Drew Barontini

Product Builder

Issue #85 •
• 15m read

Skill Routing

I’m reviewing engineering résumés for an engineer position we need to fill. After a while, they blend into a single blob of white PDFs listing the same information: contact information, short bio, work history, skills. If you strip away the contact information and short bio, a résumé is really just two things: skills and experience. Here are some things I did using these skills. The better ones will articulate the impact of their work. The worst ones will just list a bunch of technologies without any context. You can only presume about their knowledge and process.

Skills are gained through experience. The skills we pick up are then used to acquire new positions to practice those skills, sharpen them, and attain more skills and more positions. Experience grows and, hopefully, how we’re compensated for said skills, too. More skills means more value for the output.

But how does AI change things?

Does a skill still warrant the same value?

Does a résumé show how you think?

Or how you leverage AI?

Or the quality of your judgement and intuition?

AI is changing everything. It’s noisier in the tech bubble, but the technology is and will fundamentally change and reshape the world.

Knowledge workers are already facing challenges in the nature of work. Most companies were premature reducing their engineering teams for the promise of autonomous agents writing perfect code and sustainable systems. And the debate of who in the product trio—designer, engineer, strategist (PM)—will be supplanted by AI still rages on.

Because we don’t know.

There’s no way to forecast a future full of uncertainty. AI is an agent of change permeating every complex system of society.

Chaos will emerge.

Reading those résumés, I couldn’t decipher what “skills” would matter now. If two people know React and Python, how are they different? Is it their specific experience or amount of time using those skills? And is the best approximation for how they’ll perform in this role predicted by how well they performed in jobs like this one?

Adaptation has always been the key to survival and longevity. I’m only here writing these words because I belong to a species that learned to adapt. Thriving in a world of change requires a new set of skills. That’s why I couldn’t reason with a pile of résumés devoid of any hint of how someone thinks, how they approach problems, or how they adapt to new situations, environments, and technologies.

In Claude, you can create a skill, which is written instructions for how to do something so the AI can repeat a task with the same level of quality each time. If the models get better, then the output of the skill should get better, too. That’s like people. I wasn’t nearly as good at writing React when I started with it in 2015 than I am now. The underlying skill is the same, but I’m better at it. Why? Through repetition and application in real experiences. But it’s not the highest-leverage skill for me to focus on anymore. I’d rather let Claude Code write the code and I can review and refine it. Or I can spend time on the strategic plan that precedes the React component.

Skills, experience, and knowledge are inputs.

Intuition, comprehension, and judgment are outputs.

Skills only matter if you know how to leverage them effectively. You need a process, a perspective, a philosophy.

The idea of skills is interesting. If you think about everything you do as a set of skills, you can externalize work into reusable artifacts. You can make everything visible so you can decide what’s worth your time and energy.

What should you do yourself?

What should you collaborate on?

What should you fully delegate or automate?

Even better: what do you care about, like doing, and want to do more of?

These questions have always been part of the decision-making process of knowledge work.

This is Attention Allocation—finding work that is both high engagement and high expertise. The shift? There’s now an exponential agent of change doing more of the work. It feels more efficient, but there’s an insidious layer behind every new skill subsumed by AI.

Human intelligence develops entropy.

How do we prevent this? Skill Routing

The three pillars are:

  1. Craft yourself.
  2. Collaborate with others.
  3. Command the machines.

Intelligence Orchestration coordinates flow*.*

Human Calibration applies quality.

Attention Allocation focuses your energy.

Operating Cycle moves work forward.

Skill Routing defines the capabilities.

Craft

I love Leonardo da Vinci. For my birthday, my wife got me a now-out-of-print edition of Leonardo’s Notebooks edited by H. Anna Suh, which is a collection of scans and translations taken directly from his innumerable notebooks.

The work comprises more than 7,000 pages of writings and drawings. He died before he could organize them in any discernible order, but the messiness illustrates the depth of his creative genius. He wrote formulas to represent light, perspective, and the human form. He observed nature to model natural phenomena in his art. Art like the Mona Lisa, The Last Supper, and Virgin of the Rocks emerged from a creative process that let ideas develop naturally.

Sometimes an idea comes in a flash; sometimes an idea needs space and nurturing.

While painting The Last Supper, Leonardo would show up in the Santa Maria delle Grazie in Milan to paint all day; other times he would come in, stare, and paint one stroke; other times he wouldn’t show at all. He’s often incorrectly (in my opinion) identified as someone who abandoned a lot of his work. It’s true, but I think it’s a feature not a bug. He was an obsessive perfectionist who was distractible because his mind was constantly sifting through deep thoughts and ideas.

Leonardo knew that ideas need time.

