The AI Skill Every Working Adult Needs to Learn

I sat down with Mark Cleveland on the Parallel Entrepreneur Podcast for an 80-minute conversation that covered almost everything I care about: learning, music, community, generalism, and the shift AI is forcing on every company and every working person. Mark was in the first cohort of Pioneer Species and came to Un Poco de Planning in Mexico City, so he knew exactly where to push.

Here are the main ideas, roughly in the order we got to them.

Curiosity is the common thread

Mark opened by asking what connects music, software, education, and AI. My answer: curiosity and learning. I want to figure things out, and once you've figured something out, the natural next step is to teach it. You can build businesses around that teaching process. Almost everything I've started came from there.

The intermediate plateau

Every skill has the same shape. The jump from zero to roughly 60% capability is fast and exciting. Then you hit what I call the intermediate plateau, where you can put in a year of real work and feel like you're spinning your wheels. In guitar, about 90% of players stop there: open chords, barre chords, the pentatonic scale. Good enough for a campfire.

Getting past it means going back to fundamentals. It's like cooking. Recipes get you pretty far, but to actually be inventive you have to understand the principles. On guitar that means scales, dyads, triads, arpeggios, seventh chords, inversions, and how the instrument actually maps, then applying that architecture across styles. After five years of studying music around the world, the thing that struck me most is that the principles are the same everywhere. The rhythm and idiomatic phrasing change. Harmony mostly doesn't.

I got through my own plateau with a great teacher and a small piece of software I built on the side while I was a product engineering leader at Patreon. It was a spaced repetition app that emailed me a practice plan every day and resurfaced whatever I was about to forget. Within about a year I broke through. That app was the first version of Sonora.

The world invites a business out of you

Sonora was an accident. Friends asked to use the system, then their friends did, then I ran out of capacity and hired a teacher. The pandemic hit, we had amazing musicians on staff who needed consistent work, and we had to build real marketing and admissions functions. Seven and a half years later, nearly 6,000 students have gone through the program and seven of them have won Grammys.

Pioneer Species happened the same way. I didn't set out to start it. Enough friends at our New Year's retreat said "teach me what you're doing with AI" that I organized it, and now every cohort fills through word of mouth. A lot of people ask "what business should I start?" I think the better move is to notice when the world keeps asking you for something. Product-market fit is undeniable when it's there. It smacks you in the face, and everything gets easier.

Motivation is the real bottleneck in learning

There's a whole field of pedagogy around sequencing and digestible learning paths, and it matters. But the biggest thing that stops people from learning hard skills is the motivation wheel. Learning something difficult forces you to confront your limiting beliefs. Good educators embed small wins along the path so there's always a carrot pulling you to the next step. Mark calls these "little victories" and opens his meetings by asking everyone for one, which I love.

Community turns intentions into action

We spent a while on Un Poco de Planning, the New Year's retreat I help organize. This year around 40 creatives and entrepreneurs spent a week in Mexico City reviewing the past year, building ten-year visions, and getting tactical about the next year and quarter.

The magic is what happens when a room full of people stops letting you talk yourself out of an idea. My friend Connie arrived wanting to interview people, paint their own words on their bodies, and photograph them, with no idea how to do photography. Someone in the room had lights. Someone had a camera. Someone volunteered as a model. They shot the next week, and a month later she was showing the work at Zona Maco. That's the power of setting a clear intention around people who say "here's how you'd do it" instead of "are you sure?"

Mark and his wife Jenny left the retreat committed to building a company together, Happy Overall, which is a perfect example of the same thing.

Generalism creates agency

Mark has spent a 25-year career doing many things at once, and he asked about what I've written on generalism. Most domains overlap. Jack Conte, who I recently had on my podcast, is a great CEO and a great bandleader, and each role makes him better at the other. Charlie Munger's idea of learning the hundred core mental models across disciplines is the same insight.

The piece I emphasized is that generalism creates agency. Entrepreneurship is about how fast you can go from "I want to fix this" to a working solution. I grew up coding, so when I see a problem I immediately ask how software could solve it. Then building companies forces you to bolt on accounting, sales, team management, ads, and fulfillment. Those skills generalize to every kind of business, and it turns out almost everything in life is kind of a business.

Learning anything means following the constraint

Languages and instruments are both embodied skills with the same mechanic: memorize information, then make it immediate and intuitive. Knowing a scale is very different from expressing yourself with it, the same way knowing vocabulary is different from holding a conversation. Spaced repetition and good sequencing, where each new thing builds on the last, matter in both.

The deeper principle is the theory of constraints. In a business you find the biggest constraint, form a hypothesis, test it, and move to the next one. In music you find constraints by putting yourself in uncomfortable situations. I'm learning bluegrass right now. At a jam someone points at me to take a solo, I'm fine over the G chord, and then the C chord shows up and I have nothing. So I go home, learn ten licks in C, come back, and see whether they come out in the moment. Skills, languages, and companies all come down to aiming your attention at the biggest constraint and building small environments to fix it.

Managing agents is the essential skill

This was the center of the conversation. When Claude Opus 4.5 came out at the end of November 2025, agentic coding went from mostly broken to genuinely good. I spent December rewriting software for Sonora and building an app for the retreat, and over the following months it became clear that the way we work had fundamentally changed.

I now believe that learning to manage agents, for both personal and professional use, is the most essential skill for every working adult in the world. It doesn't matter if you're in customer success, fulfillment, marketing, engineering, or running a solo business. Every single thing you can do with a computer, agents can now do if you design the right environment for them. My own work looks completely different than it did six months ago. I start the day spinning up 15 to 30 agents and spend most of my time on QA and keeping the plates spinning.

