AI Leadership
Faster Delivery Without Fragility
AI is changing how software gets built, and it’s changing faster than most operating models can absorb. AI is an amplifier. It multiplies whatever system it touches. If your delivery system is clear and disciplined, AI accelerates outcomes. If it’s vague and bottlenecked, AI accelerates confusion.
I focus on responsible, production-grade AI adoption: literacy, guardrails, and measurement, so teams move faster without trading away reliability, security, or trust.
MY STANCE
A few principles I keep coming back to
These aren’t predictions or trends. They’re positions I’ve tested through delivery, adoption programs, and leadership at scale.
1 – Accountability doesn’t move to the model
Teams still own what they ship; quality, security, operability, and outcomes. AI generates; humans decide and are responsible.
2 – AI won’t fix unclear priorities
If coding gets faster and lead time doesn’t improve, the constraint is upstream or downstream: decision latency, dependencies, validation, operability, and ownership.
3 – Guardrails are not bureaucracy
They’re the load-bearing structure that lets you scale speed safely. Without them, every acceleration creates a new category of risk.
4 – AI fluency is table stakes, but not enough
Fundamentals, judgment, and systems thinking are what keep organizations durable. Tooling proficiency without engineering depth is brittle.
5 – Measure the system, not the individual
AI changes output rates; what matters is whether the value stream moves faster end-to-end. Cycle time, not lines of code.
PLAYBOOK
My AI adoption approach
From ad hoc experimentation to responsible production use. This is the system I run, not a theoretical framework.
Define the risk boundaries
What data can be used, where it can go, what cannot leave the environment, and what must be reviewed. This covers IP, PII, regulated data, and customer data. Start here or nothing else is safe to scale.
Choose tools with intent
Don’t treat “AI” as one thing. Select tools based on use case, risk, and cost, then standardize enough to reduce chaos without stifling exploration.
Build AI literacy, not AI hype
Run cohorts. Create internal champions. Teach people how to prompt, manage context, and verify outputs, especially the engineers who feel threatened or skeptical.
Integrate AI into the delivery system
AI is most valuable when it tightens feedback loops across the delivery workflow.
Better specs and thinner slices · Stronger test thinking · Safer PRs and reviews · Faster debugging and incident learning
Add evaluation and review workflows
Treat AI-assisted changes like any other acceleration: repeatable checks for correctness, security, and regressions. Speed without verification is debt, not progress.
Measure outcomes and tune the system
Track what improves: cycle time, change failure rate, rework, incident rates, on-call load, escaped defects, and where the constraints move next.
ENDORSEMENT
What others say
“Bridging Strategy and Humanity in AI Adoption”
Phil brings a rare combination of operational rigor and human-centered leadership to AI adoption. His approach makes the transition practical, accountable, and real.
– Miguel Dias, Value Stream Management Leader
CURRENT THINKING
What I’m exploring right now
I’m actively testing how agentic workflows change planning, execution, and review, and what new leadership behaviors become necessary when implementation gets cheap.
Like many senior technical leaders, I’ve re-engaged with code again. Agentic planning and coordination are moving fast. I still love setting long-term technical direction, but as AI compresses the time to build “good” systems and good plans, even strategy looks different than it did six months ago.
My bias: if the system is clear, AI becomes a force multiplier. If the system is unclear, AI becomes a force amplifier of dysfunction.