A bank of toggle switches in mixed states.

enzwa lab · concept

Governor

An AI adoption engagement run at two altitudes: the whole estate of tools seen from above, and one team putting a single one of them to work.

A tool for seeing every AI tool a company is using, and for getting one of them properly adopted.

The problem

AI tools arrive from everywhere: bought centrally, expensed by a team, or opened in a browser tab by one person. Nobody holds the list, so nobody can say what is approved, what is risky, or what any of it is returning.

What it does

Governor works at two heights. From above it shows the whole estate of tools and the state of each one. From ground level it follows a single team through the five stages of putting one tool to work.

what leadership can actually see

The estate is already running. Most of it is running unseen.

Nobody has to be persuaded to adopt AI any more. The average enterprise employee now uses 4.7 AI tools a week and 1.2 of them are approved. The average enterprise carries more than 1,200 unofficial AI-connected apps alongside the sanctioned ones, and eighty-two per cent of CIOs say their people are building agents faster than IT can govern them.

78%

of executives believe they have a clear picture of AI use in their organisation

23%

is what employee surveys put the real figure at

seen and governedrunning anyway

Shadow-AI and tool-sprawl figures: published 2026 enterprise research. Everything drawn from here on is Governor, run on a synthetic portfolio.

why the frameworks do not help

Six frameworks get shown as a menu. They are stages.

Maturity models, autonomy ladders, augmentation spectrums and governance frameworks are nearly always presented side by side, as though the job is to pick the one that suits the organisation.

In a real engagement they sequence. You assess where you are, frame what AI should do and who decides, design the loop, govern the scale-up, then embed it with the team. Each decision compounds into the next, which is why choosing a framework is not the same as having a programme.

Assess where you are

Not the ambition, the position. Where the organisation actually sits, function by function, before anything is promised to a board.

frameworkmaturity ladder

Frame what AI does, and who decides

How far a tool may act on its own, and which decisions stay with a person. Set per use case rather than once for the whole organisation.

frameworksautonomy vs controldecision rights by use case

Design the loop

Where the tool sits in the work: what it drafts, what it checks, what it hands back, and what a person still owns end to end.

frameworkaugmentation spectrum

Make it learn

Feedback has to be a discipline rather than a goodwill exercise, because error and drift compound over iterations and nobody notices until the output is wrong in a meeting.

frameworksthe learning loopfeedback disciplineerror & drift

Govern the scale-up, then embed it

The stage most programmes never reach. Governance that survives more than three use cases, and a teaming model the floor can actually work inside.

frameworksgovernancehuman–AI teaming

the choice that decides the programme

How many use cases you open at once.

Governor scores a portfolio across eight use-case domains, then asks the only question that changes the shape of the engagement. Every other decision is downstream of this one.

portfolio optimisationlocation strategyFM ops & maintenance sustainability managementlease administrationworkplace & occupancy design, fit-out & buildemployee experience

contained

The top three use cases only. Enough surface to prove the model works and to find out what governance actually has to cover.

prove it, contain risk

building

Five at once. Enough running in parallel that the organisation starts to expect it, and enough load to test whether the loop holds.

build momentum

everything

All eight domains together. The fastest route to scale and the widest exposure, taken before the governance from stage five is anywhere near ready.

max surface, max exposure

where programmes stop

The production chasm.

A pilot proves a tool can do the work. Production asks a different set of questions: who owns the output, what happens when it is wrong, which system of record it writes to, and who is accountable in an audit.

Those questions belong to stage five, and most programmes are still congratulating themselves on stage one when they meet them. Governor models the chasm explicitly, so the cost of crossing it is in the plan before the pilot is signed off rather than discovered afterwards.

drop down

Then it stops being an estate and becomes a Tuesday.

The second view follows one team through the same five stages with one tool. This is where a governance model either survives contact with the work or turns out to have been a slide.

Proficiency does not arrive with the licence. It ramps, unevenly, and it dips first: the early weeks cost time rather than save it, because people are learning where the tool is reliable and where it is confidently wrong. A programme that reports on month one kills itself.

swipe the chart →

hours the team gets back hours it costs them to learn it

Illustrative proficiency ramp for one team on one use case. The shape is the finding; the values are Governor’s synthetic portfolio.

what the visibility is worth

Governance is not the brake. It is the thing that lets you go faster than three use cases.

The organisations that can see their estate are not the ones running fewer tools. They are the ones able to approve a sixth use case without a committee, because the questions that stop everyone else were answered at stage two.

$287k

saved a year by organisations with strong AI governance

$400k

spent a year on incidents and lost productivity where visibility is poor

4.7 : 1.2

tools used each week against tools that were approved

Governance saving, visibility cost and tool counts: published 2026 enterprise research. The engagement model, the eight domains and the proficiency ramp are Governor, on a synthetic portfolio.