At Sonora we rewrote every piece of software we used to pay for, from the CRM to scheduling to contracts, plus tools for processes that never had software before. That saved about a quarter million dollars a year. More importantly, once all the data lives in one place and you own every hook, agents can operate across the whole business. Claude knows every aspect of the company. It's like the most incredible board member or advisor, one who knows second to second what's working and what isn't.

It also makes individual contributors far more capable. Instead of a month-long campaign project across several people, I can ask for a marketing campaign and it writes the landing page and emails, picks the audience, and loads it into the delivery system. When something is wrong, you don't just fix the output. You fix the skill or context that produced the mistake so it never happens again. That's where the compounding comes from.

The middle manager layer is going away

Mark works in mergers and acquisitions, and he pointed out that most founders aren't ready when they go to sell. Assembling a data room takes months. We effectively have a live one running 24 hours a day: the Sonora OS, a single app holding everything from accounting to the CRM to AI reviews of every customer call, with Claude as the intelligence layer on top.

That changes planning. Quarterly planning can now look like asking Claude to look at the business and tell us what to work on. It suggests the priorities and then asks if it should just go do them. We keep a human approving the suggestions, but you could close the loop entirely.

My view, which is a hard pill for a lot of owners and private equity groups, is that the middle management layer is going to disappear. Conceptually, middle management is a data normalization function: pass executive intentions down, synthesize front-line reality back up, and keep people accountable. AI may be the best tool in the world at that. I think organizations flatten into executives and individual contributors, with senior player-coaches spending about 70% of their time doing the work and 30% teaching.

What's left for humans is what computers can't do, which is relational. We didn't touch our mentor team in the transformation. Mentors are the sauce. Guiding someone through a deep mastery path is a human transmission, and humans aren't leaving that loop. What AI does is review every lesson across the whole program and ping me on Slack the moment something falls outside the process we know works. That becomes a training moment for the next team meeting. No manager could ever review that volume.

The Time article, and what I got wrong

Mark asked about the Time piece, which started with meeting a reporter at a party. His story was about displacement at small businesses, and we became the main example, since some roles existed only to maintain SaaS tools and agents took over much of our outbound setter work. I followed up with a longer post because I think the more accurate narrative is upskilling. The team is still growing, and I want everyone on it to be excellent at directing agents.

On mistakes: any change in a company carries a tax, and in a few places we did a hard switch and paid it. The real mistake was the timing of communication. My default is to go in like a bull in a china shop, learn it, execute, and tell the team afterward. During an AI transformation people are absorbing a steady media diet of fear, and their CEO is suddenly obsessed with AI. Of course they wonder if they're next.

What I'm learning is to keep the forward momentum while communicating incrementally, so the team participates in the transformation and feels ownership of the outcome instead of feeling nervous to adopt the tools. The upskilling itself is non-optional. Within a few years, not knowing how to work with agents will be like not knowing how to use a cell phone or answer email. The interfaces will get friendlier, but the way of thinking is the real shift: instead of doing the thing, you manage the environment and context that lets agents do it.

Detachment and curiosity in hard seasons

We've had plenty of eating-glass seasons at Sonora. In 2020, election-cycle ad costs spiked and we were hemorrhaging cash. I had about $100,000 in payroll and payables due the next month and $8,000 in the account. I sat down, half panicking, and wrote out every option, starting with crawling under my desk. Then a revenue-based financing loan, raising capital, a promo to current and prospective students, getting on the phones myself, affiliate deals with companies I already knew. Twelve options. I executed on all of them, three or four worked, and we got through.

Two things came out of seasons like that. First, detachment from the outcome. Write down the worst case. Nobody dies, we're teaching music, and you'd learn enough to make the next thing better. Nothing generative comes out of a cortisol-spiked, clenched state. Second, curiosity is the opposite of contraction. It's very hard to be fearful and genuinely curious at the same time. So ask why it isn't working, and what it would take to 10x from here.

Success as alignment

Ten years ago I'd have defined success differently. Now it feels like alignment plus freedom of choice. Am I spending my time on things that feel right, with people I'm aligned with? Do I want to wake up and do the work? Getting to have conversations like this one with like-minded people is part of it.

Music teaches presence

Ed Cook, who created the Sonic Sphere, sent in a question: if I could take a 30-channel microphone rig anywhere in the world to record a song tradition, where would I go? I talked about Voices of the World, the 45-minute spatial tour of vocal traditions I built for the Sphere at The Shed. My answer was Georgian folk choirs. The compositions are incredible, singers are encouraged to paint outside the lines, and they fold in screams, guttural sounds, claps, and stomps. It's probably the most creative vocal music I've heard.

Mark also asked what my time in Ireland taught me. Music, especially session and jam music, is about presence and listening, and only about 10% playing. It's active meditation: getting inside the other players' heads while staying in your own body enough to execute. Nobody in a session announces that they're going to be present. It's just the only way to play that music well.

The same is true of interviewing. On Playback I've sat down with players like Tommy Emmanuel, who has mastered improvisation, and the conversation works the same way a great jam does. Prepare, then surf the moment, because the next question becomes obvious the closer you are to the present.

The transfer rate is the asset

We closed on two ideas Mark had pulled from my writing: trust the compound interest, and the transfer rate is the asset. When bodies of knowledge overlap, pattern recognition gets strong. Jamming teaches interviewing. Running a company teaches running a band. Marketing skills transfer to every project.

It reminded me of the improv classes I took at Endgame in San Francisco. They teach you not to be clever and to say the obvious thing, because your obvious thing is someone else's genius. Your job is to build a wide foundation of different skills and knowledge so that what's obvious to you turns out to be novel to everyone else in the room.

Thanks to Mark for a generous, wide-ranging conversation. If the AI section resonated, Pioneer Species is where I teach all of it, and it's built for non-technical people in any field.