AI Fluency
How I teach AI, advise on adoption, and help organizations make sense of what to do next.
AI education isn’t a tool training problem. It’s a judgment problem. People don’t need to know how every model works — they need to know how to evaluate output, structure context, manage trust, and decide when AI is the right answer. My work is built on the premise that the human always gets the final call, and the job is to make sure they’re equipped to make it well.
The Clarity AI Index
A diagnostic for AI fluency. Two dimensions, four capabilities, twelve levels — designed to surface where individuals and teams sit today, and what to invest in next.
Awareness
Know how AI works and where it fails
Craft
Direct and build with AI
Judgment
Evaluate output and spot failure
Responsibility
Own outcomes and sustain trust
Most AI training lives in one quadrant. Real fluency lives across all four.
Your progression in AI knowledge — from understanding what AI is, to taking action with it. This is the AI maturity curve. It moves from learning what AI can do, to using it deliberately, to designing systems with it.
How you build a working partnership with AI. The machine side: understanding how AI works well enough to direct it. The human side: judging what it produces and owning the outcome. A trustworthy partnership requires both.
Each capability has a maturity ladder.
Most people aren’t at the same level across all four. The gaps tell you what to build.
The Index spans from foundational AI literacy through expert-level building and evaluation. An executive learning to evaluate AI proposals, a product manager directing AI workflows, and an ML engineer designing agent systems all sit somewhere on this map — usually at different levels across the four capabilities. The Index doesn’t teach ML research or deep model engineering itself; those are their own technical curricula. What it does is help individuals and organizations see where they are, where the gaps are, and where to invest.
Find out where you land.
A short assessment to map your current fluency across the four capabilities. Twelve questions, about three minutes. The result shows your level on each capability, the profile you most resemble, and where to focus next. See all six profiles →
Twelve questions. About three minutes. You’ll get a level on each of the four capabilities and a recommendation for where to focus next.
Who the Index serves.
AI fluency looks different by role. An executive deciding on an AI investment, a manager rolling AI out to their team, and an ML engineer building agent systems all need fluency across the four capabilities — but at different levels, with different emphases. Six common personas, with where each typically starts and where they need to grow.
Executives & Leaders
Typical starting pointSurface awareness, reactive judgment.
Growth edgeStrategic judgment, principled responsibility.
Focus areasGovernance, decision frameworks, AI investment evaluation, organizational posture.
Managers
Typical starting pointOriented awareness, varying judgment.
Growth edgeDiscerning judgment, intentional responsibility.
Focus areasTeam adoption, coaching, setting AI norms, performance and feedback in an AI-augmented workflow.
Manager fluency increasingly overlaps with IC fluency as ICs learn to manage agents and AI workflows. The capabilities of delegation, evaluation, and coaching apply to both human teams and AI systems.
Knowledge Workers
Typical starting pointUninformed-to-oriented awareness, prompting-level craft.
Growth edgeCalibrated awareness, directing craft, discerning judgment.
Focus areasEveryday fluent use, prompt design, output evaluation, in-flow-of-work integration.
Includes GTM, sales, marketing, content, HR, finance, operations. Solopreneurs and small business owners typically blend this with the Manager path.
Software Engineers
Typical starting pointOriented awareness, prompting-to-directing craft.
Growth edgeDesigning craft, strategic judgment.
Focus areasBuilding AI features, evaluation frameworks, working with agents, MCP, RAG, and the broader AI tooling stack.
ML/AI Engineers & Scientists
Typical starting pointCalibrated awareness, designing craft.
Growth edgeUsually judgment and responsibility — the human side.
Focus areasEvaluating real-world outputs, owning downstream impact, communicating with non-technical stakeholders.
Technical depth doesn’t guarantee fluency. ML engineers who can design sophisticated systems may still paste AI-generated output into a deck without reading it. Building powerful systems is one skill. Judging what they produce — and owning the consequences — is another.
Data Professionals
Typical starting pointStrong awareness of evaluation, varying craft with LLMs.
Growth edgeDirecting-to-designing craft with AI tooling, strategic judgment for AI-augmented analysis.
Focus areasData preparation for AI systems, evaluation design, AI-assisted analysis workflows.
Programs.
Programs are how individuals and teams develop fluency. Each maps to one of the four capabilities in the Index, and each can be delivered as a workshop, a cohort, or embedded into a broader curriculum.
Foundations — AI Literacy
A facilitated session introducing how generative AI works, where it sits in the broader machine learning landscape, and how to interact with it effectively. Covers the evolution from traditional computing through machine learning to generative AI, single vs. multimodal models, and hands-on practice with text and image generation. Designed for non-technical and semi-technical audiences moving from "I’ve heard of ChatGPT" to "I can use AI deliberately in my work."
Human Instinct — Judging AI
A facilitated session that teaches participants to evaluate AI output the way an analyst evaluates data. Participants prosecute a flawed AI-generated engagement survey summary, learning to spot three failure modes — flattery, gap-filling, and puffery — and building a six-point Analyst’s Checklist (Factual, Cuts, Contradictions, Limits, Action, Audience). The session uses Claude Projects with system instructions and memory to teach four techniques for controlling AI output: instructions, memory, tools (apps, MCP, and skills), and style control.
“You’re not looking at AI. You’re looking at what you gave it.”
Building with AI
A hands-on program for builders moving beyond prompting toward designing AI systems. Covers Claude Projects, custom instructions, memory and context management, tool integration (MCP and skills), and the four-technique framework for reliable, repeatable AI output. Participants leave with their own working AI project, a reusable validation checklist, and a model for evaluating new AI tools as they emerge.
AI Consulting.
The Index handles fluency. But organizations face harder questions that don’t fit neatly into a framework: Should we build this AI capability ourselves or use a vendor? Which of these hundred new AI tools are worth our attention? How do we adopt AI without creating chaos? What’s the difference between using AI, integrating it into our product, and actually building models — and which do we need?
I help leaders think through these questions. Not with a fixed methodology — every organization’s posture on AI is shaped by its product, data, talent, and risk tolerance. The work is structured thinking partnership: surfacing the right questions, mapping the real tradeoffs, and helping leaders make decisions they can defend.
Common areas of work.
- Build vs. buy.When to develop proprietary AI capability vs. integrate vendor models vs. simply use tools off the shelf. Each has different cost, risk, defensibility, and capability implications.
- Tool stack strategy.Cutting through AI tool sprawl to a coherent set that reinforces rather than competes. Distinguishing tools worth adopting from tools that overlap with what’s already deployed.
- Adoption & change.Getting an organization to actually use what’s deployed. Building internal capability, growing power-user networks, navigating resistance, setting governance and norms.
- AI posture.Distinguishing between using AI (internal productivity), integrating AI (vendor models in your product), building AI (proprietary models), and researching AI (frontier work) — and matching each posture to the right use case.
The human always gets the final call. The job is to make sure they’re equipped to make it well.
Book a conversation