AI direction & roadmap
Prioritised opportunities, a decision basis and a plan with owners and dependencies.
AI & Automation
AI can move your business forward.
Value has to be built into the work.
I help leaders and teams choose where AI should make a difference, test it on real tasks and establish new ways of working that people can use in their daily work.
From opportunity to shared practice
- Value
- Workflow
- Adoption
Direction · Capabilities · Agents · Ownership
Allan Bau Madsen
Technologist. Adviser. Leader.
Your starting point
Where should AI
make a difference?
Choose a situation you recognise. Different teams may have different starting points. Together, we find the next step that creates value.
Choose another starting point ↑
01 / We are starting
Start with a task that matters.
You see the possibilities but need a shared view of where AI and automation can make a difference. I help connect the ambition with the work your people and customers need to get done.
- The task
- The opportunity
- First trial
How I help
We examine workflows, data and capabilities with the people who know the task. We choose a bounded use case and a responsible owner.
What you take away
A prioritised list of relevant tasks, a concrete first trial and agreed goals for value and quality.
Ready for the next step?
We move forward when the purpose, owner, data and success criteria are clear.
Choose another starting point ↑
02 / We are testing AI
Turn promising experiments into a shared direction.
There is activity and enthusiasm, but it is hard to decide what should continue. I help consolidate the learning and prioritise trials that can improve an entire workflow.
- Experiments
- Evaluation
- Roadmap
How I help
We compare trials on business value, quality, dependencies and effort. What employees learn becomes reusable working methods and clear next choices.
What you take away
A focused AI roadmap, shared evaluation criteria and a plan for training and adoption.
Ready for the next step?
We move forward when a trial works on relevant cases and someone owns the new way of working.
Choose another starting point ↑
03 / We want to embed
Make AI work in everyday tasks.
You have an AI tool such as Copilot or a prototype that needs to become part of the work. I help connect the solution with employees’ tasks, capabilities and responsibilities.
- Workflow
- Learning & ownership
- Daily practice
How I help
We adapt workflows, training and support with domain owners. Where the solution uses agents, we also define the mandate, integrations and knowledge sources and test routine cases and exceptions.
What you take away
A tested workflow, clear owners and a plan for training, maintenance, operations, support and handover. For agents, this also includes a mandate and relevant skills.
Ready for the next step?
We move forward when quality, human oversight, access and ownership work in practice.
Choose another starting point ↑
04 / We want to scale
Scale what creates value for the whole organisation.
Several teams use AI, but solutions may compete for data, capacity and decision rights. I help establish a shared way to prioritise, operate and improve AI.
- Shared foundations
- More teams
- Lasting improvement
How I help
We connect AI initiatives, platform choices, capabilities and decision rights. We agree what to reuse, who maintains it and how to follow outcomes and operating costs.
What you take away
A shared operating model, a prioritised portfolio and an agreed schedule for reviewing outcomes and improving the work.
Ready for the next step?
We expand when value can be assessed and the organisation can sustain operations, learning and the next investment.
How I can help
From defining the work
to taking responsibility.
I can help you choose the first task, lead an AI initiative or take an interim role. You work directly with me. We agree what you need to take away, who needs to be involved and the responsibility I will take.
Adoption & ways of working
Embedding the change with leaders and employees through practice, domain owners and learning through daily tasks.
Explore learning and adoptionAgents & automation
I develop and test prototypes and reusable instructions for agents — skills — with your team.
I connect the business and technical disciplines. Security, legal matters, integrations and production operations are addressed with the relevant owners and specialists.
Practical experience behind my contribution
I use AI myself.
And lead the change.
Selected experience from previous roles and my own working practice. They illustrate different parts of the AI journey.
AI in employees’ daily work
In LEGO Me Chat, I shared product leadership for an AI agent for knowledge and tasks, with an intended audience of 30,000 employees.
The scale describes the intended audience, not a measured number of active users.
Copilot pilot & capabilities
I have initiated and steered the change around Copilot in the daily work of 60 communications employees, focused on developing their AI knowledge.
Focus: AI knowledge and new ways of working in employees’ daily tasks.
From need to prototype
At LEGO, I used AI to build a working prototype for dynamic user segmentation, combining knowledge of the need with UX, product management and architecture.
Delivered as a prototype. Before production use, integrations, code approval and a process for testing and releasing the solution were still needed.
Reusable professional practice
I have built an agent with skills for targeted communication and decision briefs. I use this method to translate complex knowledge into value, choices and success criteria.
My own working practice, including decision briefs and monthly reporting.
LEGO Me Chat · Joint product leadershipA team focused on the employee’s task.View the example +
The task crosses boundaries.
The solution needs to do the same.
In LEGO Me Chat, I shared product leadership with another Product Manager. The team worked on a simpler digital employee experience connecting knowledge with everyday tasks.
Among the solutions we built were meeting and room booking and navigation to the room. For the employee, these are parts of the same task, requiring collaboration across disciplines, systems and responsibilities.
I bring that experience into AI transformation: shared direction, understanding of the task and a team that can turn technical possibilities into something useful.
The employee’s task
- Book the meeting
- Find the room
- Get there
Simplified illustration of the capabilities described.
Context: the team’s product work at LEGO. The example describes built capabilities and shared leadership responsibility.
From ambition to prioritised work
A roadmap with
choices and learning built in.
We follow a concrete task from need to outcome and build the plan around what needs to improve. Scope and pace depend on your starting point, data and capacity.
- 01
Find the value
Follow the task from need to outcome. Identify delays, repeated work, quality issues and decisions that need better support.
Agree: goal, starting measurement and accountable owner. - 02
Test in practice
Build a bounded trial with domain experts. Compare quality, usefulness and total effort with the current way of working on relevant cases.
Decide: continue, adapt or stop. - 03
Embed in the work
Adapt workflows, access, training and support. Give agents, skills and data sources an accountable owner.
Agree: mandate, operations and handover. - 04
Review the outcome
Review the whole workflow: quality, lead time, use and costs. Use the learning to prioritise the next improvement.
Decide: reuse, expand or change.
↶ Learning feeds back into priorities. A roadmap needs to change when reality does.
Adoption is part of delivery
Learning through daily work.
Value for the whole team.
We build capabilities around real tasks. Experience becomes better workflows, sources and examples that colleagues can use.
- 01Choose the taskSomething that matters in daily work.
- 02Try it togetherUse domain knowledge to judge the result.
- 03Share what worksMake examples and skills useful to others.
- 04Improve the practiceFix errors, update knowledge and try again.
learning
Embedding change needs
agreements about daily work.
- Leadership makes room
- Prioritise concrete tasks and time to learn. Clarify what changes and where people still need to exercise judgement.
- Domain owners maintain quality
- Give sources, skills and workflows an owner. Use relevant tests and practical experience to approve and improve changes.
- The team learns and contributes
- Practise on your own cases, share good examples and make help accessible. Feedback needs a path to a concrete improvement.
My contribution is to connect learning with leadership priorities and the team’s work, so AI capabilities become part of the organisation’s own capacity.
Explore in more detail
Agents, automation
and shared knowledge.
Open the topic that matters to you. Explore how technology, accountability and professional knowledge become a useful solution.
Assistant, automation or agent?The task determines the right support. Explore the differences and practical examples.
The task determines the solution
The right support.
The right responsibility.
We choose technology around the work. Often the best solution combines fixed rules, AI and human judgement.
AI as an assistant
When a person leads the task.AI helps research, summarise or prepare a draft. The colleague evaluates and uses the result.
Example: a decision brief with sources and alternatives.
Automation
When the workflow follows defined rules.Agreed steps connect systems and move the work forward. Ownership and failure handling need to be explicit.
Example: an approved case automatically creates a task.
AI as an agent
When the solution needs to choose the next step.The agent uses knowledge and tools to complete a bounded task. Its mandate defines what it may do and when a person takes over.
Example: identify missing information and propose the next action.
What may the agent do?Understand, propose or act — with agreed boundaries and human takeover.
Agents with a clear mandate
What may the agent
do on your behalf?
More autonomy is a choice that needs justification. Select a level to see how responsibility and control change.
I help translate the mandate into concrete requirements with the business, IT and relevant specialists. Ownership, testing and the ability to stop are part of the work from the start.
How does knowledge become shared skills?Methods, sources, quality criteria and a named owner make professional practice reusable.
Expertise that can be reused
Turn your best practice
into shared skills.
A skill describes how to perform a bounded task: the method, examples, quality criteria and boundaries. It makes professional knowledge usable by an agent and visible to the people maintaining it.
I help identify repeatable tasks, turn experience into working instructions and establish testing and improvement with domain owners and technical teams.
How we make knowledge ready for AI
Trusted knowledge needs an owner.
- Sources
- Identify authoritative knowledge, resolve conflicting versions and make sources visible.
- Ownership
- Agree domain owners, review dates and responsibility for keeping knowledge current.
- Access
- Limit and test which information the agent may use for the specific task.
- Feedback
- Make errors easy to report. Improve answers, sources and working methods when something fails.
Optional exampleWhat a skill can look likeView the example+
EXAMPLE SKILL
Prepare a decision brief
- Purpose
- Make a concrete choice clear to leadership.
- Inputs
- Approved sources, audience and decision context.
- Method
- Describe value, alternatives, assumptions and the next choice.
- Quality
- Show sources. Flag uncertainty. Never invent outcomes.
- Ownership
- A domain owner approves changes and checks examples.
Describe → Test → Approve → Maintain ↶
When several teams work with AIMake agents and skills a shared responsibility.See example +
I help create a shared view of purpose, domain owners, access and quality requirements. New versions need testing, changes need to be traceable, and responsibility for operations and improvement needs to be agreed.
Illustrative example of an overview. Roles and boundaries are agreed in your organisation.
Scroll horizontally to see all four columns →
| Task | Skills | Domain owner | Mandate & oversight |
|---|---|---|---|
| Customer overview | Find sources · Summarise | Customer service | Understand: bounded access and visible sources. |
| Draft quote | Pricing basis · Quote checks | Sales | Propose: approval before anything is sent. |
| Routine follow-up | Check mandate · Create task | Process owner | Act: known boundaries, traceability and a path to human takeover. |
Shared practice: version control · testing on relevant cases · named owners · ongoing review
Follow value end to end
Faster at one step.
Better for the whole journey?
An AI-generated quote helps the customer when the business can stand behind it. We also examine queues, quality and handover to the next step.
01Customer need
02AI-assisted proposal
03Expert reviewCapacity? Waiting?
04Agreed delivery
Quality
Errors, rework and the user’s experience.
Flow
Time from need to outcome, including between teams.
Adoption
Use in relevant tasks and the ability to take over.
Economics
Development, integration, learning, operations and maintenance weighed against value.
This is an illustrative workflow. We agree relevant measures and a baseline in your organisation before assessing the outcome.
Let’s begin with your reality
Which task should AI improve?
Tell me where you are today and what you want to make possible. We will clarify the next valuable step and how I can help.