The maturity illusion
One ladder is carrying four different questions.
The ladder looks orderly only because it hides what is moving independently.
One reassuring answer
Maturity models have a seductive power. They turn messy organizational change into a clean ladder: begin here, climb through these stages, arrive at maturity. The shape offers leaders something reassuring. Movement has a direction. Progress has a score. The future appears manageable.
That logic may work when one capability accumulates in a predictable order. AI asks the ladder to do something much harder.
One AI maturity score is expected to tell a leader whether the investment is creating value, whether the necessary capabilities exist, whether the organization can use them reliably, and whether somebody is prepared to operate and govern what has been built.
Those are four different questions.
A company can be far ahead on one and exposed on another. It can have ambitious use cases with no capability to deliver them. It can build impressive agents that move no meaningful economic measure. It can prove a workflow once without establishing it in the organization. It can establish valuable AI work without deciding who monitors it, who corrects it, or who is allowed to stop it.
Calling all of that “level three” does not simplify reality. It hides the part leadership needs to see.
What is one ladder trying to measure?
When we stepped back from the maturity ladder, four separate views of the company appeared.
The first is value. Where does money move? Does the work increase revenue, remove cost, reduce risk, release capital, or strengthen an economic machine the company depends on? An activity can become faster without producing a material business outcome. Value asks for the causal path, not the promise.
The second is capability. What must exist for that value to become possible? Data, workflow design, identity, integrations, evaluation, recovery, governance, and operating infrastructure may all be required. A use case is not a capability simply because a model can demonstrate it.
The third is maturity. What is verifiably true today? Which practices remain personal, which have spread into teams, which are named and funded, which are wired into real work, and which operate as standing capability? Maturity asks for signs and receipts rather than intention.
The fourth is the operating model. Who runs the capability? Who owns the decision, handles the exception, teaches the system, reviews performance, and remains accountable when the work crosses a consequential boundary?
These views belong to one company, but they do not answer one another. Economic potential does not prove capability. Capability does not prove adoption. Adoption does not create ownership. Ownership does not prove value.
That is why one ladder breaks. It compresses four movements that do not advance together.
Four views of one company
The questions meet in the same organization. They do not answer one another.
Each view asks leadership to inspect a different kind of reality.
What happens when the four do not meet?
The gaps are recognizable.
Value without capability is an ambition. Leadership can identify a costly bottleneck and describe the upside of removing it. If the data is inaccessible, the workflow has no stable boundary, or success cannot be verified, the opportunity remains a slide.
Capability without value is a demonstration. A team can build an agent that searches, summarizes, reasons, and acts. If nobody can show which financial or operating result improves, technical sophistication becomes its own justification.
Value and capability without maturity are a pilot. The system may work in careful hands. One person knows how to start it, what to inspect, and how to recover when it fails. The company possesses the result but not yet the capability.
Value, capability, and evidence without an operating model become orphaned automation. The work runs, but ownership is vague. Exceptions move sideways. Corrections are made but not retained. A human gate exists only because somebody responsible happens to be watching.
A diagnosis, not a grade
Every gap has a different name and a different remedy.
The useful question is not how high the company sits. It is which connection remains open.
None of these conditions means the company has failed. Each identifies a different unfinished connection. The mistake is assigning all four the same maturity label and prescribing the same next step.
What shape does maturity take if it is not a ladder?
This leads to a different proposition.
An organization becomes mature with AI when value, capability, evidence, and operating ownership line up around real work.
The word “line up” matters. It does not mean every workflow becomes autonomous, every capability is built, or every part of the company reaches the same condition.
A routine invoice workflow may run without supervision on modest machinery. A consequential research process may require deeper capability and permanent human review. The second is not less mature because a person remains in the design. The question is whether the operating arrangement fits the value, risk, and nature of the work.
Nor does maturity require every part of the map to be filled. Capability the company does not need can remain absent. Automation that would damage trust can remain unbuilt. A task that properly ends in accountable human judgment can rest there.
Maturity is therefore not maximum AI. It is organizational coherence. The company can explain why the work matters, show that the needed capability exists, produce evidence that it operates, and name the people and agents responsible for running it.
Coherence is not uniformity
The right operating arrangement depends on the value, risk, and nature of the work.
More automation is not the finish line. Fit and inspectability are.
Modest machinery can be mature when the arrangement fits the work.
Permanent human judgment can be part of a coherent design.
One open connection names the next move without grading the whole company.
How can a leader see coherence in a real company?
Begin by walking into the work.
Ask a team which AI-supported activities matter economically. Then ask what machinery makes each one possible. Ask to see the last real result, not the demonstration. Ask who owns the next run, the exception, the correction, and the decision that the system must not make alone.
The work provides the contact point because it is where all four views must meet.
A customer-support workflow may have a clear value measure and working AI assistance. If no dated evidence shows that resolution quality held, the maturity connection is open. A research agent may be well tested and governed. If its output does not change a decision or remove meaningful effort, the value connection is open. An invoice workflow may produce value every day. If one employee remains the only person who can recover it, the operating connection is open.
This is where tasks belong in the model. They are not the subject of the company. They are where claims about the company encounter evidence.
The same discipline applies at larger scales. A department, a customer journey, or an operating function can be examined through the same four questions. Where does value move? What capability is required? What is true now? Who runs it?
Where does the AI roadmap come from?
The roadmap is not the distance between one maturity level and the next. It is the set of broken connections across the company.
When value is clear but capability is missing, the next move is to build or acquire what the work requires. When capability exists but evidence is thin, the next move is controlled operation, measurement, and learning. When the work is proven but ownership is informal, the next move is an operating model. When all three exist but no economic result moves, the next move may be redesign or removal.
This changes investment decisions. A company no longer funds generic movement toward “advanced AI.” It funds a specific connection whose absence is preventing value from becoming durable.
Some connections are inexpensive. A named owner, an exact operating trigger, or a reliable receipt may be enough. Others cross identity, data, architecture, workforce, and governance. Their cost should come from the capability they require, not from the maturity label attached to them.
The roadmap becomes a portfolio of joins to complete, each with a reason, an owner, evidence, and a business consequence.
The leadership view
Replace the readiness score with a ledger of open connections.
The roadmap becomes specific enough to fund, own, verify, or stop.
How can leaders see where the company actually stands?
Remove the colored box that says the company is AI-ready.
In its place, show the important AI-enabled work and four things about each area:
- the economic outcome it is meant to move;
- the capability required to move it;
- the evidence showing what is true today;
- the person, team, or governed agent responsible for operating it.
Then show the missing connection.
One area may have value and capability but lack operating evidence. Another may be proven but economically trivial. A third may create meaningful value while depending on one person's memory. A fourth may have the capability, evidence, value, and ownership aligned well enough to widen its scope.
Leadership can now see the company. Not as an average score, and not as a collection of disconnected experiments. It appears as an operating system whose economic aims, technical capabilities, organizational evidence, and workforce either meet or fail to meet.
The ladder was attractive because it gave one answer. The better model gives four questions that remain distinct long enough to become useful:
Where does the money move? What must exist? What is true today? Who runs it?
AI maturity is not the climb toward more technology. It is the company's growing ability to turn AI into governed, repeatable value.
If you were designing your company for AI today, would you manage that change as a stage to reach, or as a system whose parts must finally line up?