Three families, nine packages.

Format

two to three weeks · in-person interviews · in-person validation session

What happens

• Interviews with decision makers, typically between six and ten conversations.
• Mapping the decisions that actually move the business, and those that only consume time.
• Selection of one to three indicators that already existed in the management report.
• Measurement of the current baseline, with data, not estimates.
• Validation session with the board, in person.

By the end, the organisation can

• Identify the decisions currently being made on insufficient information.
• Choose one to three business indicators that AI has to move.
• Measure the baseline for those indicators, without estimates.
• Prioritise where to start, and justify that order to the board.

What you keep

Target Charter, a single page with indicators, baseline and definition of success, validated and signed.

Format

one day in person · prior assessment · risk, compliance, IT and business in the same room

What happens

• What the regulation requires, who it applies to and on what deadlines.
• Inventory of the AI systems in use across the organisation, built during the session.
• Classification of those systems by risk level.
• Prohibited practices and what has to stop immediately.
• Transparency and literacy obligations, and what they mean in practice.
• What to do over the next ninety days.

By the end, participants can

• Distinguish the obligations that fall on providers from those that fall on deployers of AI systems.
• Classify the systems in use across the organisation according to the risk levels in the regulation.
• Identify the practices the organisation has to stop immediately.
• Prioritise the actions for the next ninety days, each with a named owner.
• Justify to the board what requires investment and what does not.

What you keep

Preliminary inventory of the systems in use, risk map and ninety-day plan.

Format

quarterly cycle · individual sessions with each member and group workshops · in person

What happens

• Individual sessions, to work through the concrete decisions in each area.
• Group workshops, to align criteria among those who decide.
• Cases applying AI to team management and to running the company.
• Critical reading of supplier proposals, with real proposals on the table.
• Internal communication of the change, and what to say to those who are afraid.

By the end, participants can

• Distinguish market noise from what changes the business of the organisation.
• Assess an AI supplier proposal without depending on the supplier to interpret it.
• Decide where to invest, what to postpone and what to refuse, with explicit criteria.
• Design the internal message about AI, including what is said about the effect on people.

What you keep

Board decision agenda for the year, with priorities and investment criteria.

Format

half a day in person · up to 40 participants · teams of four to six people

What happens

• Short framing: what an agent is and what it does not do.
• Each team picks a real problem from its day-to-day work.
• Building the agent, with support in the room.
• Testing, fixing and preparing the presentation.
• Final presentations, with voting and a gamification element.

By the end, participants can

• Identify, in their own work, a process with potential for automation.
• Build a working agent from that problem.
• Test the agent and fix what does not work first time.
• Present the result and estimate the gain for the team.

What you keep

One agent built per team, and a prioritised list of the ideas left undone.

Format

two days in person · up to 15 developers · work in a real team repository

What happens

• Agent-assisted development methods, and where they fail.
• Specification: how to write a request that produces usable code.
• Critical review of generated code, focused on security and dependencies.
• Tests and verification of what the agent produced.
• Integration into the existing workflow of the team.
• Defining the house rules on what is not delegated.

By the end, participants can

• Apply an agent-assisted workflow to a real repository.
• Write specifications that produce usable code without a full rewrite.
• Review generated code, identifying security flaws, performance problems and undue dependencies.
• Measure the effect on the cycle time of the team.
• Decide, as a team, which parts of the work are not delegated to an agent.

What you keep

Documented workflow and the team rules on the use of agents.

Format

four to six weeks per function · observation, redesign, build and training · in person

What happens

• Observation of the real work of the team, on site.
• Mapping and measurement of the current process.
• Redesign of the process, with the team, not for the team.
• Building the solutions and agents required.
• Training the team on the new process, not on the tool.
• Drawing up the list of what is not automated, with justification.

By the end, the team can

• Describe and measure the current process, with numbers.
• Run the redesigned process without outside support.
• Identify what was deliberately left human, and justify why.
• Measure the effect on the indicator the organisation chose.

What you keep

New process documented, solutions in production, team trained and the list of what is not automated.

Format

project · scope defined case by case · in-person sessions at the decision and delivery points

What happens

• Framing the problem with the people who do the work, not only with those who describe it.
• Designing the solution and defining the success criteria, before writing code.
• Building, with intermediate validations.
• Testing with real users, in real work.
• Going into production and transferring knowledge to the internal team.

By the end, the organisation can

• Run the agent in production, inside the existing work process.
• Change and maintain the agent with the internal team, without depending on the supplier.
• Measure the effect on the indicator defined at the start.

What you keep

Solution in production, technical and operating documentation, and an autonomy plan.