Skip to content
Multigrid

Enterprise adoption & change management

The artefacts an AI programme actually runs on — the scoring rubric, the pilot charter, the budget structure, the risk register — rather than a maturity model.

There is a large literature on adopting AI in a company and almost none of it can be acted on. It is made of maturity models, pillars, readiness scores and unattributed percentages, and its central claim is usually that success depends on culture — which is true, unfalsifiable, and of no help to the person who has to write something down by Friday.

These pages are built the other way round. Each one carries a thing you can take away: a rubric with the questions filled in, a pilot charter with a stopping rule, a build-versus-buy calculation with its terms defined, a rollout whose stages have exit criteria, a risk register with real rows, a budget shaped so that a metered bill and an annual plan can coexist. Where an argument can rest on arithmetic — what a per-seat licence costs at 400 seats against metered usage, why a pilot with no control condition cannot attribute anything to itself — it rests there.

You will not find an adoption statistic anywhere in this cluster. The figures that circulate about pilot failure rates, ROI multiples and productivity gains are nearly all unsourced or built on methods nobody published, and a page that repeats one has borrowed a credibility it did not earn. Where a number would have made the point, the point is made without it.

Why Most Corporate AI Pilots Never Ship

Why the failure-rate figure you came for is not knowable, what actually stops a pilot from becoming a decision, and the six-field charter that fixes it before the pilot starts.

6 min read

Choosing Your First AI Use Case

A weighted scoring rubric for picking a first AI project — measurability, data readiness, failure cost, blast radius and value — with two candidates scored end to end.

5 min read

Build vs Buy for AI Features

The four layers the build/buy line can be drawn at, the cost terms each option carries, and the seat-count at which a per-seat licence stops being cheaper than metered usage.

6 min read

Proving ROI on an AI Project

The measurement design that has to exist before the pilot starts — one primary metric, a baseline window, a comparison condition and a pre-registered analysis — plus the full cost denominator.

5 min read

The Pilot-to-Production Gap

The four things that change between a demo for twenty people and a feature for ten thousand, and a readiness gate with the specific evidence each row requires.

5 min read

Getting Legal and Security to Approve an AI Project

What a security and legal review is structurally trying to establish, the one-page data flow it actually wants, and the question list with the artefact each question is asking for.

6 min read

AI Vendor Due Diligence: A Checklist

A vendor checklist with pass conditions rather than topics — data terms, commercial terms, continuity and exit — plus the clauses that only matter once you depend on the thing.

5 min read

Change Management for AI Rollouts

A five-stage rollout with numeric entry and exit criteria per stage, including the support-load and cost-per-user constraints that decide when the next stage is affordable.

5 min read

Training Your Team to Use AI Well

A three-hour internal workshop with four exercises, each aimed at a specific failure — miscalibrated confidence, no verification habit, prompt superstition, and unclear data boundaries.

5 min read

Shadow AI: Employees Using Tools You Didn't Approve

Five places unapproved AI use is already recorded in systems you own, what each source over- and under-counts, and a triage rule that ranks findings by data reach rather than by tool popularity.

5 min read

Writing an Internal AI Policy People Follow

The three-part test a rule must pass to be followable, six vague rules rewritten into checkable ones, a one-page template, and the exceptions process that keeps the policy honest.

5 min read

Centre of Excellence vs Embedded Teams

The five things an AI function can centralise, why three of them centralise well and two do not, and the numeric triggers that say a structure has stopped working.

5 min read

Data Readiness: Are You Actually Ready for AI?

A five-check assessment you can run in a week, one check per day, with the specific query or test for each and a scorecard that says which use case to pick instead when it fails.

5 min read

Procurement: Budgeting for Usage-Based AI Costs

Why an annual fixed budget and a metered meter fight each other, a three-part budget structure that lets them coexist, and the enforcement architecture that makes the number binding.

5 min read

Measuring Productivity Gains Honestly

Why a self-reported time saving cannot answer the question, three study designs ranked by what each can rule out, and the five confounds to pre-register before anyone is switched on.

5 min read

AI Governance Inside a Company

A four-question tiering rule, what each tier is required to produce, a published response time the board is measured on, and the pre-approved patterns that keep most work out of the queue entirely.

5 min read

Risk Registers for AI Systems

The columns an AI risk row needs, eleven filled rows covering the failures specific to model-backed systems, and a scoring scheme that avoids inventing precision it does not have.

5 min read

Retiring an AI Feature Nobody Uses

The annual carrying cost of a live AI feature, the arithmetic that compares it with what the feature returns, three endings other than deletion, and a dated sunset runbook.

6 min read

Communicating AI Changes to Customers

Three announcements companies routinely confuse, a trigger table saying which changes require notice and on whose clock, and a message skeleton with the sentences that cause trouble.

5 min read

The Realistic Timeline for an Enterprise AI Project

Why engineering estimates miss by months on these projects, a worksheet that separates work time from queue time, and the three serial constraints that cannot be parallelised away.

5 min read

Other topics

Enterprise adoption & change management · Multigrid