup first if you want to prove it.
- Build the version that keeps working as an asset, in your own environment, with a human keeping the last word.
- Measure the win in time or revenue, and use that win to tackle the next process.

That is where our [AI consultancy](https://www.whatsnext-ai.com/ai-consultancy) begins: looking at one process together and whether the data you already have is enough to begin. Want to explore that? [Book a free consultation](https://www.whatsnext-ai.com/contact).

### Frequently asked questions

#### How can a small business use AI in practice?

Start with one recurring process that costs time or blocks revenue, for example answering customer questions, writing quotes, or producing product copy. Use the data you already have as input, and measure whether AI genuinely saves time before you expand.

#### Do I need a data team or data scientist to start with AI?

No. For a first project you need one well-defined process and existing, reasonably structured input, such as a spreadsheet you already fill in. Multi Ratio (25 employees, no data team) started exactly that way.

#### How much does an AI consultant cost for a small business?

It depends on the process and whether you test with a cheap setup or build for the long term straight away. For an honest breakdown of how the costs are structured, see our guide on [how much an AI agent costs](https://www.whatsnext-ai.com/blog/how-much-does-an-ai-agent-cost).

#### Is a no-code tool like Zapier or Make enough?

For a test or demo, fine. But no-code connections break with every change and run in an account you do not own. Once a process becomes business-critical, you want to build it as an asset in your own environment.