Start with the value, then design the work
In my earlier LinkedIn posts, I asked whether we would rebuild the same organisation chart. That is still the question I bring into AI Transformation. A department describes where people sit. An outcome describes what the business needs to achieve. Start with a customer request, a completed order or a released product. Follow it from beginning to end. Where does information wait? Where does somebody copy it into another system? Where does a decision actually require experience?
Take a quotation as an example. The commercial outcome is a useful, accurate response while the customer is still interested. Preparing it may require sales context, delivery capacity, technical expertise and pricing authority. Giving each department an AI assistant can speed up individual tasks. Designing the whole journey can remove the waiting between those tasks. That is a different conversation.
Assign responsibilities to people, agents and systems
I map responsibilities before assigning agents. Who owns the result? Which inputs count as approved information? Which decisions may an agent make? When must a person step in? An agent can collect the request, identify missing details and prepare an offer brief. A commercial owner can retain authority over price and commitments. The system should show what is complete, what is uncertain and who needs to decide next.
This changes the role of the employee. People spend more time defining priorities, judging exceptions and improving the process. They need enough understanding of adjacent roles to recognise when an apparently excellent output creates a problem elsewhere. Deep expertise still matters: broader responsibility gives that expertise more reach.
Use an Agentic Twin to make the alternative visible
An Agentic Twin lets us explore this operating model beside the existing business. I start with synthetic data or approved copies, explicit boundaries and an isolated environment. We can show teams, tasks, handovers and results without granting direct access to production systems or company assets. Any later integration is a separate decision.
The visual world helps people understand the organisation. The actual value comes from the agents, tools, instructions and feedback that perform the work. A moving character is a representation of a role; completed, checked work is the evidence. The useful demo follows a real process all the way to an outcome, including what happens when information is missing.
Build one working loop before scaling
Choose one repeatable process with enough volume to learn from. Record its present completion time, waiting time, rework and human effort. Build an initial loop, run it within agreed limits and compare the outputs. Expand its authority only when the evidence supports it. If it fails, the owner should see the failure and have a clear recovery path.
My ambition is dramatically faster execution and more capacity for growth. The practical first step is specific: choose the work, define good, build the loop and measure what changes. That is how a bold vision becomes a credible operating decision. I help leadership teams make those choices and get hands-on with the systems needed to put them into practice.