Score your company's AI maturity.

We have a written, function-by-function map of how the business actually runs: who does what, in which system, with which handoffs.

AI spend is run as a portfolio with a named owner: some bets to cut cost, some to win work, some to build the foundations, each with a number attached.

We will rebuild a workflow we only finished months ago if a better way now exists, rather than defend it because it was expensive.

Directors use AI on their own work every week. They build with it; they do not just sponsor it.

Leadership meetings review AI progress against agreed measures (hours freed, margin recovered, risk reduced), not anecdotes.

Everyone in the office and on site has an approved AI assistant they use daily, paid for by the company, with our data rules applied.

Proven prompts, instructions and templates live in one shared place, so a good result in one team becomes standard practice in every team.

Meetings, site visits, calls and decisions are recorded and written up, so the knowledge exists somewhere AI can use it.

Human judgement is placed deliberately at the start and the end of a process, with AI doing the middle.

Each department has a named person who owns finding and proving AI improvements in that team's work.

Our documents, drawings, emails, project records and management accounts can be searched and questioned in one place, not across five systems and a shared drive.

The rules of how we work (standards, templates, naming, approval steps) are written down and kept current, so AI can follow them.

Site and field information (photos, diaries, inspections, deliveries, timesheets) is captured digitally at the point of work, in a form AI can use later.

Finance runs continuously: costs, cash and forecasts update as the work happens, not at month end.

Project and commercial decisions are checked against a live financial model, so people see the margin effect before they commit.

Non-technical staff write plain-English descriptions of what a workflow should do, and those descriptions are the brief for anything built.

Subcontractor and supplier documents (RAMS, insurances, accreditations, competence records) are checked and chased automatically, with a person only handling the exceptions.

The high-volume paperwork of our sector (RAMS, valuations and applications, variations, O&M and handover packs, client reports) is drafted first by AI from our own records and checked by a person.

Our routine workflows learn from their own results: the output is checked, feedback goes back in and the next run is better.

AI moves up a ladder of trust for each workflow (observe, suggest, act with approval, act alone) and only after it has earned each step.

Bids, tenders and PQQs start from a maintained library of past submissions and evidence, with AI producing the first draft and people editing.

Estimating reuses our own cost history: past jobs, rates and outturn costs are structured so the next estimate starts from evidence.

We can show a client, on request, exactly how AI is used on their project, which of their data it touches and the confidentiality rules that apply.

We match the AI tool to the job: careful, slower tools for planning and high-stakes documents, quick and cheap ones for volume work, rather than one default for everything.

We measure the hours freed and margin recovered by each AI improvement, per workflow, and compare it with what it cost.

Guardrails come before features: AI can only see what the person asking is allowed to see, enforced at the data level, not by trust.

We keep a standard set of tests for our core AI workflows, so when a new model is released we can check it in a day rather than argue about it.

Legal, HR and IT sit with the people driving AI, so the business is covered without slowing things down.

Cyber and data protection cover AI use: which data may enter which tool, which tools are approved, and how AI-assisted attacks are detected.

Every AI-produced output can be traced back to the source data, the prompt and the person who approved it.