Sign in Subscribe
8 min read

Learn How the Best Work with AI — and How to Train Others to Work Just Like Them (part 1)

I don't know if it's just me, but AI at work right now feels a bit like putting everyone in the cockpit of an F-35 fighter jet with a one-page flying pamphlet, and expecting them to speed out on a mission. The planes work as advertised. And the pamphlet is accurate. But most people are still puttering around trying to get off the landing pad, while a very small number of 'keeners' are flying circles around everyone else.

I notice it in my own team and in pockets across companies I interact with. There is about 2% of people who 'get' AI, are aggressively learning the latest tips and tricks and using it effectively and consistently in their day-to-day. It genuinely helps them be more productive, although it's still difficult to quantify exactly how much more productive. From my observations, these people are usually already experts or high performers in their role — they know exactly what they want to do and how to do it, and if they could just duplicate themselves the team would be humming.

The other 98% are some combination of indifferent, afraid it will replace them, or unsure where to start. Like the personal computer, the internet, and the smartphone before it, everyone will eventually 'get it.' But right now, there's a massive gap between the promise of AI — everything faster, better, easier — and the reality that unlocking those benefits takes real effort, and not just dollars spent. Things won't magically complete in half the time just because your company has access to the latest models and tools.

Here's what I've observed about that 2%: they're not just using AI more. They've developed an intuition for when to use AI and when their own judgment matters more. They know which parts of their work can be delegated or accelerated, and which parts require human creativity, strategy, or relationships. That's the skill gap — and it's not something you can close by reading a prompt engineering guide.

That gap is the seed of a problem I think needs to be solved immediately. The benefits magnify as you advance from solving it as an individual, to a team, to a company, or even as a nation that wants to invest in AI and increase productivity.

The Idea: Build A Coach Trained by Experts to Transfer Their AI Skills to Everyone

One solution I think I can meaningfully contribute to: an AI-powered coach that learns from experts who have mastered the use of AI in their roles — and then helps everyone else work like them.

Imagine this: You're a product manager who wants to get better at pitching ideas to executives. Instead of reading articles or watching videos, you're dropped into a simulation. An AI persona plays your CEO — she's direct, time-pressed, and asks tough questions. She assigns you a real task: put together a 10-slide pitch deck for the board meeting on Thursday.

As you work, a separate AI coach is watching. Not doing the work for you, but asking questions that sharpen your thinking: "What would make the board say no? Have you addressed that?" When you submit, the CEO persona debriefs you — walking through your deck slide by slide, pushing on weak assumptions, showing you how a stronger version might look.

You're not learning about pitching. You're practicing pitching — in a safe environment where mistakes are cheap and feedback is immediate.

The coach doesn't just tell you what to do. It helps you develop judgment about what requires your thinking versus where AI can accelerate the work. For a pitch deck, that might mean:

  • Your thinking: Who is the real audience? What would make them say no? What's the narrative arc?
  • AI accelerates: Competitive research, drafting speaker notes, pressure-testing your logic with tough questions

The goal isn't to make people dependent on AI. It's to make them more capable — faster at the mechanical parts, sharper at the strategic parts.

How It Works

The system has three main components:

1. Training Agent: Your Personal AI Coach

This is what you interact with. The coach adapts to how you want to learn:

  • Guided mode — Step-by-step walkthrough of new AI-augmented skills. The coach explains what to do, shows you how AI can help at each stage, and checks your understanding before moving on.
  • Simulation mode — Practice skills in realistic scenarios with AI personas playing stakeholders (your CEO, a skeptical customer, a resistant engineer). Safe environment to make mistakes and learn from them.
  • Feedback mode — Upload real work for critique before you submit it. The coach reviews your PRD, pitch deck, or analysis and helps you strengthen it — Socratically, so you build the skill, not just get the answer.
  • Shadow mode — AI observes your work silently in your native tools, available for consultation or gentle nudges when it spots opportunities.

The coaching style matters. Unlike tutorials that give you answers, the coach asks questions that develop your judgment. If you ask "How should I size this market?", it might respond: "What approach would be most credible to your specific audience? What data would they trust?" — helping you think through the problem rather than handing you a formula.

2. Skills Graph: The Knowledge Structure

Behind the coach is a skills graph — a structured map of roles, skills, tasks, and decision points. Each node represents something learnable:

  • Roles (Product Manager, Security Analyst, Software Developer)
  • Skills (Pitching Ideas, Writing Requirements, Incident Response)
  • Tasks (Size the market, Structure the narrative, Build the slides)
  • Decision points (Where human judgment is required vs. where AI can handle it)

The nodes of the graph will also include templates and guidance including:

  • Quality rubrics ("A good PRD includes X, Y, Z")
  • Common pitfalls ("Most people skip the 'why now'")
  • AI touchpoints ("Use AI to pressure-test your competitive positioning here")
  • Reference examples (anonymized expert work showing "what good looks like")

This is what makes the coaching contextual rather than generic. The coach knows that "market sizing" is a different skill than "narrative structure," and can give you targeted practice and feedback on each, grounded in leading experts best practices.

