The UK government published its AI Adoption Plan for Advanced Manufacturing this month. It proposes a three-stage pathway: Scan, Pilot, Scale. The diagnosis is sound. Pilot purgatory is real. The barriers are real. The competitors are real.
But the pathway has no levels.
The ruler
Technology Readiness Levels changed how defence and aerospace developed systems. Not because the nine-point scale was intellectually profound. Because it gave everyone in the room, engineers, programme managers, procurement, a shared language for where something actually was. TRL 4 means something. It means the same thing in Yeovil as it does in Toulouse.
Manufacturing readiness levels did the same for production. MRL separated "we made one" from "we can make a thousand." That distinction saved programmes.
Both frameworks succeeded because they made progress falsifiable. You could not claim readiness without evidence. You could not skip a level by renaming it.
The gap
Industrial AI has no equivalent.
Firms describe their AI maturity in adjectives. "Early stage." "Exploring." "We've done some pilots." These words carry no information. They do not tell you whether the data infrastructure exists, whether the model is integrated into operational technology, whether the economic case has been validated at fleet scale, or whether a single human being on the shop floor trusts the output.
A plan to escape pilot purgatory needs a way to know when you have left it. Scan-Pilot-Scale describes a direction of travel. It does not describe a position.
The pattern
This is not a novel problem. Every domain that has scaled complex technology through organisations has eventually needed a levelled readiness framework. Aerospace needed TRL. Manufacturing needed MRL. Software needed capability maturity models. The pattern is consistent: narrative pathways stall until they acquire measurement.
The absence of a measurement framework is not a minor gap in the plan. It is the structural reason adoption stalls. Without levels, firms cannot benchmark. Government cannot evaluate programme impact. Supply chains cannot communicate readiness to primes. Investors cannot assess deployment risk.
The instrument
This is why AIRL™ exists. AI Readiness Level is a nine-level framework mapping the journey from initial feasibility to autonomous operation at economic scale. Four stages. Nine levels. Each with defined criteria, evidence requirements, and assessment methodology. Concept, Development, Demonstration, Production.
AIRL™ does for industrial AI what TRL did for technology and MRL did for manufacturing. It makes the question answerable. Where are you? What is the evidence? What does the next level require?
It is not a diagnostic quiz. It is a measurement instrument. The kind that lets a supply chain say "we are AIRL™ 4, moving to 5" and have that statement mean the same thing in Birmingham as it does in Gothenburg.
The multiplier
Government proposes a national front door. A single entry point. That is centralisation dressed as accessibility.
The readiness frameworks that actually scaled were never centralised. TRL did not require NASA to assess every programme. MRL did not require a government visit to every factory. The frameworks worked because they standardised the question, then let every organisation answer it locally, with evidence, against shared criteria.
That is how decentralisation drives impact. Not by removing structure, but by distributing it. A common measurement spine lets a thousand firms benchmark simultaneously without waiting for a lighthouse to be built in their postcode.
The discipline
This is not theoretical. The principals behind AIRL™ deployed AI across machining cells, visual inspection, robotic automation, process planning, and digital twin environments in production facilities in Asia and Europe. Six years before the government published its first adoption plan.
The framework is open. Anyone can use it. The measurement spine is public because the problem is systemic. What is not public is the assessment methodology, the rubrics, the integration logic, and the twenty years of shop-floor pattern recognition that make the levels mean something in a live manufacturing environment.
Manufacturing Engineers build readiness. They qualify processes, validate data pipelines, design the feedback loops that turn first-article evidence into nth-of-a-kind confidence. They are the discipline that sits between "it works in the demo" and "it works on Tuesday, on second shift, with the B-team, on the third batch."
Kaipability builds at this interface. AIRL™ is the measurement layer. The assessment, roadmap, and implementation support are the work. If your AI programme needs a level, not just a direction, start here.