Strategy
AI strategy and operating model
I set what gets built, who owns it, which data it may use and how it is funded, before assistants or agents reach production.
Enterprise AI
How I run Enterprise AI.
How I make AI work inside large organizations, and what that looks like in practice.
Perspective
Most enterprise AI programs do not fail because the technology is inadequate. They fail because the organization cannot say who owns the output, which data the system is allowed to see, what happens when it is wrong, or how the result reaches the people doing the work.
The questions that decide the outcome are rarely about model selection. They are whether a decision can be traced, whether a control still holds, who is accountable once a step is automated, and whether anyone's work changed.
This is why the gap between a pilot and production is wider than it looks. A demonstration needs one convincing example. A production system has to be right on the unremarkable ones, every day, inside processes that are audited and workflows people already depend on.
Generative AI and agents are more capable and less predictable at the same time. That combination makes governance more valuable.
None of this is an argument for moving slowly. It is an argument for building AI where the work already happens, with the controls that work already has.
Practice areas
Delivery
AI inside core business processes
I design AI around named workflows: master data, approvals, planning, reporting, service and the audit trails that connect them.
Governance
Responsible and privacy-aware adoption
I specify model risk, data sensitivity, access design and human oversight during implementation, and test them the way any other control is tested.
Executive priorities
Accountability
Who owns the output
Every AI-supported decision has a named owner, a defined escalation path and a record of what the system saw when it produced the result.
Measurement
What it produced
AI portfolios reported against cost, cycle time, error rates and compliance outcomes, on the cadence the rest of the business already reports on.
Controls
What happens when it is wrong
Failure modes identified before deployment, with the oversight, controls and rollback that the surrounding process already requires.
Newsletter
Between the Lines
Insights on AI, data and automation beyond the hype, uncovering real impacts hidden between the lines of technology.
- Published monthly on LinkedIn
Editions