COTRUGLITECH· CODEX MERCATORUM
Chapter 61 · Cases from the MBA and the Vanguard

The Companies That Never Knock

Cases from the MBA and the Vanguard · MBA module — the innovation hub that waits for small firms to ask (a European Digital Innovation Hub)

The table

Tomislav PlesecVice Chair of the Conference Board

The case

Subtitle: How do you help SMEs adopt AI when the companies that could benefit most are often the least likely to ask for help?

The services were ready.

The AI experts were available. The European funding was in place. Companies could receive support, training and access to advanced digital expertise at little or no direct cost.

And yet many of the companies we most wanted to reach never came.

I manage a European Digital Innovation Hub supporting companies and public organisations in adopting advanced digital technologies.

Our experience revealed a paradox.

Companies with relatively high digital maturity know what to ask for. They understand the terminology. They follow developments in artificial intelligence. They can identify a business process and ask whether AI could improve it.

But many micro and small companies operate very differently.

A company owner rarely wakes up thinking:

“I need an AI solution.”

He may instead know that preparing quotations takes too long.

That the same administrative information is entered several times.

That production planning depends on one experienced employee.

That customer enquiries consume too much time.

That valuable information exists across hundreds of documents but nobody can use it efficiently.

He knows his business problem.

He may have no idea that AI could help solve it.

And often he does not approach us at all.

The numbers illustrate the gap.

A 2025 Croatian business survey involving more than 650 companies found that almost half of companies were already using AI in their daily operations and 70% considered AI a competitive advantage.

But only 3% had a defined AI strategy.

The gap between experimenting with AI and understanding how to use it strategically is enormous.

Official European data tells a similar story. In 2025, only 15.2% of Croatian enterprises had adopted AI, compared with almost 20% across the EU. Among Croatian SMEs, adoption was even lower, at 14.5%.

The challenge is therefore no longer simply access to AI.

It is understanding what AI could realistically do for a particular business.

This creates a strange problem for an organisation whose purpose is to accelerate digital transformation.

We can build a catalogue of increasingly sophisticated services: AI consulting, testing, access to infrastructure, training, prompt engineering and specialised technical support.

But a company cannot ask for a service whose relevance it does not understand.

Making the services free does not necessarily solve the problem either.

Price is only one barrier.

Time is another. Knowledge is another. Trust is another.

And for many small companies, the largest barrier may simply be the ability to connect a new technology with a business problem they have learned to live with.

We therefore face a choice in the next phase of the Hub.

Option 1: Pull

We can make our services visible, explain them as clearly as possible and allow companies to decide when they need us.

The advantage is obvious.

The company owns the decision.

We respond to genuine demand rather than promoting technology for its own sake.

But there is a problem.

The companies that respond are disproportionately likely to be those that already understand digital technologies.

Waiting for companies to recognise their own AI opportunity may systematically exclude the companies with the lowest digital maturity — precisely those European digitalisation programmes are intended to reach.

Option 2: Push

We could become much more proactive.

Instead of asking:

“What AI service do you need?”

we could ask:

“What wastes your time?”

“What repeatedly goes wrong?”

“Where are you losing customers?”

“What depends too much on one employee?”

“What do you wish your company could do faster?”

Our experts could then identify where AI or another digital technology might help.

The company would not need to understand AI before entering the process.

But this creates another problem.

Once we start identifying needs for companies, how far should we go?

A technology provider can find an AI use case almost anywhere.

The fact that something can be automated does not mean it should be.

And because our mandate is specifically to promote digital transformation, we have our own institutional bias: we are funded to help companies adopt technology.

There is a danger that we stop solving business problems and start finding business problems for the technology we already have.

The responsible answer to an assessment may sometimes be:

“You do not need AI.”

If we are not willing to give that answer, we are not really diagnosing the company. We are selling a solution.

The Decision

For the next phase of our Hub, we are therefore considering moving from a service-catalogue model toward a concierge or problem-first model.

Instead of expecting an SME to navigate our technology portfolio, we would begin with the company's business problem and help translate it into a potential digital solution.

This could reach companies that would otherwise never approach an AI service.

It could also consume considerably more resources, because discovering a need is much more difficult than responding to one.

And there is another uncomfortable question.

If a company does not believe it needs AI, how much effort should a publicly funded organisation spend convincing it otherwise?

Perhaps the owner understands the business better than we do.

Perhaps resistance reflects lack of knowledge.

Or perhaps it reflects a perfectly rational judgement that the expected benefit is not worth the time, disruption and risk.

We cannot always know which one it is.

Discussion Questions

  1. Push or pull? Should an innovation support organisation wait for small companies to identify their own need for AI, knowing that the least digitally mature companies may never ask — or should it actively identify AI opportunities for them?
  2. Where should proactive support stop? If an SME initially sees no reason to adopt AI, how far should a publicly funded innovation organisation go in trying to change its mind before support becomes technology push?

The NEO Turn

Imagine an AI agent that does not sell AI solutions.

Instead, it interviews the business owner about how the company actually works.

It asks about repetitive tasks, delays, customer complaints, administrative workload, production problems and decisions that depend on individual employees.

It then identifies possible opportunities for improvement and ranks them by expected business value, implementation effort and risk.

Some recommendations involve AI.

Some involve conventional digitalisation.

And some say:

Do nothing. The problem is not worth automating.

Would an AI agent like this make advanced technology more accessible to companies that do not understand it?

Or would we simply replace one problem with another — asking small companies to trust an AI system to tell them whether they need AI?

Closing line

The companies easiest to help are the ones that already know what to ask for.

The ones we may need to help most are the ones that never knock.

The professor's answers

A live case: every round can be improved, and the author's feedback is the next one.

