Five skills AI makes more valuable, not less

The usual picture of AI is a rising tide: every few months the water takes another of your skills off the beach. A more useful picture is the opposite. As AI gets more capable, a handful of human skills are gaining leverage, because each one now has far more practical power behind it.

That's worth acting on before planning next year. Look again at the projects you ruled out: the service you couldn't launch without hiring, the problem your team understood but never had time to fix, the idea that needed a technical person you couldn't afford. Those decisions were based on what the work required then, and a lot of that has changed. More of us can now carry a whole piece of work that used to need several people. Doing it well takes five skills: taste, agency, focus, resourcefulness and follow-through.

1. Taste: act on what you care about

Taste is usually described as recognising good work. It's also acting on it. It starts with attention: caring enough to notice distinctions other people skip. And it isn't the same as seniority; people early in their careers can have excellent taste in an area they've studied closely and passionately.

Say new customers keep getting confused at the same step, and you can almost predict their question. You used to write a memo and wait for another team. Now you can use AI to build a rough interactive guide or a better onboarding flow, and put it in front of someone. Trying it reveals details the complaint never did: maybe customers understand the explanation but not which option applies to them, or the steps are in the wrong order. The AI can also suggest approaches you hadn't considered and explain why a design you admire works, so your taste improves as you make.

For teams, that means letting the people closest to a problem show a better version rather than describe one. Their first attempt often uncovers what a meeting never would. For yourself: pick the explanation that always confuses people, or the report that never answers the question, and make an alternative.

2. Agency: revisit what you can take on

Agency is taking the initiative on what you can influence, and it's measured by the gap between saying and doing. AI helps with the parts that used to stop you, and supplies an almost unlimited number of plans. Making a real attempt is still on you, and it matters more now: if solving a problem used to need six months of someone else's time, it may not anymore. Your willingness to act has a far more direct effect on whether things get solved.

You also don't have to know the whole plan before you start.

A sales leader wants to know which deals need attention but can't describe the system in advance. An agent builds a rough first version that flags deals with no activity for 14 days. Looking at it, the leader realises that's not the signal: the real worry is a contract that's sat with legal for 10 days with no next step. The imperfect attempt drew out knowledge the leader had but couldn't explain. Building helps you find the plan, so raise your sights, and let each attempt shape the next decision.

3. Focus: concentrate as options multiply

AI makes it possible to try many things, but that doesn't mean trying many things gets the most out of it. When almost anything is cheap to attempt, the scarce resource is sustained attention: how long you can stay with something difficult or boring, and how determined you are to see it through. The tools are converging for everyone, so higher standards and refusing to settle for slop become a real difference.

Picture two people with the same tools. One spends a month starting fifteen projects and moving on whenever the work gets awkward. The other picks a single customer problem and stays with it. The second person's attention compounds: they understand the problem better, and people are already benefiting and giving feedback. Explore widely to find a problem, but once you've chosen, focus pays off disproportionately.

AI can also make distraction look very productive. By the end of a day you might have several prototypes, three plans and an elaborate test of an idea you weren't pursuing that morning, and none of it far enough along to matter to anyone. For leaders, choosing where to concentrate becomes a bigger part of the job, because teams will bring more credible proposals than ever. Before taking something on, ask what you'd keep working on after it stops being exciting, and what you're willing to leave alone.

4. Resourcefulness: get comfortable with unfamiliar parts

Resourcefulness is finding a way forward when you don't know how. People are naturally flexible tool users, and that's what lasts while the tools keep changing. Give a group the same idea, say a simple system to capture information and find it again later, and they'll build it with whatever fits: one in Notion, another in Obsidian, another in the software their company allows. What they share is an understanding of what the system is for, and that lets them swap the parts.

AI changes what it's like to be a beginner. Bring it the confusing error from a project you care about, ask what's happening and whether there's a simpler way, and the explanation has somewhere to land. Each finished project makes the next one possible. For leaders, a CV says less than it used to; watch how people approach unfamiliar problems: whether they ask a useful question, reframe it, and come back with another attempt of their own.

5. Follow-through: carry it until someone benefits

We've become very good at starting things. A first version appears quickly and can look polished. Then someone has to stay with it until it matters. The onboarding guide has to actually get a new customer through the confusing step: maybe it needs to appear sooner, maybe a handoff needs to change, maybe it still assumes something the customer doesn't know. The sales leader has to change how the pipeline is actually tracked, not just brainstorm a better one.

Where you could once offer an analysis or a recommendation, you can now deliver something that works and stay involved while people find what still needs changing. As impressive first versions become ordinary, the people who repeatedly carry them through earn a different kind of trust. Good follow-through often goes past the edges of a job description, like the engineer who knows shipping to production isn't the end. Before the next project, ask who it's for, what they'll be able to do afterwards, and what doing the whole job would take.

Five skills, one loop

Together they cover the distance from an idea to something useful: notice what's worth improving, make a real attempt, give it sustained attention, learn what you need along the way, and carry the result through to the person it's for. More of that distance is now open to anyone who understands a problem and is willing to stay with it, whatever their job title.

AI has expanded what we can do faster than most of us have changed what we're willing to attempt. So look at the project you've been sitting on, and give yourself the chance to find out whether you can do more than you thought.