CXaiS Institute  ·  Insight Paper 01

Operational AI Readiness

Why enterprise AI fails before implementation, and how leaders build the foundation for measurable value.

By Rachel Lane, Founder, CXaiS

Executive summary

The argument in one page

Boardrooms have stopped asking whether to invest in AI. They are asking how quickly they can show a return. Yet most enterprise AI work stalls before it reaches the people it was built for. The reason is rarely the model. It is whether the organisation was ready to run AI at scale before it started building.

It then sets out six dimensions that determine operational readiness, a maturity model leaders can use to place themselves honestly on the curve, and the sequence that turns readiness into commercial value. The framework can be applied independently. Nothing in the next ten pages requires a consultant to act on.

The problem

Technology is not the constraint

 

The evidence on enterprise AI is unusually consistent, and unusually stark. In 2025, MIT's Project NANDA studied more than 300 deployments and found that 95 percent of enterprise generative AI pilots delivered no measurable impact on the profit-and-loss account. Only 5 percent created significant value. MIT was explicit about the cause. The divide was not driven by model quality. It was driven by approach.

Other researchers reach the same place by different routes. RAND found that more than 80 percent of AI projects fail, roughly twice the failure rate of conventional IT projects. S&P Global Market Intelligence found that 42 percent of companies abandoned most of their AI initiatives in 2025, up from 17 percent the year before. Gartner expects more than 40 percent of agentic AI projects to be cancelled by the end of 2027, on cost, weak value, and poor controls.

Exhibit 1  ·  The failure evidence

Read those numbers alongside one fact. The technology has never been better. Models improve every quarter, budgets are approved, vendors are selected, and the work still does not reach production. When capability rises and outcomes do not, the explanation has to lie somewhere other than the capability.

It lies before go-live. It lies in the foundation the deployment was built on, and in whether anyone checked that foundation before committing the budget.

The diagnosis

The winners fix the workflow first

 

McKinsey's own research points in the same direction. Only 7 percent of companies have fully scaled AI across their organisations, and more than two-thirds of high performers name data, not models, as the primary obstacle to enabling AI. In McKinsey's 2025 survey, the organisations reporting significant financial returns were about twice as likely to have redesigned their end-to-end workflows before they selected their modelling approach.

The failure modes are predictable, and they repeat across sectors. Organisations automate an inefficient workflow, so AI performs the wrong work faster. They build on fragmented data, so a small flaw becomes a systemic one at machine speed. They deploy without clear ownership or governance, so no one is accountable when the numbers do not move. And they select a platform before defining the business outcome, so the technology arrives in search of a problem.

None of these is a technology failure. Each is a readiness failure. And each is visible before a single line of production code is written, provided someone looks in the right place.

Received wisdom

Four assumptions worth challenging

 

Some of the most expensive AI decisions rest on assumptions that sound reasonable and do not survive contact with a real programme.

The framework

Six dimensions of operational readiness

 

Operational readiness is not a single measure. It resolves into six dimensions, each of which independently determines whether AI will perform in a given workflow. A programme can be strong on five and fail on the sixth. The value of assessing all six is that it surfaces the one that would otherwise have stalled the rollout after the money was spent.

The maturity model

Placing yourself honestly on the curve

 

Readiness is not binary. It is a curve, and most enterprises are further down it than their AI ambition suggests. The following five levels describe what operational readiness looks like in practice. Leaders can use them to locate their programme before they commit to the next wave of spend.

Most enterprises sit at Explore or Build. The distance between Build and Operational is where value is won or lost, and it is an operational distance, not a technical one. No new model closes it. Only readiness does.

The sequence

From readiness to value

 

The dimensions and the maturity model resolve into a single sequence. Each step depends on the one before it, and the order is not negotiable.

Strategy → Readiness → Assessment → Roadmap → Implementation → Commercial value

The point of the sequence is simple. Value sits downstream of readiness, and readiness sits upstream of technology. Run it in this order and AI has a foundation to perform on. Reverse it, and you rejoin the 95 percent.

Applying the framework

From insight to decision

 

Leaders can use this framework independently. The six dimensions are a discipline any capable operator can run against their own programme, and the maturity model is a mirror any leadership team can hold up to itself. If reading this paper changes the order in which you make your next three AI decisions, it has done its job.

Where an organisation wants an independent read, this is the framework CXaiS applies. We assess readiness across the six dimensions, upstream of platform selection, and return a ranked portfolio with a clear verdict per workflow, expressed in commercial terms. The work is deliberately independent of any vendor, because the value of a readiness verdict depends entirely on it being honest. When we say a workflow is not ready for AI, that judgment is based on whether AI will perform there, not on whether anyone needs the deal to close.

See the value. Act on it.

To apply the framework to your own programme, start with an independent readiness read. One workflow, assessed across the six dimensions, with a verdict on where AI will perform and where it will not.