AI readiness
Asks whether the data, infrastructure, governance frameworks, and skills are in place to deploy AI. It is mostly a capability and tooling question.
Category · Definition and diagnostic
AI implementation risk. The risk your organisation can't absorb the decisions your AI generates.
AI implementation risk is structural, not technical. It is what stalls AI programmes after the model works and the pilot lands. The decision architecture diagnostic - WorkLattice - is built to find it before it becomes a board problem.
AI implementation risk is the gap between what a working AI can produce and what the organisation around it can actually absorb, ratify, and act on coherently.
It is distinct from model risk (will the AI work?) and from AI readiness (do we have the data, tooling, skills?). Implementation risk sits in the human architecture the AI plugs into: authority, definitions, enforceability, escalation. It is the load-bearing layer that decides whether a working AI lands or stalls.
Where it sits
AI readiness and AI implementation risk are not the same thing.
AI implementation risk
Asks whether the organisation can absorb the decisions a working AI produces. It is a structural question - about authority, definitions, enforceability, and the human architecture that has to receive the outputs.
Symptoms
Four signatures of unmanaged AI implementation risk.
01
Pilots succeed; scale doesn't.
The pilot delivered a working model and credible numbers. When the rollout begins, local teams stop trusting central outputs, escalation paths get clogged, and the explanation drifts toward "adoption problem". The structure that absorbed the pilot is not the structure being asked to absorb the scale.
02
Recommendations arrive faster than they can be ratified.
Model outputs land in the operating layer at machine speed. The authority required to act on them is human-speed: ratification chains, sign-offs, working groups. The gap widens, queues form, and either the recommendations are ignored, or they are acted on without authority and the organisation absorbs the risk silently.
03
Shared terms mean different things to different functions.
Qualified, approved, complete, escalated. Definitions coexist quietly for years because human-speed coordination smoothes the differences. AI imposes one definition at speed; the disagreement surfaces at the execution layer and looks like the system is broken.
04
Rules that exist on paper are routinely overridden.
The control framework says X must precede Y. In practice, political weight regularly overrides it. AI does not see the override pattern and runs the rule as written. The exceptions that used to be invisible become incidents, and the people who used to manage them have no authority to intervene in machine-speed flows.
Structural causes
Why AI implementation risk is structural, not behavioural.
The instinctive explanation for a stalled AI programme is behavioural: people aren't using it, the change management was thin, the comms didn't land. Sometimes that is true. More often, the behaviour is rational given a structure the AI exposed but didn't cause.
Three structural conditions reliably predict whether the programme will land: decision rights that are clean enough in practice to build on; shared terms that mean the same thing across functions; and rules that hold under political pressure rather than only on paper. When any of those three is weak, an AI deployment magnifies it rather than absorbing around it.
Sagentivum is an independent Sydney-based advisory practice that diagnoses AI implementation risk through a graph-based decision architecture analysis called WorkLattice. The diagnostic gives a sponsor a structural read on whether the programme they're about to scale, pause, or rebuild has the conditions it needs to land.
AI didn't create the gap between what your organisation knows and what it can do. It removed the slack that used to hide it.
Human-speed decision-making absorbed structural incoherence for years. AI runs at machine speed against a structure built for human speed, and the gap that was tolerable becomes load-bearing and visible. That is why a category of risk that was previously dispersed across "change", "operating model", and "governance" has now consolidated into something worth naming and diagnosing in its own right.
Worried your AI programme has structural risk you can't see yet?
A thirty-minute conversation tests whether a decision architecture diagnostic fits your situation, the structural questions worth answering, and how the work would be scoped.