Inspiration is a conduit of energy. And energy is temperamental. You can’t rush something great. So if you want to do great work, you need to develop a deep intuition, explore widely, and iterate relentlessly.

Craft is the first stage of Skill Routing. In the model of Attention Allocation, this is where your Creative Work belongs—the work that is high energy and high expertise.

If AI collapses execution time, then there’s space to get messy. You don’t have to rush an answer without exploring widely. You always diverge before you converge.

Every day I receive AI-generated artifacts with the expectation to review them. If you don’t take the time to write something, why should I take the time to read it? Add a “Written by AI” line like the old “Sent from my iPhone” signatures used to excuse brevity and typos.

Writing is thinking. And thinking is in increasingly short supply. Knowledge work is becoming more about knowledge of AI tools than about using all available intelligence. Blindly following AI is like driving into a lake because the GPS said so. You will lose.

That’s one of many reasons I love Leonardo da Vinci and all the great thinkers throughout history. The human brain contains around 86 billion neurons. It’s the most amazing piece of technology there is. And we don’t even fully understand cognition or how it all works. Craft is where you, us, the humans leverage the miracle traveling around in our skull to do messy, innovative, and creative work.

Spend time on your Craft, no matter what it is.

The Five Craft Skills

The Craft Skills are designed to generate signals, define direction, and shape what’s worth building.

  1. Sense Problems: Observe users, systems, and environments to identify meaningful problems, unmet needs, and tensions.
  2. Frame Problems: Define the core problem by articulating who it affects, why it matters, and what success looks like.
  3. Explore Possibilities: Generate a range of potential approaches, ideas, and directions without prematurely converging.
  4. Model Solutions: Sketch and simulate potential solutions by mapping how they would work, interact, and create value.
  5. Define Intent: Synthesize the problem, constraints, and solution direction into a clear, actionable intent.

Collaborate

Craft is for your work—the work you enjoy doing because it challenges you and delivers impactful outcomes. Collaborate is where you integrate multiple intelligences through collaboration. AI is a collaborator just like your teammates are collaborators. You can’t let AI remove you from the process. Stay involved in lively debate, or be there to fastidiously review its outputs. You’re responsible for the outputs, so make sure they meet your personal standards for quality. Hold the line.

In Attention Allocation, Exploratory Work is the type of work where you leverage the collective intelligence of your team and AI.

And Judgement Work is where you review the outputs and refine them. That applies here because collaboration is a rigorous debate of ideas to catalyze the highest yield.

I was preparing for a user interview. I used the PostHog MCP in Claude to look the user up and create a detailed usage report. It gave me their background, how they’re using the product, and suggested questions to ask.

I created a posthog-report skill to package this up into a reusable process.

After the interview, I returned to share my notes and draft a report of the interview so I could record a Loom to share with the team.

I created an interview-snapshot skill to package this up into a reusable process.

I could have reviewed the PostHog data and compiled the interview report myself. It would have taken time, but it’s important work. It’s not Creative Work. It’s Exploratory Work and System Work combined with Judgement Work to review all outputs.

It was still collaborative. I stayed engaged in the conversation with the user, following the curvature of the discussion organically. The AI usage report preloaded my brain with context before the call, and the interview report let me quickly funnel the information to the team to capture feedback and ignite discussion.

One of the most important parts of Collaborate is comprehension. You must understand every input and output. When you hand work off to AI or another person, you aren’t exempt from understanding the work. To give effective feedback, you have to understand what you’re giving feedback on.

Writing code falls into Collaborate. I’m not writing code by hand, but I am reviewing every line. I ask questions and give feedback because, without comprehension, coherence in the system is lost. And that’s why I’m there—to maintain coherence.

The Five Collaborate Skills

The Collaborate Skills are designed to shape, refine, and align thinking through collaboration—with AI and other humans.

  1. Aggregate Signals: Gather and organize inputs from multiple sources to create a unified view of the problem space.
  2. Synthesize Insights: Identify patterns, themes, and tensions across aggregated inputs to generate meaningful insights.
  3. Shape Solutions: Refine and structure potential solutions into coherent approaches that align with the problem and constraints.
  4. Align Perspectives: Communicate and reconcile different viewpoints to create shared understanding and commitment.
  5. Refine Outputs: Iterate on artifacts and outputs to improve clarity, quality, and effectiveness.

Command

The sales team regularly requests usage data to present to customers. Given GDPR requirements, there are often gaps in the data when organization users deny cookies. But that creates a problem for organization admins who want to understand how their team is using and getting value out of the product. So I came up with a solution to reconcile the usage data in our analytics tool with session data we store in the database. We get verified active user data to pair with a sampling of more granular product usage, which, collectively, creates a cohesive picture of received value.