3. Expertise Capture: Learning from the Best

The skills graph is built by observing and interviewing experts — the 2% who have figured out how to work effectively with AI.

The capture process extracts:

  • Workflow steps (what they actually do, in what order)
  • Reasoning and heuristics (why they make certain choices)
  • Quality standards (how they know when something is good enough)
  • AI integration points (where they use AI, where they don't, and why)

This can use output from tools like Scribe or Tango, combined with structured interviews that take ~10 minutes beyond normal task completion. The result: decision rules, quality criteria, common pitfalls, and AI-opportunity spots — all captured in a way that can be turned into training.

The Transformation

Here's what this would look like in practice:

Before: A product manager spends 4 hours writing a pitch deck from scratch. They're unsure if it's any good. They send it to their manager, get vague feedback ("make the market size more compelling"), revise blindly, and hope for the best in the meeting.

After: The same PM works through a guided workflow with their AI coach. They spend 20 minutes clarifying what the audience actually cares about — with AI helping them anticipate tough questions. They spend 30 minutes on narrative structure — AI generates three options, they pick and refine. They spend an hour building slides — AI drafts speaker notes, they sharpen the headlines. Before submitting, they run it through a simulated Q&A where an AI board member pokes holes in their logic.

Total time: Maybe 2.5 hours instead of 4. But more importantly: the deck is stronger, they can defend every slide, and they've built skills they'll use next time.

Customization: Train with Your Own Experts

The out-of-the-box skills graph gets you started. But the real power comes from customizing it with your own organization's experts and standards.

Some of the Inputs you can provide to customize it:

  • Captured workflows (video, screenshots, Scribe flows)
  • Past artifacts (docs, PRDs, incident reports, templates)
  • Interview transcripts from your internal experts
  • Your company's rubrics, policies, and quality standards

What you get:

  • A skills graph tuned to how your best people work
  • A coach that references your templates and standards
  • Training that reflects your company's way of doing things

The goal: turn a handful of AI-fluent experts into a system that trains everyone else to work like them. You get a living model of how your best people actually work — and a coach that helps everyone else close that gap, in the flow of work.

Why I'm Building This

I know so many talented and hardworking people right now who are in transition, or nervous about being forced to transition in their careers as current skills and expertise become less relevant (or commoditized with AI). Everyone from software developers, statistical analysts, economists, musicians, professors, researchers, consultants, artists, writers, editors — people with decades of experience across a range of roles and industries. I'd like to help people like that seize this opportunity to get back to the top of their game! Or for other groups (like my teenage kids thinking about future careers) just get into the game in the first place. I feel that I am in a unique position where I have access to the latest technology, the right background, and plenty of reasons to contribute to solving this problem. If I can play some small part start by helping train dozens, then hundreds and potentially thousands of people to be hyper productive as they navigate the shift from the information age to the AI age. At the very least, I feel compelled to try.

I feel confident tackling this problem for at least a few skills where I've achieved some level of mastery. My academic background is B.Math and M.Sc. Computer Science from University of Waterloo and University of British Columbia, plus over a decade in software development and architecture, followed by another decade in product management, individual contributor and leadership roles at tech startups in Canada — spanning cybersecurity, mobile gaming, algorithmic trading, data science, payments systems, proptech and a few failed startups along the way.

I've been pushing my limits with AI since ChatGPT launched in late 2022, and keeping up with the latest capabilities is nearly a full-time job. But patterns are emerging. I've been watching how top enterprises and tech companies are adopting AI, and I believe these patterns can be taught.

I'm starting with roles I know: Product Management, Software Development, CyberSecurity, and Entrepreneurship. And expand with expert colleagues and other masters willing to work with me. My goal is to capture leading AI expertise from people thriving in these fields and build coaches that help others work like them.

AI has rekindled my desire to build. It's making it easier to build software, to build a business, and to share knowledge. It's the same excitement I felt during the dot-com era and again when mobile broadband and smartphones changed everything. I want to spend my energy on something meaningful — and helping people thrive in the AI age feels like the right place and the right time.

What's Next

In Part 2, I'll go deeper on the Training Agent — specifically how the simulation and coaching modes work. I'll walk through a real example: training someone to "Pitch an Idea," from understanding the ask through building slides and defending them in a mock Q&A.

If this resonates, follow along. And if you're one of the 2% — an expert who's figured out how to work effectively with AI — I'd love to talk. The first step is capturing how you work so others can learn from it.

→ Subscribe for Part 2
→ Reach out if you want to contribute expertise or try an early version

Published by: