Companies aren't ready for AI.

Daniel Miessler, in an essay published on 2 May 2026, makes a case worth restating bluntly: the bottleneck of AI adoption in companies isn't the technology, it's internal coherence. Most organisations don't know themselves well enough to point an intelligent system at something useful. We share the diagnosis. We don't share the cure.

The diagnosis is correct

When an individual professional fails to extract value from a generative model, the problem is rarely the model. The problem is they can't formulate the question. Companies, at scale, do exactly the same thing. Kickoff meetings where nobody can say who the customer is, which three indicators truly matter, who owns which process, how much budget is actually allocated, and what differentiates this company from another fifty doing the same thing. AI needs all those inputs to be hooked to anything real. Almost no one has them ready.

Public data backs the thesis. Industry research from the last two years — Gartner, MIT NANDA, RAND — converges on a narrow band: between seventy and eighty-five per cent of corporate AI initiatives never reach production, or reach it without moving the metric they were started for. Not because the technology is missing. What's missing is a defined perimeter: which metric must change, by how much, in what horizon, under whose responsibility. Without those four variables, any model — even tomorrow's best — produces demos, never systems.

Many companies considered successful are disordered inside. They live off two or three positive accidents that worked for reasons no one ever reconciled: a channel that runs, a legacy customer that pays, a product that found its audience by chance. Revenue covers the noise. Adding automation on top of a process like that orders nothing — it multiplies the confusion. Confused teams don't get clearer with an AI agent; they move faster in the wrong direction.

There's a diagnostic test that works better than any paid audit. Gather the leadership and ask, separately: who is the target customer, which KPIs are we tracking this week, which initiatives are open right now, who owns each, what's the budget. Healthy companies answer in real time, and the answers match. Fragile companies produce five different versions of reality, or silence. Strategic stability over time is another indicator: rewriting priorities every quarter signals incoherence, not agility.

Five concrete signs you're not ready

An operational checklist, drawn from our first conversations with companies that have contacted us in the last eighteen months. If you see three or more of these in your organisation, any AI project will start managed but die unmanaged.

1 · No written ICP (Ideal Customer Profile). An ideal customer profile that lives in two or three people's heads, never moved into an operational document. When a targeting decision arrives (campaign, segmentation, lead prioritisation), it's made by informal vote. An AI agent without a written ICP can't qualify anything — it can only paraphrase the last opinion it heard. Side symptom: salespeople close different profiles and no one has ever consolidated why.

2 · Ambiguous owners on the top-five processes. The five processes that generate half the revenue don't have a single name next to them. They have a team, or "everyone", or "it depends". An automation touches a process, makes a choice, fails: who fixes it? If the answer is a meeting, the automation has already been written to fail. Our experience: without a single decision-maker per process, the twenty-one-day cycle becomes thirty-five, then fifty, then inertia.

3 · Monthly P&L not available in real time. If the CEO doesn't know, on any given day of the month, how much is being billed and how much is being spent, AI has no compass. You can automate things, but you won't know if they've moved anything. Quarterly P&Ls are historical data useful to the accountant, not to the person running execution.

4 · Tech stack that requires three people to change a workflow. When integration between CRM, ERP and email is managed by external vendors who bill by the hour, any change introduced by an AI agent must pass through three ticket queues. This isn't a technical problem — it's an ownership problem. The company doesn't own its operational layer. Without ownership, AI is a guest, not a partner.

5 · Customer data in four or more silos. Marketing has theirs, sales theirs, customer success theirs, finance theirs. None of them match. Reconciling a single customer takes someone a full day. An AI agent asked "what is Customer X worth" will produce four different answers, one per silo. The system is incoherent before AI even touches it.

The competitive threat isn't what you think

The danger isn't that AI replaces confused companies. It's that smaller, clearer, more disciplined competitors can now punch far above their weight. A six-person outfit that knows exactly what it sells, to whom and why, can today orchestrate what required fifty people five years ago. The lever has moved from scale to sharpness. Incoherent companies aren't replaced by AI — they're replaced by coherent rivals using AI.

The pattern is already visible across markets. AI-native marketing agencies, with teams of eight to twelve people, closing contracts that until 2024 were the preserve of international networks with hundreds of staff. Boutique law firms with three partners and two document automation systems competing — and winning — on complex matters, because their marginal cost per new case is near zero. Niche e-commerce operations replacing a twenty-person customer service department with three specialised agents and a single supervising operator. In all these cases the advantage doesn't come from the AI model — any competitor could access the same model — but from the speed at which the small structure integrates AI into its operational flow, because the operational flow is already written and governed.

The temporal asymmetry is the point. A small, clear structure moves the next step in three weeks. A large, incoherent structure moves it in nine months, because every decision crosses four functions that don't talk to each other. When the next step costs ten per cent of what it used to — that's what AI changes, in practice — the company that takes it fifteen times before its competitors compounds an advantage that can't be recovered. You don't lose a market in a year. You lose it in three twenty-one-day cycles while you're doing an audit.

Where we disagree

So far the diagnosis is precise. The usual conclusion, however, slips into a trap we know well: first twelve months of internal alignment, then we'll talk. Workshops, off-sites, cultural realignments, strategic clarity frameworks. It's the security blanket of classical consulting, and it produces exactly the same result it did before AI: voluminous decks, deferred decisions, momentum dissipated.

