The Blocker Is Almost Never the Idea

The same scenario, everywhere

The same scenario plays out from one company to the next, almost word for word. A leader in B2B distribution, with about a hundred employees, has been carrying five or six artificial intelligence ideas for months. The budget exists. A credible vendor is within reach. The teams are fairly willing. And yet, eight months later, nothing has started. No line of code, no test, no task made lighter on the ground.

It would be convenient to see this as a lack of willingness, or a gap in technical culture. It is almost never that. Each idea has remained an intention: "improve customer service," "put our data to use," "automate quotes." An intention, though, does not get launched; it gets discussed again. At every committee meeting, an idea is added to the list, none is decided, and the meeting ends with an agreement in principle that commits no one. The list grows; the company does not move forward.

This blocker is surprising, because the appetite is real enough. Among leaders of French micro-businesses, SMBs, and mid-market companies, use of generative artificial intelligence rose from 31% to about 55% in a year. The desire is not lacking; what is lacking is a first project precise enough to launch.

The real question, then, is not "which idea?" but "what first project can actually be launched, and finished?" Everything that follows lies in that gap, between the ambition that gets stated and the project that gets decided.

The idea is not rare

The first instinct, when nothing starts, is to look for the right idea, the one that would unblock everything. That is looking in the wrong place. AI ideas are not a scarce resource: a motivated company produces ten of them in one meeting. The spread of AI use confirms this, even among smaller companies long thought to lag behind.

In France, in 2024, 9% of companies with 10 to 49 employees use at least one AI technology, compared with 15% for those with 50 to 249 employees and 33% for those with 250 employees or more. The gradient is clear, but the shift is underway everywhere, at every size. What separates a company that moves forward from one that stalls is not the number of ideas: it is their ability to turn a single one into a project.

The evidence also shows up in stated intentions. Among leaders of French SMBs and mid-market companies, 43% say they have defined an AI strategy, and 58% consider AI important, even very important, to their company's survival over the next three to five years. The strategy is set, the ambition is named. The ground, though, often remains untouched. This gap between the stated intention and the first project actually carried out is not a detail of execution: it is the subject.

This point deserves attention, because the misreading is common. It is often said that an SMB suffers from an "idea gap," that it lacks imagination or vision. The numbers say the opposite: the strategy is written, use is growing, the appetite is there. The friction point sits one level down, at the exact spot where an ambition should turn into a bounded project. That step, and only that step, separates companies that move forward from those that pile up notes of intention. Looking for one more idea is like adding fuel to an engine whose problem is not the fuel, but the clutch.

An intention cannot be decided

The apparent cause of the blocker is doubt about the technology: is it mature, reliable, suited to the task? The real cause is simpler, and more unsettling: an intention cannot be decided. An executive committee knows how to decide on a project — a scope, a budget, a deadline. It does not know how to decide on an intention, because an intention has neither boundary nor end.

"Put our data to use" can mean ten projects over three years or a six-week test. As long as the phrase stays vague, everyone around the table imagines something different: one thinks dashboard, another sales forecasting, a third an overhaul of the information system. The meeting ends on an apparent consensus that actually covers three different projects, which means it covers nothing decidable. No one lied, no one blocked anything; the wording itself was undecidable.

This is where the costliest misunderstanding takes hold. AI gets treated as a problem of choice — which idea to pick — when it is really a problem of form: nothing put on the table has the shape of a decision. A topic gets discussed; a project gets decided. Confusing the two condemns every committee meeting to reopen the same debate.

A committee decides on a specific project. Never a vague ambition.

The distinction seems slight; it governs everything. Moving from "put our data to use" to "cut the time spent drafting a quote in half by summer" is not refining an idea: it is changing the nature of the object placed before the committee. The first cannot be decided. The second can be decided, implemented, and finished.

AI magnifies the blocker

AI does not invent this blocker: it magnifies it. A vague intention, without AI, costs one line in meeting minutes. The same intention, once a budget and expectations are attached to it, becomes a topic discussed in committee for months without anything reaching the ground. The broader the ambition, the more comfortable it is: it creates the feeling of progress, precisely because it requires nothing.

The phenomenon is not limited to hesitant SMBs. Globally, 88% of organizations report using AI in at least one function; yet only one organization in three has moved past the pilot stage, and few report more than a 5% impact on operating results. Adoption is massive; scaling remains the exception. Between the two lies exactly the same void as in the committee that cannot decide: many open intentions, very few projects carried through to completion.

The lesson from this gap is direct. The dominant problem is neither the availability of the technology, which is largely there, nor the absence of ideas, which are abundant. The problem is the crossing: moving from the reassuring trial to the committing project. And that crossing does not play out in a tool, but in how a committee frames what it decides.

There is even a trap specific to broad ambition: it feeds on itself. The broader a topic, the more preliminary studies, scoping exercises, and vendor comparisons it calls for — all activities that look like work and that push back the moment of decision. It is possible to spend a year "preparing" an intention without ever having decided it, while feeling reassured about progress. A bounded project offers no such refuge: its deadline is a weekly reminder that, by a known date, a result will have to be shown. That is uncomfortable, and precisely why it produces something.

Stepping down a level

The natural instinct — listing ideas, prioritizing them, scoring them from one to five — solves nothing, because it stays at the level of intention. Prioritizing five vague ambitions produces a vague ambition at the top of the list, not a project. The prioritization matrix is reassuring, but it ranks topics that, one by one, remain undecidable. The next day, the committee reopens the debate at the same point.

