AI adoption for SMEs: consultant, free course or playbook? How to choose

An SME that has decided to adopt AI soon realises that the hard question isn't which tool, but whom to turn to. Within a few weeks it receives three proposals that seem to answer the same need and instead compete for the same budget and the same attention: a free course offered by a trade association, the quote from a consultant or an agency, and — rarer, but the direction the market is moving in — a ready-made playbook to graft into your processes. These are different things that resemble each other only on the surface. Choosing between them at random is the most common way to waste the first, precious attempt.

It's worth starting from the context, because it explains why these offers all exist together right now. Italy is among the last in Europe for AI adoption in business — around 18th out of 27 member states — and most Italian SMEs use AI in an isolated and non-integrated way: a few scattered trials, no process genuinely redesigned. Those selling awareness, those selling consulting hours and those selling a reusable method are each, in their own way, trying to close that same gap. Understanding where each one stops is all you need to choose well.

The free course: valuable for understanding, but it stops short of doing

The most visible development of 2026 in Italy is the AI accelerator for SMEs born from the collaboration between OpenAI and Confartigianato — the largest Italian association of artisans and small businesses — with other supporting partners. It's free for admitted member businesses, and the format is one of literacy: hands-on workshops, real-case demonstrations, online modules. The stated aim is to move Italian SMEs from experimentation to structural adoption.

It's an excellent thing and worth taking advantage of. But it's important to read it for what it is: it's training and awareness, not implementation. It teaches you to see the opportunities and ask the right questions; it doesn't build the workflow for you nor put it into production in your company. It's not a substitute for a real engagement — it's what comes before. If anything: an entrepreneur who has done the accelerator and then wants to move to action is a more informed counterpart, not a lost opportunity. The risk is another: stopping there, mistaking "I've understood AI" for "I've adopted it", and staying exactly in the condition of isolated use you wanted to leave behind.

Traditional consulting: three very different tiers under the same name

"AI consultant" is a label that covers at least three different trades, and confusing them is costly. Read for what it delivers — and for how it tends to fail — the offer splits like this:

• The freelancer or micro-agency. They configure the tools for you and connect them. Fast and affordable, useful for a one-off trial. The limit: it ends with the installation, leaving no method behind. When the tool changes or the process evolves, you're back to square one.
• The structured SME partner. They bring the method of a large firm — ordered phases, measurable objectives — at a pace and scale sustainable for a small business. This is the tier where an SME usually finds the best balance between rigour and practicality: enough method not to squander the work, without the machinery of a big consultancy.
• The large consulting firm. Scale, depth and timelines of a big enterprise. For the vast majority of SMEs it's oversized — not in quality, but in pace and budget: a reference ceiling, not a starting point.

For most SMEs the sensible choice lives in the middle tier. But "middle tier" isn't enough to tell a good engagement from a bad one: for that you need two concrete filters, the red flags and the shape of the journey.

The red flags of a bad consultant

Industry analysis agrees on which signals, in a first conversation, foreshadow a project destined to stall. There are four, and they're easy to check before signing:

• They propose the tool before having observed the process. If the solution arrives before the questions about how you work, you're buying a demo, not an engagement.
• They skip training the team. A workflow nobody in the company can govern gets abandoned as soon as the consultant walks out the door.
• They ignore data privacy and security. No questions about where the data lives, about what you feed the model, about who answers for an automated decision: it's the sign that compliance will be discovered too late.
• They don't plan for governance after go-live. If the project ends at "go to production" with no defined controls, monitoring and responsibilities, it's designed for the signature, not to last.

Put positively, this list is also the specification of what a good engagement must visibly satisfy. They are exactly the points that our compliance overlay wires to every workflow we design — not as a downstream formality, but as part of the design.

The shape of a journey that works: phased, with a number at each phase

A serious AI-adoption engagement for an SME, when done well, has a recognisable shape: it typically lasts between three and six months and proceeds through ordered phases — literacy and awareness, then process mapping and discovery, then analysis and solution design, and finally governance and ongoing support. Concrete results usually become visible around the third or sixth month, and industry analyses report — on optimised processes — time savings of up to 40%.

That "up to 40%" should be read honestly: it's a ceiling on selected, well-chosen processes, not a guaranteed average across the whole company. The difference between those who achieve those numbers and those who don't lies not in the tool, but in having chosen the right process and having put method, training and controls around it. Anyone who promises you the high figure without the phases is selling you the conclusion without the work that produces it.

Playbook versus bespoke project: what you're really buying

Here comes the fundamental choice, the one that resembles a classic "make or buy" applied not to the tool but to the method. Traditional consulting, even good consulting, tends to rebuild the same analyses and the same slides from scratch for every client: you pay for the diagnosis every time. The alternative the market is starting to offer is the playbook: a reusable artefact — use cases by department, benefits, risks, governance controls, tool-selection criteria — already built, that adapts to your company instead of being reinvented.

The playbook on its own, though, has the same limit as the free course: it tells you what to do, it doesn't do it. And this is precisely where Innesti's position sits — in the overlap of three things that, in the landscape above, no one puts together: a ready-made playbook instead of consulting rebuilt every time; real implementation — the graft — instead of training alone; and an SME-sized structure of measurable phases instead of the pace and perimeter of a big firm. In a sentence: compared with the free course we are "what comes after the workshop"; compared with the bespoke consultant, "the method is already built, you pay for the graft, not the diagnosis".

How to choose, in practice

Lined up, the three options aren't in head-on competition: they occupy different moments of the same journey. The same road — from understanding to doing, to the method that lasts — and each option reaches a different milestone:

A practical reading for an SME:

• If you're at the start and want to understand — take the free course, if you're eligible. It costs only time and raises the quality of every subsequent conversation.
• If you have a specific and already clear problem — a micro-agency that configures the tool for you can be enough, provided you know it won't leave you a method.
• If you want to adopt AI structurally — look in the middle tier, check the four red flags and demand the phased shape with a number at each phase. This is where the real return lives.
• In any case, start by knowing where you are. Choosing the path is easier — and harder to get wrong — when you know your starting point: which department is ready, which data are in order, which controls are missing.

That last point is the true first step, and we've made it self-serve and free: our AI-readiness assessment gives you, in two minutes, a sense of where it's best to start and with what controls around it. If you then want to understand how to move from understanding to doing, let's talk — no obligation.

This article is for orientation. The adoption figures, timelines and savings cited come from market analyses and from self-reported industry sources, not independently verified: they should be read as indications of direction and not as guarantees of results. The choice of adoption path should always be assessed against the data, the controls and the context of the individual company.