We saw the same script played out on so-called digital transformation between 2015 and 2022. Companies that consumed three or four years in discovery, in mapping, in governance — and arrived at the execution phase without budget left, without internal sponsors, with a technological landscape that had moved two generations ahead. The pattern fails twice: once for opportunity cost, once for the competitive ground lost to rivals who had already delivered. The same mistake told to oneself a second time, now called "AI readiness", will produce the same two failures. There's no reason to believe otherwise.

Our operational position is the opposite. Clarity isn't reached by retreating into a meeting room. It's reached by putting something real into production. Jigen's twenty-one-day cycle starts with a diagnostic similar to the one Miessler describes — a short, sharp interview with the leadership, the same five or six points — but its purpose isn't to produce a document. It's to identify the one process worth automating first, and put it into production before the company has had time to think it over.

Three weeks of concrete execution say more about a company than three months of internal interviews. When an automation actually touches a process — contacts a customer, qualifies a lead, closes a ticket — unresolved points come to the surface immediately. Who's the owner becomes obvious because someone has to approve. Which KPI counts becomes obvious because it has to be measured. Which customer is really target becomes obvious because some answer and some don't. Clarity is a by-product of execution, not a prerequisite.

That doesn't mean starting blind. It means rejecting the idea that the company has to be perfect before touching a system. Companies don't become coherent by thinking about coherence. They become coherent when operational reality forces them to make decisions that had been comfortably ambiguous. AI, done well, is precisely that kind of concrete pressure.

What happens if you start anyway

An honest note, because it isn't the kind a standard vendor has any interest in writing. If you're at three-signs-out-of-five and start anyway, the sequence is predictable. The first audit produces a clean but generic dossier, because your internal data doesn't allow descent below a certain level of detail. The pilot phase works — you see something running in a demo, the stakeholder is enthusiastic — but the move to production gets stuck on issues you hadn't budgeted for: who approves the agent's responses, where the logs live, who pays the API call that three months in turns out to be the dominant cost driver. At six months, no system in production. At nine, you start over with another vendor.

It isn't fate. It's a consequence. It's what happens when you try to do with AI a journey that with digital transformation went wrong twice. The difference, today, is that the cost of misunderstanding burns faster — because the state of the art moves more, expectations grow more, and the competitor who got it first puts more distance between you.

The minute test

A concrete provocation for anyone running a company and thinking about "doing something with AI". In under sixty seconds, with no consultation, they should answer five questions: who is the customer we want this quarter, which number must go up for the quarter to be a success, which internal process costs more time than average, who's the person leading it, how much budget is in their hand. If the answer comes clean, AI has something to hold on to. If it comes in five different versions depending on who's speaking, the problem isn't AI, and another audit won't fix it. The fix is picking one thing, putting it in production in three weeks, and using that thing as a mirror.

Indicative answers. Target customer — good: "Italian SME, 50-200 employees, manufacturing, revenue €10-40M, already running a CRM in-house." Bad: "enlightened entrepreneurs". Number that must rise — good: "proposal close rate from 18% to 26% in the next ninety days." Bad: "more attention". Hidden-cost process — good: "inbound lead qualification, currently three operators four hours a day." Bad: "internal communication". Owner — good: a first and last name. Bad: "the sales team". Budget — good: a figure in cash, monthly, with spending authority. Bad: "we'll see".

How to prepare in thirty days

For those who have recognised their own signs and don't want to start badly, a practical path. Thirty days aren't enough to rebuild a company. They are enough to bring the four variables — perimeter, metric, owner, budget — to a minimum threshold of clarity that allows a serious vendor to work.

Week one · One-page ICP. Leadership sits down and writes, on one page, who the customer they want this year is. Industry, size, geography, specific problem, buying behaviour, reasons they choose us over a competitor. Not the perfect document. The written document, which is already more than 80% of companies have.

Week two · Single owner for the top five processes. List the five processes that generate the bulk of revenue or cost. For each, a single name next to it, with decision authority. Not a team. Not a committee. A name. If there are ambiguities, this is the week to resolve them.

Week three · Simplified P&L on a dashboard. Even rudimentary: a spreadsheet wired to the ERP, updated weekly, with the three main revenue lines and the three main cost lines. It isn't accounting — it's a compass.

Week four · Single decision-maker for the first AI iteration. One person on the leadership team takes responsibility for the first automation. They have authority to approve scope, to close the project if it derails, to sign off the budget. Without this name, any vendor — including ours — is powerless.

At this point, if the four elements are in place, the company is ready for a twenty-one-day cycle. Not perfect. Ready. The difference is substantial, and it's measured in the months that follow: those who start ready deliver in production and iterate; those who start unready deliver in slides and start over.

Miessler is right: companies aren't ready. But "ready" isn't a state you reach before starting. It's a state you reach by delivering in production. The rest is theatre.

Source: Daniel Miessler, "Most Companies Aren't Ready for AI", danielmiessler.com, 2 May 2026 — read the original ↗. Jigen reading: Miessler's diagnosis holds; the cure the market proposes (alignment first, execution later) is the same one that has already failed a decade of digital transformation. Our position is inverted — it's concrete execution that forces clarity, not the other way around.