What unblocks things is stepping down a level. Stop choosing among topics, and define one specific first project. Not "what is our best AI idea?" but "what is the first thing that can be launched next month and finished by summer?" The question changes scale: it drops the general ambition for a bounded object, dated, assignable to someone. It becomes answerable.

This shift requires accepting a small renunciation, often painful for an ambitious leader: choosing an object more modest than the ambition driving it. Cutting the time spent on a quote in half looks trivial next to "transforming the company through AI." But it is the modest that gets finished, and the grandiose that gets stuck. The larger ambition is not abandoned: it is simply put back in its place, as the direction of the journey, while the first project becomes the first mile actually covered. Nothing is given up in aiming far; what is given up is confusing aiming far with moving forward.

And a first project carried through to completion changes the conversation inside a company more reliably than ten well-ranked ambitions. Internally, proof always carries more weight than a promise. The debate then shifts from "should this go ahead?" to "where should it be applied next?" That shift — from deliberation in principle to a question of extension — is the real sign that a company has crossed the threshold. It is not achieved by discussing better; it is achieved by finishing once.

Proof carries more weight

This path — one bounded, finished first project that opens the way for others — is not abstract. It is documented. Setic Pourtier, an 85-employee SMB based in Roanne that manufactures machines for the cable industry, started with a first predictive AI project. Once that first step was taken, the company redirected its effort toward computer vision built into its machines, to provide continuous quality monitoring of the cables produced. Order matters: a specific project first, extension after — never the reverse.

The finding holds beyond a single case. The "AI Risk-Takers" survey documents about twenty AI projects carried through to completion at French SMBs and mid-market companies, with observed gains of up to 200% on the targeted tasks. Siparex / Bpifrance. What these successes have in common is not a brighter idea than the rest, but a scope tight enough to be closed out.

The gap in outcomes, moreover, does not come from the technology, which is identical for every company. One study measures an average return of $3.7 for every $1 invested in generative AI, and up to $10.3 among the most advanced organizations, for an average deployment completed in under eight months. IDC for Microsoft. Between the first figure and the second, what changes is not the tool, but the discipline: a scope held to, a short deadline, a measurement at the end.

The project in four decisions

A useful first project requires no consulting firm, no oversized committee, no three-year roadmap. It comes down to four simple decisions that any executive committee can make in a single meeting, provided it agrees to step down to the concrete level.

1. Choose one task, a single one. Repetitive and painful — the one teams name themselves when asked what takes up the most of their time. Not a domain, not a function: a task.
2. Measure its current cost. In hours per month, today, before any project. This is the starting point; without it, no result will be readable.
3. Set a dated goal. Modest and verifiable — cutting that time in half, for example. A quantified goal turns hope into a criterion.
4. Give it a short deadline. Six to eight weeks, at the end of which the decision is made: it works and it gets extended, or it does not work and it gets stopped without hesitation.

The quote-drafting task, often cited in distribution and manufacturing, illustrates well what a tight scope can achieve. At a robot manufacturer with a catalog of more than 120 configurable models, the time needed to produce a commercial quote went from eight days to one day. Tacton (Yaskawa). This is not "put our data to use"; it is a named task, a measured time, a verifiable result. The difference is not one of degree, it is one of kind: progress can be tracked, the finish line can be observed, and success can be called or not.

The contrast with the original intention is striking. Where "automate quotes" could sit on the agenda for months, "cut quote-drafting time in half within eight weeks" gets decided in a single meeting. The phrasing is not more ambitious; it is simply decidable. And a short deadline has another virtue, rarely measured: it makes failure cheap. A bounded eight-week project that fails costs eight weeks; an intention that drags on for a year costs a year, with no one ever daring to declare it closed. Brevity is not timid caution: it is what makes trying possible.

An agenda rule

None of this holds if the committee keeps mixing intentions and projects on the same list. The discipline that makes the method possible comes down to one agenda rule, applied consistently. It costs nothing and prevents most of the bogging down.

This rule produces two immediate effects. It clears vague ambitions off the table, sending them back to where they are useful — reflection, vision — without clogging the decision. And it protects the project underway from dispersion, a subtler ailment than technical doubt: instead of launching five projects to "avoid falling behind," one gets finished. Restraint, here, is a method, not timidity.

An objection arises at once: doesn't limiting things to a single project move too slowly when so many possibilities are opening up? Experience says the opposite. Five projects opened in parallel, in a company that has never carried a single one through to completion, do not move five times faster; they compete for the same people, the same attention, the same data, and stall together. One project, finished, moves faster than five simultaneous projects never closed out — and it leaves behind a skill set and a proof point that speed up the next one. Sequence beats a broad front.

This rule also changes the committee's role. It stops being the place where ideas pile up and becomes, once again, the place where projects get decided. A leader who adopts this single discipline will soon find that AI meetings get shorter, that unproductive round-tables disappear, and that something, finally, starts. The rest — tooling, the vendor, upskilling — gets sorted out along the way, once a real project exists in which to sort it out.

What moves forward

The blocker, almost always, is therefore not the idea. It is the absence of a first project that a committee can actually decide on. Ideas abound, the technology is available, the budget can be found; what is missing is the form — a bounded, quantified, dated object that can be implemented and finished.

The day an intention becomes a quantified, dated scope, it stops being a topic of conversation and becomes a project underway. It moves forward. And once a first project is closed out, the company's question is no longer "should this go ahead?" but "where should it be applied next?" — the only debate worth opening in committee.

The right question, for a leader, is not "what is our best AI idea?" It is "what specific first project can be decided this month, and finished by summer?" Answering it with a scope and a deadline already means having started.