Human judgment stays central
AI can support analysis and execution, but people remain responsible for decisions and their consequences.
04Summary
Technology should make people stronger.
My principles come down to one idea: build technology that helps people make better decisions and do better work, without giving up responsibility, safety, or quality.
Trust must be engineered
Accuracy, transparency, privacy, security, and clear boundaries belong in the system from the beginning.
Technology should empower people
Good AI and good leadership expand people’s capability without creating dependency or blind trust.
Quality remains the standard
Speed matters only when solutions remain reliable, maintainable, measurable, and appropriate for their context.
“AI should amplify human capability, not replace human judgment.”
How I build
Principles before prompts.
These principles guide the day-to-day decisions: what to automate, how to review a change, when to simplify, and where a person needs to stay involved.
01
Truth over agreement
State what is known, what is uncertain, and what needs checking. An AI system should challenge a flawed assumption even when agreement would be easier.
02
Human accountability
Give decisions that affect people, money, or safety a named human owner who can review, explain, and overrule the system.
03
Transparency
Make AI involvement visible. Show the sources, limitations, and uncertainty people need to assess an answer.
04
Privacy by default
Collect only the data the task needs. Protect it in transit and storage, and set clear limits on retention and access.
05
Security from the start
Define permissions, authentication, and secure defaults in the design. Keep dependencies and remediation part of ongoing maintenance.
06
Help developers do better work
Use AI to handle repetitive tasks, explore options, and review work. Keep engineers involved in understanding and validating the result.
07
One quality standard
Apply the same review, testing, and security standards to AI-generated work as to anything written by a person.
08
Preserve the freedom to change
Prefer open standards and portable data. Make dependencies clear and keep a practical route to another provider.
09
Match oversight to impact
Higher-impact automation needs stronger review, monitoring, recovery, and ways for people to intervene.
10
Learn from real use
Evaluate actual behavior, collect feedback, and investigate failures. Use that evidence to improve the model and the system around it.
What I optimize for: accuracy, evidence, simplicity, maintainability, human empowerment, security, privacy, open ecosystems, and measurable quality.