Selected writing on AI governance, business transformation, technology leadership, architecture and operations. Most pieces are published on LinkedIn, where the conversation happens; this page curates them by topic.
AI & GovernanceBusiness TransformationExecutive LeadershipArchitecture & TechnologyOperations & Business Performance
€825 million. That’s the fine Uber received for automated decisions.
Who is responsible when AI agents make decisions that affect people? Agents need defined boundaries that do not depend on a system prompt alone — and logging is not explainability.
Read on LinkedInOnce e-commerce is significant it changes the operating model. A better website cannot compensate for fragmented inventory, disconnected systems or fulfillment that does not fit.
Read on LinkedInA new ERP doesn't make a retailer faster. A new POS doesn't either.
The real measure of transformation is how quickly the business can change its mind. Modern architecture on an unchanged operating model just lets a slow organization run on newer technology.
Read on LinkedInWho is responsible when AI agents make decisions that affect people? Agents need defined boundaries that do not depend on a system prompt alone — and logging is not explainability.
AI dependency becomes business risk when teams cannot operate without the tool. Decide in advance what continues, what switches provider, what falls back to the old process and what stops.
Adopting AI mainly to cut costs widens the attack surface and breeds shadow AI. Why automation needs architecture, permissions, governance and continuity planning from the start.
Many AI products are thin layers over someone else's model, pricing and roadmap. A useful test for any company buying or building on AI: what do you own beyond model access?
The real measure of transformation is how quickly the business can change its mind. Modern architecture on an unchanged operating model just lets a slow organization run on newer technology.
The hardest legacy is often organizational — tribal knowledge, spreadsheets as an operating system, unclear ownership. Replacing technology without changing the operating model buys a more expensive version of the same company.
Transformation has to start from how the company makes money, where work slows down and who decides — which is why the modern CIO sits between the CEO, COO, CFO and technology.
Private equity starts AI from business value pools, not AI strategies or pilots. The modern CIO should think like an operating partner: baseline, business ownership, process change, then technology — measured on the P&L.
Revenue protected, margin, working capital, customer service, risk removed and decision speed — not uptime and tickets. A new ERP is not an outcome; better inventory turns might be.
Eight things a CEO should hear from a CIO in business terms: growth, risk and continuity, real ROI instead of decorative innovation, operational visibility, and technical debt expressed as business risk.
In fast-moving technology environments indecision is often more dangerous than a wrong call. Practices for deciding with speed and judgment when the data will never be complete.
Age is a poor measure of technology risk; complexity that outpaces the organization's understanding is the real danger. Modern and resilient are not synonyms.
Where AI genuinely helps with mainframe and legacy modernization — code analysis, data handling, testing, migration risk — and why it is not a standalone solution without experienced people to interpret it.
How language models combined with retrieval-augmented generation (RAG) can personalize customer interactions, building on earlier omnichannel customer-service work.
Once e-commerce is significant it changes the operating model. A better website cannot compensate for fragmented inventory, disconnected systems or fulfillment that does not fit.
Heroics don't scale. Store openings need repeatable standards — configuration, validated integrations, security baselines, monitoring and fallback — so growth carries less uncertainty each time.
Register revenue records where a sale is booked, not what the store contributed to it. Store KPIs — and the systems that produce them — have to change with the role of the store.
AI is becoming a capital-allocation decision. Not every workload needs the most expensive model, not every company needs to own AI infrastructure, and not every attractive AI project deserves capital.
AI deserves the same capital discipline as any major investment: where it earns funding, how value is measured, and when to stop — decided by the CIO together with the CEO and CFO.
Follow new writing on LinkedIn.