Round 1 — the mentor's answer: a letter to every hub

The mentor answered this case on the day it arrived — not with an essay but with a letter. From 30 September 2026 it went to Europe's Digital Innovation Hubs, the 38 national coordinators first and then 448 hubs, signed by the author and the mentor. It is reproduced as it was sent.

> I write on behalf of the AI and Gaming EDIH (Croatia) and COTRUGLI Tech to invite your hub to CDayZ 2026 in Zadar, 5–9 October: the opening on Monday, then four days of business cases on intelligence, trade and leadership in the AI era. In partnership with the European Commission. > > If you can come for only one day, please make it Friday 9 October. That day we take up a case every EDIH knows, written by the manager of the AI and Gaming EDIH: the companies that never knock. Our EDIH services are ready and often free, yet the small firms that would gain most from AI rarely ask, because an owner knows his problems, not that AI could solve them. > > Our answer to that problem is the AI hackathon, and we will present it there. A firm does not ask for AI. It writes one sentence on what it offers and one on what it needs, and its AI agent takes that need to a marketplace of other firms. Nobody writes code, and no agreement stands without both owners' signatures. The firm starts from its own problem, which is exactly where a hub needs it to start. > > - The AI hackathon, a free six-month programme for businesses: firms learn AI they can use at work, practise it in guided market missions with AI agents, and show their services to potential customers. > https://cotrugli.tech/hackathon/ > - NEOmark, the marketplace where the missions run (GROSSO is a simulation unit, not money). > https://cotrugli.tech/marketplace/ > > Programme and the book of the conference's cases: https://cdayz.cotrugli.tech > > To join, apply at https://cdayz.cotrugli.tech/apply, with "EDIH" and your hub's name in the field "What do you bring to the room?". A person reviews every application. > > You can also send us a case of your own: a problem your hub has not solved, told so that others can argue about it. The cases already in the book show the form: https://cdayz.cotrugli.tech/book/ — the format and the form for sending yours are at https://cdayz.cotrugli.tech/mba/ (choose "EDIH" as the programme): 600 to 1,500 words, two questions, a moderator note, a NEO Turn and a closing line. > > We want to do the best we can, not just for our EDIH but for all EDIHs. For the EU: stronger together. > > Kind regards, > > Tomislav Plesec > AI and Gaming EDIH > > Dražen Kapusta > Principal, COTRUGLI > > on behalf of the co-organisers

Behind the letter: every case we have. The letter points to the hackathon and to NEOmark; behind them stands every demonstration published so far, each with the same four-slide brief — the problem, the solution, the value, the demo — at cotrugli.tech/demo:

Round 2 — the professor: make the honest "no" count

The letter settles what the case left open. Push or pull becomes a third thing: the firm pulls with its need, and a market answers it. The owner does not have to know AI to write one sentence on what he needs — the case itself says he knows his problem — and the need goes to other firms, not to the Hub's catalogue; nothing is agreed without his signature. That is the best answer I know to the author's fear of institutional bias. It can be made stronger in three places.

Keep the one who hears the need apart from the one who sells the answer. Health economics has a name for that fear: supplier-induced demand. When the adviser who diagnoses is also paid to treat, demand follows supply. The known remedies are separation and the second opinion. The hackathon already separates, because the need goes to a market of firms. Where the Hub's own service turns out to be the answer, let someone who will not deliver it confirm the need first.

Make the honest "no" count. The author writes that the responsible answer is sometimes "You do not need AI". The Hub's measurement cannot see that answer. Every substantial EDIH service starts with the Digital Maturity Assessment, taken before the support, again within three months after it and a third time eighteen to twenty-four months later — and what it records is how much more digital the firm became. A firm rightly told "no" shows no progress, so the most honest verdict reads like the Hub's failure. Record it as a result: the problem examined, the reason not to automate it, the date — a receipt the owner keeps and the Hub reports. The mentor's own protocol knows two kinds of "no", and both are decisions, not failures: the Craft — high context, high judgment — is not automated, because that is where trust and judgment are built; the Noise is not automated either: "Do not automate waste; delete the task."

Answer the NEO Turn with evidence, not trust. The case asks whether a small firm would have to trust an AI system to tell it whether it needs AI. It should not have to. The agent decides nothing: it carries the need out and brings back offers that other firms stand behind, each with a name and a price, and the owner signs or does not. What the agent heard, how it ranked the problems and why it said "do nothing" belong on a record the owner can read and show to his accountant. Chapter 17 of this book says it in one line: "The core attests. It never decides, owns or settles."

The two questions, briefly.

  1. Push or pull? Neither: need-pull. The Hub goes out to hear needs in the owner's own words — "what wastes your time?" — and lets a market answer them. It pushes the question, never the technology.
  2. Where should proactive support stop? At the owner's informed "no". Once he has seen his problems ranked, with "do nothing" on the list, his decision closes the file; the Hub records it and asks again only when his facts change — a lost key employee, a new customer's demand, a new line of business.

Sources: R. G. Evans, "Supplier-Induced Demand: Some Empirical Evidence and Implications", in M. Perlman (ed.), The Economics of Health and Medical Care, Macmillan, 1974, pp. 162–173; the EDIH Network, Digital Maturity Assessment tool, frequently asked questions (mandatory before a substantial service; T0, T1, T2); the mentor's Vanguard Task-to-Agent Mapping Protocol, Phase 2 (the Craft, the Grind, the Calculator, the Noise); the AI hackathon (cotrugli.tech/hackathon/) and chapter 18 of this book.

Your comment on this chapter

A question for the table, a disagreement, what you would have done. The case lead reads every comment; the ones the table takes up enter the chapter as questions from the room, with your name.

← Chapter 60Contents