A laboratory of mental models.

AI Workflow Map

Most teams jump straight to tools. This map starts one step earlier: before automating anything, decide what to transform, and where human judgment stays. The failures rarely come from the model. They come from the organization around it.

Two questions, two phases

• Prototype, "can we build it?" Accuracy in a controlled setting, clean data, a single use case. Necessary, but not the goal.
• Performance, "can we sustain value?" Adoption and ROI, real-world resilience, governed and compliant, scalable. This is where value actually lives. The gap between the two is where most projects stall.

A 90-day path

1. Days 1–30, pick the pain. Start from a real business problem, not a pilot for its own sake. Size the value before you build.
2. Days 31–60, build in shadow. Run the AI alongside the human process, so you get credibility and a fair comparison instead of a leap of faith.
3. Days 61–90, go live and tell the story. Document what happened and share it. Credibility comes from the story, not only the result.

Treat AI as an operating capability, not a project: size value before the proof of concept, and govern to accelerate.

Prompt Design Principles

A prompt is closer to a brief than a command. Good prompting is part interface design, part cost discipline.

1. Match the model to the task. Don't use your most powerful model for everything, match capability to what the task actually needs.
2. Give context before instructions. Who it's for, what it's part of, what "good" looks like.
3. Set constraints, not just commands. Define the boundaries of a good answer, and guide the model to be brief where verbosity adds nothing.
4. Manage context deliberately. Reuse the stable part of a prompt, and clear the history when you start something new or unrelated.
5. Optimize for outcome, not price. Think in cost per successful outcome, not just price per token.
6. Keep the judgment. The model proposes, you decide what ships. Make it wrong before you trust it.

Compliance Knowledge System

A control framework is really a model of how an organization believes it works. Most have AI principles, policies and ambitions, but lack the operational ownership, repeatable processes and decision infrastructure that actually govern. Policy alone doesn't govern AI, people and systems do.

Three operational layers

1. Inventory, for visibility. Track use cases, owners, models, vendors, risk classification and business purpose. If you cannot see it, you cannot govern it.
2. Policy, for decision boundaries. Define accepted use, restricted use, escalation requirements and human-oversight expectations. Policy creates rules, operations enforce them.
3. Intake, to govern before deployment. Capture business purpose, AI capability, data involved, risk profile, approval pathway and third-party dependencies. If you cannot intake it, governance starts too late.

What makes it hold

• Clear ownership. Models without an owner don't get maintained.
• Risk-based assessment. Proportionate to each use case, over-governance kills speed, under-governance kills trust.
• Responsible-AI principles in delivery. Fairness, transparency, human oversight, privacy, built in, not a checklist.
• Lifecycle governance. From idea to retirement, most failures happen after deployment.
• Technical guardrails (MLOps). Monitoring, logging, versioning, audit trails, as enablers, not overhead.
• Portfolio-level oversight. Govern the portfolio, not individual models, with shared standards.

It scales through relationships, not documents

• Engineering / Product / IT: what are we building or buying?
• Legal: what must we prove?
• Audit / Risk / Assurance: can we evidence this?

Cognitive Load Checklist

Attention rarely vanishes in one dramatic moment, it leaks. And clarity, not speed, is what protects it. Run this on a typical workday.

• Am I switching tasks more than I'm finishing them?
• How many open loops (unfinished, unparked) am I holding in my head right now?
• Do notifications interrupt deep work, or only at moments I chose?
• Is my environment asking me to remember what a system could hold for me?
• Did I schedule the hardest thinking for my best energy, or my leftover energy?
• Am I confusing being busy with being loaded?
• Where am I paying "attention residue", still half on the last task?
• What's one thing I can close, automate or write down to free working memory?

Reflective Life Design Canvas

A one-page canvas for designing a full life with intention, not optimization. A sentence each, revisited every season.

• Intentions: what does a full season look like, in your own words?
• Constraints: which limits are real, and which are inherited? Which will you design with, not against?
• Energy: what reliably restores you, and what quietly drains you?
• Curiosities: what are you drawn to explore, with no outcome attached?
• Directions: the two or three threads you're holding now. You're allowed more than one, that's multipotentiality, not scatter.
• Enough: what would "enough" look like here, so you can stop optimizing?

Systems Thinking Reading Map

A path through the ideas, roughly in the order I'd read them, because AI doesn't fix complexity, it amplifies it, and you can't run intelligent tools on top of a fragmented organization.

1. Start: Donella Meadows, "Thinking in Systems: A Primer". The clearest door in.
2. See the loops: feedback, stocks and flows, delays. Meadows' essay "Leverage Points: Places to Intervene in a System".
3. Organizations as systems: Peter Senge, "The Fifth Discipline", on mental models and learning organizations.
4. Map it: learn causal-loop diagrams, then draw the system you're actually in.
5. Practice: turn it on your own week or team, where are the loops, the delays, the leverage points, the places power is quietly shifting?

The 9-Strategy Framework

How to build influence and lead without relying on authority, in three movements: earn credibility, build connection, then execute. In that order.

Pillar 1 · Credibility

1. Leverage your expertise. Self-promote without being a know-it-all, share accomplishments naturally.
2. Simplify things. Busy people don't have time to decode complex proposals. Clarity is power.
3. Spread positivity. Regular recognition plus a growth mindset equals magnetic leadership.

Pillar 2 · Connection

4. Build relationships across the org. Active listening, empathy, and being genuinely approachable.
5. Understand your audience. Know their motivations, backgrounds and goals, then frame accordingly.
6. Recognize work styles. Read the room, adapt to brainstormers versus decisive decision-makers.

Pillar 3 · Execution

7. Create an open environment. Transparent communication plus collaboration equals psychological safety.
8. Be a team player. Deliver on promises, offer good ideas, and know when to take a back seat.
9. Expect resistance. Allow grievances without becoming defensive. Resistance often means they care.