The Peruze framework

From AI that helps your people to AI that becomes your business.

Each stage brings a bigger prize (productivity, then capacity, then shareholder value) and asks a harder question. Most organisations are in more than one at once.

1Productivity
2Capacity
3Shareholder value

Stage 1

AI helps your people work

People write, research and analyse faster with assistants and copilots. The value arrives quickly when people are trusted to use AI well.

You are here if

People use AI assistants one task at a time and decide what to do with the answer · AI has appeared in software you already pay for · nobody has a complete list

The board asks

What AI is our organisation already using that nobody approved?

Read: Nobody Owns Your AI Estate →

Stage 2

AI does the work

AI agents (software that completes tasks on its own) take on whole tasks and processes. The organisations that get this right treat agents more like staff than software, with an owner, a defined job and limits.

You are here if

A process or a person sets the task, and AI works out the steps and acts in your systems · teams build their own automations · usage charges appear on invoices

The board asks

If an agent made a costly mistake tonight, who would know, and who would own it?

Stage 3

AI becomes your business

AI shapes your products, services and business model. Organisations that reach this stage are running a different kind of business, and boards start asking what AI does to the value of the business.

You are here if

Customers, or people and systems outside your organisation, instruct your AI directly, or it decides outcomes for customers on its own · AI is part of what clients buy from you

The board asks

When our AI learns our business, who owns what it learned?

What AI costs at each stage

The bill changes form as AI does more.

Each stage changes what you pay for and how you should judge it. The cost sits with the part of the business that decides to use the AI and gets the benefit.

AI spend and the cost of AI →

  1. Stage 1Seats and add-onsJudged by cost per active user, and whether people with a seat use it.
  2. Stage 2Usage, and the people who check itJudged by the cost of each outcome against what the same work costs when people do it.
  3. Stage 3AI inside what you sellJudged as a share of revenue for each product or customer.

Board principles

Decide how far your AI goes.

A stage is set by who instructs the AI and how far it acts, not by the tool. No use should move up a stage without a decision by a named person, on evidence. A use that works beyond its approved stage without that decision has drifted, and good governance shows it.

Above every stage

Your board sets a clear AI policy: the harm it will accept, what AI may never do without a person deciding, what each gate must show before a use moves up, and when it must be told that the business’s AI risk has changed. Management approves individual uses within it.

Stage 1

AI helps your people work

Delivered through AI Usage Oversight

  • Business and client data goes only into approved AI. Unapproved use is found and acted on, and there is a clear route to ask for new tools.
  • Every AI output is checked by the person using it before it is used, and managers make sure that check happens.
  • You know what AI you use, what it can do, what it costs and the value it returns.
Gate out of stage 1 · into stage 2Any agent can be stopped within an agreed time, the work carries on without it, and both have been tested.

Stage 2

AI does the work

Delivered through the AI governance service

  • Every agent has a named owner, a defined job and written limits it cannot go beyond without a person’s approval.
  • What agents do is recorded, their cost is capped, and failures are reported and handled.
Gate out of stage 2 · into stage 3Your incident, crisis and regulatory reporting plans cover AI failures, and have been rehearsed.

Stage 3

AI becomes your business

Delivered through the AI governance service, plus customer-facing controls

  • Outcomes AI produces for customers are explained, evidenced and monitored, so you can account for them to the customer and the regulator.
  • Any critical use of AI can be removed or replaced in a known time and at a known cost, without losing your data, your intellectual property or your choice of supplier.

Your evidence is organised the way ISO/IEC 42001 and the NIST AI Risk Management Framework expect, and mapped to the EU AI Act where it applies to you. How our services put these into practice →

The step from stage 2 to stage 3 is mostly a change to the business itself, and only partly to its systems.

Find out which stage you are in, using your own business.

Book a 30-minute walkthrough
Not ready to talk? Take the board questions into your next meeting. The AI questions your board should be asking →