SELECTED WORK 02 · VOLVO CARS

From AI opportunity
to operating value

Building a repeatable way to decide where AI belongs, what deserves investment, and how proven solutions reach real operations.

Systems thinkingAI portfolioProductionisationOperating ownership
01
Start with the system

Find what limits the outcome before choosing the technology.

In complex organisations, the largest visible inefficiency is not necessarily the most valuable place to intervene. I start with the business system: the outcome that matters, what currently constrains it, and which decision or process is creating that constraint.

“Where can AI be used?” is usually the wrong first question.

The stronger question is: what limits the outcome — and what intervention actually changes it? Sometimes that intervention is AI. Sometimes it is software, process redesign, better data, or no intervention at all.

01UnderstandThe system and outcome
02FindWhat limits performance
03ChooseThe right intervention
04ProveSystem impact
05CommitCapital + capacity
06ProductioniseFor real operations
07LearnReassess the system
02
Evidence before commitment

Increase commitment as uncertainty falls.

A prototype can prove that something can work. It does not prove that the organisation should invest in it. Early learning should be cheap. Larger engineering, data and organisational commitments should follow stronger evidence.

Is it worth solving?Is the constraint material to the business outcome?
Is there enough evidence?Does the intervention improve the decision or process?
Should we own it?Build, buy, enable, stop — or wait?
Can it operate?Workflow, data, reliability, controls, economics and ownership.
03
FLAGSHIP EXAMPLE · CHINA MANUFACTURING

From “automate maintenance” to a second pair of eyes for engineers.

The opportunity initially looked obvious: maintenance was costly, expert knowledge was unevenly distributed, and troubleshooting could be improved with AI. But the goal was not to automate the most inefficient task.

The product vision became a copilot — not a replacement.

The system was designed to surface relevant history, likely causes and troubleshooting information while keeping the maintenance engineer responsible for the decision. The work also made data readiness part of the product problem.

04
Different problems, different interventions

Not every valuable problem deserves bespoke AI.

I used different investment paths depending on differentiation, maturity and the economics of ownership.

ROUTE 01Buy / EnableUse vendor capabilities when ownership adds little advantage.
ROUTE 02Enable / GovernUse shared platforms when broad adoption matters most.
ROUTE 03Build / OwnInvest when capability is strategically differentiated.
05
What I took forward

The goal is not to maximise the amount of AI in the organisation.

01
Start with the system, not the use case.

Understand the outcome before selecting the intervention.

02
Find what limits the outcome.

The largest visible inefficiency may not be the system constraint.

03
Separate feasibility from investment evidence.

A working prototype is not yet a business case.

04
Increase commitment as uncertainty falls.

Learn cheaply before making expensive organisational commitments.

05
Production includes ownership and economics.

Deployment alone is not operating value.

06
Be willing to stop.

Not building can be the right AI investment decision.

The real job is to make better decisions about where AI creates durable operating value.