I later found a solution to anonymously track users within organizations, but that’s not relevant to this problem story.

The sales team comes to me to run the report, which is a Python script I built with Claude Code to pull data from PostHog and the database to generate a PDF and CSV file to present to customers.

This is a perfect example of what I call System Work in Attention Allocation. It requires little expertise from me to execute with low engagement to run the script. This is an excellent candidate for machine work.

Why? Time. My skills aren’t needed and I don’t enjoy doing it. So why would I invest energy without an adequate return?

Craft is where my intuition led me to the signal of what the solution should be.

Collaborate is where I engaged with the sales team and Claude to build something.

Command is where I created the solution to deliver with precision and efficiency.

Command requires Judgement Work to iterate on system outputs. This isn’t much different from the timeless process of finding ways to automate knowledge work. It’s just infinitely more efficient with AI. But the end result is the same: more time, more space, and more energy for the Creative Work.

That’s where you belong.

The Five Command Skills

The Command Skills are designed to execute, scale, and operationalize work through systems, AI, and automation.

  1. Specify Work: Translate a defined solution into clear, structured specifications that can be executed by systems or builders.
  2. Sequence Execution: Break work into ordered steps and dependencies to ensure efficient and coherent execution.
  3. Deploy Systems: Leverage AI, tools, and automation to execute defined work at speed and scale.
  4. Validate Outputs: Evaluate outputs against defined intent, constraints, and quality standards to ensure correctness and effectiveness.
  5. Operate Feedback: Monitor results, capture feedback, and feed learnings back into the system to drive continuous improvement.

The Practice

The practice of Skill Routing is the Skill Registry, a collection of prompts across the lifecycle of Craft, Collaborate, and Command.

There are three parts:

  1. Skill Library: The 15 product skills.
  2. Skill Cards: Each skill as a runnable unit.
  3. Skill Profile: Your own scoring + routing.

Skill Library

Craft (5)

  1. Sense Problems
  2. Frame Problems
  3. Explore Possibilities
  4. Model Solutions
  5. Define Intent

Collaborate (5)

  1. Aggregate Signals
  2. Synthesize Insights
  3. Shape Solutions
  4. Align Perspectives
  5. Refine Outputs

Command (5)

  1. Specify Work
  2. Sequence Execution
  3. Deploy Systems
  4. Validate Outputs
  5. Operate Feedback

Skill Card

Here’s the template for each skill.

Scoring System

Attaching quantitative scores to each skill allows you to rank them. Evolve this to map to yourself and your team.

Scores (1-5)

  1. Strength (S): How strong are you at this?
  2. Energy (E): How much energy does it give you?
  3. Taste (T): How much human judgement is required?
  4. Leverage (L): How much does this matter?
  5. Clarity (C): How easily can this be defined and executed?

Formula

Routing Score = (Strength + Energy + Taste + Leverage + Clarity) / 5

Examples (for me)

Skill Profile

Here’s an example of one of the skills mapped to how I score it and think about it.

Skill: Frame Problem

The Throughline

In a world rapidly accelerating and expanding, slowing down is important. Look at the slow progression of scientific thought for life’s biggest questions about the universe.

From Copernicus arguing that the Sun, not the Earth, is at the center of the cosmos in the 16th century to Galileo and Kepler studying the motion of planets to Newton’s Principia introducing laws of motion and universal gravitation to 19th-century advances in gravity, light, and fields to Einstein’s general relativity to the birth of quantum mechanics, quantum field theory, and the standard model.

These ideas developed slowly over centuries, growing knowledge with each iteration.

While a new onboarding experience in your product is less consequential and grandiose than questions about the universe, it’s still an evolution of information. AI will advance knowledge in every field. I’m working in a space to help researchers advance the world’s knowledge. But we’re not offloading the problem entirely to AI.

Clarity emerges from human thought.

Skills, knowledge, and experience are multiplied by a philosophy about the work to shape intuition, comprehension, and judgement.

Craft is intuition.

Collaborate is comprehension.

Command is judgement.

Taking a big thing and breaking it down into small things is how we reduce complexity.

Skills are atomic units of knowledge work that guide how and where you focus your time and energy. Leonardo da Vinci wasn’t reading productivity books or drawing the Eisenhower Matrix to plan his day. He was mindful, intentional, and curious. He followed his energy and built a rich skillset developed slowly over time.

When you zoom into the lives of most great thinkers, they don’t look productive by modern measures of productivity. But if you zoom out, their output and impact were extraordinary. They shaped the world.

So slow down, focus your energy, and cultivate the skills to move you forward in life. And make sure you focus on the right skills.

Clarity Current Strategic Momentum

Enjoying this issue? Sign up to get weekly insights in your inbox: