Writing
Long-form writing on governed AI systems, agentic AI in production and the architectural decisions that determine whether technology survives real operating conditions. Written for CTOs, AI leaders, platform teams and business leaders moving from experimentation to production.
The agents nobody authorised
One in five breached organisations reported a breach involving shadow AI (IBM, 2025). Shadow agents are worse than shadow apps — they act at machine speed, hold access indefinitely, and leave no evidence trail to reconstruct.
September 3, 2026 · 9 min read
How would you know it worked?
Only about 5% of integrated GenAI pilots deliver measurable P&L impact (MIT NANDA, 2025). The cause I keep coming back to is the absence of a success metric defined on day one — and agents fail in a way that makes the usual metrics lie.
September 3, 2026 · 10 min read
Your AI pilot isn't blocked by the model
Most CTOs call technical debt their biggest challenge, and 86% of surveyed CFOs say legacy systems limit AI readiness. The reason is more specific than 'legacy systems' — you cannot grant autonomy over an estate nobody can map.
September 3, 2026 · 10 min read
The agent that spends the money shouldn't be the one that judges the spend
Separation of duties is older than AI, and agents dissolve it silently — because nobody reads the union of an agent's tools the way they read an org chart.
September 3, 2026 · 10 min read
Why your AI pilot never shipped
Only about one in ten agentic AI use cases reached production last year (Camunda, 2026). The demo tested whether the model could. Production tests four things the demo structurally cannot.
September 3, 2026 · 10 min read
The hidden cost of "we'll fix it later"
A forensic accounting of deploy infrastructure debt across the four cost categories most CTOs systematically undercount.
April 25, 2026 · 9 min read
Why your engineers are firefighting deploys at 2am
A diagnosis most CTOs miss — and the architectural lever they own.
April 25, 2026 · 10 min read
Six indicators your deploy infrastructure has crossed from working to dangerous
Six observable, binary diagnostic questions for whether your deploy pipeline has crossed from friction into architectural debt.
April 25, 2026 · 10 min read
The staging-to-production gap
Seven dimensions where staging structurally diverges from production — and why your team can't see them from inside.
April 25, 2026 · 10 min read
Already past the reading stage and considering an audit? Here’s how I run them →
Practical Training for AI Builders and Leaders
I run live, hands-on cohorts for professionals and business operators who want to design and deploy AI agents with defined roles, tools, guardrails and human oversight. Explore current training through AIToken Labs, or join the list for upcoming cohort dates and new writing.