5 Human Skills That Matter More as AI Gets Better, Not Less
As AI makes execution cheap, taste, agency, focus, resourcefulness and follow-through become the scarce skills. Here's why they're more valuable now.
What human skills actually grow more valuable as AI improves?
Five skills gain more value as AI capability increases: taste, agency, focus, resourcefulness, and follow-through. The common thread is that AI has made execution cheap, which means the bottleneck shifts to the human choices around execution: what’s worth building, whether you actually start, where you concentrate effort, how you improvise when stuck, and whether you finish. None of these skills are new. What’s new is how much more they pay off now that the mechanical cost of doing the work has dropped.
TL;DR
- Taste means acting on something you care about, not just recognizing quality, and AI lets people close to a problem build a fix instead of writing a memo about it.
- Agency is measured by the gap between what you say you’ll do and what you actually do, and AI closes that gap by removing a lot of what used to stop people from starting.
- Focus has become more valuable precisely because AI makes it easy to start many projects at once, and sustained attention on one problem now compounds faster than it used to.
- Resourcefulness shows up as flexible tool use: people solving the same problem with completely different stacks (Slack, Notion, Zapier, Discord, Obsidian) because the specific tool matters less than the willingness to improvise.
- Follow-through is what turns a working prototype into something that actually ships and gets used, and it’s the skill most likely to separate people with identical AI access.
- The underlying argument is that skill requirements have changed, so projects shelved last year because they needed a hire, a budget, or a specialist may be worth a second look now.
Built like a system. Not vibe-coded.
Remy manages the project — every layer architected, not stitched together at the last second.
Why does AI increase the value of these skills instead of replacing them?
The standard narrative says AI erodes human skills one at a time, like a tide taking things off a beach. The counterargument here is that AI removes the execution barrier, not the need for judgment. When building something no longer requires months of someone else’s time or a technical hire you couldn’t afford, the constraint moves to you: your willingness to act, your standards for what’s worth doing, your ability to stay on one thing, your knack for improvising, and your discipline to finish.
This matters for planning. If you ruled out a project last year because it required a developer, a budget, or a dedicated team, that requirement may no longer hold. The skills that let someone carry a whole piece of work that used to need several people’s help are exactly the five above.
How does taste work differently now than it did before AI?
Taste is often defined as the ability to recognize what good work looks like. That’s incomplete. Taste is acting on something you care about. The distinction matters because recognizing a problem and doing something about it used to require different resources: recognizing a bad onboarding flow took attention, fixing it took a developer.
AI collapses that gap. Someone who notices the same customer confusion over and over, the kind of friction a support person or operations person has flagged for years without ever getting a fix, can now build a rough working alternative instead of writing it up and waiting for another team to prioritize it. That rough version does two things: it gives a colleague or customer something concrete to react to, and it often reveals details that weren’t obvious from the original complaint (the customer understands the explanation but doesn’t know which option applies to them, or the sequence of steps is wrong).
The collaborative part is underrated. Working through a build with AI can sharpen your own taste, not just execute it. The model might suggest an approach you hadn’t considered, or explain why a particular design choice works. Taste improves through the act of making, not just through observation.
What does “agency” mean in practical terms, and how do you build it?
Agency is the ability to act on what you could influence, measured informally by the distance between saying you’ll do something and actually doing it (the “say-do ratio”). AI doesn’t close that gap by making you more motivated. It closes the gap by removing a lot of what used to block the start: access to a specialist, weeks of setup work, a plan you didn’t know how to write.
That shifts responsibility. If a problem required six months of someone else’s time, and AI now lets you do much of that work yourself, the advantage sits with your willingness to act, not with external constraints. A useful pattern here: a rough first version of a tool or system, even an imperfect one, often surfaces what you actually want better than trying to specify every requirement in advance. A first attempt at flagging stalled sales deals might flag the wrong signal at first, but seeing that wrong signal is often what clarifies the real one. Building helps you find the plan, rather than the plan coming first.
The practical move for anyone with a stalled project: ask what’s actually blocking the start, not what blocked it last year.
Why does focus matter more when AI lets you try almost anything?
This is the most counterintuitive of the five. A common claim in AI circles is that cheap experimentation means you should try many things at once. The argument against that: trying many things is easy, but the real advantage comes from sustained, intense attention on one thing.
Two people with identical AI tools can produce very different outcomes. One spends a month starting fifteen projects and moving on each time the work gets awkward. The other picks one customer problem and stays with it long enough that understanding compounds and real users start giving feedback. AI can make distraction look productive, by end of day you might have three prototypes and an elaborate test of an idea you didn’t even have that morning, but none of it may be far enough along to matter to anyone.
The practical questions worth asking before adding a new project: what deserves attention, what would you keep working on after the initial excitement fades, and what are you willing to leave alone. As AI capacity grows, it’s easy to fill all of it with new starts and still leave the important work unfinished.
What does resourcefulness look like when everyone has access to the same AI tools?
Resourcefulness is the ability to find a way forward without a clear playbook, using whatever’s available. In practice, it shows up as flexible tool use. People solving the same kind of problem (say, combining a messaging tool, a notes app, an automation layer, and an AI model) will land on completely different combinations: one person builds it in Slack and Zapier, another in Discord, another inside Obsidian, another in whatever software their organization permits. The specific stack matters far less than the underlying willingness to improvise with constraints and get something working.
This is the skill that keeps a project alive when the “correct” tool isn’t available, a budget doesn’t exist, or a team member isn’t around to help.
Frequently Asked Questions
Are these five skills new, or have they always mattered?
They’ve always mattered. What’s changed is their payoff. AI lowers the cost of execution, so the human judgment calls around execution (what to build, whether to start, where to focus, how to improvise, whether to finish) now have a much bigger effect on outcomes than they did when execution itself was the bottleneck.
Does this mean technical skills don’t matter anymore?
No, but the specific technical requirement for a given project has often dropped. Problems that used to need a dedicated developer or months of specialist time can sometimes be tackled directly by the person who understands the problem best, using AI to handle a large share of the execution.
Why is focus listed as a skill that matters more, not less, with AI?
Remy doesn't write the code. It manages the agents who do.
Remy runs the project. The specialists do the work. You work with the PM, not the implementers.
Because AI makes it cheap to start many projects, which creates a temptation to spread effort thin. The people who concentrate sustained attention on one problem tend to get further, build deeper understanding, and accumulate real user feedback, while scattered effort across many starts often doesn’t survive contact with the boring, difficult middle part of any project.
What’s the difference between agency and follow-through?
Agency is about starting, closing the gap between intention and action. Follow-through is about finishing, carrying a project through the parts that get tedious or difficult after the initial build is working. Both matter because AI has made the initial version of almost anything easier to produce, which puts more weight on whether someone actually launches and sustains it.
How do I know if a shelved project is worth revisiting?
Ask what specifically blocked it: a hire you couldn’t afford, a technical skill you didn’t have, access to someone else’s time. Then ask whether AI changes that specific requirement today. If the blocker was really about execution capacity rather than judgment or demand, it’s worth a second look.