What AI Can't Replace: Intuition, Alignment, and Taste
“But what about AI??” In the age of AI-assisted output, clarity and taste are what will set early-stage teams apart.
The past year has changed how we build.
AI tools aren’t just helping with small tasks around the edges anymore. They’re making it possible to design, prototype, and ship faster than most of us imagined, especially for lean teams.
This post was sparked by a recent fireside chat I did with my friend Vince Law’s product management class at Haas. We talked about how AI is changing the way product is built, and what today’s PMs and early-stage teams should be paying attention to.
We discussed Julie Zhuo’s post, The Death of Product Development. She describes how we’ve moved past the era where building was the bottleneck. Now we’re in the era of product decision-making. Exploration is easy. The hard part is pruning: deciding what matters, what doesn’t, and where to focus.
One metaphor that came up in our conversation stuck with me: a rowing team. With AI, it’s like every rower suddenly gets access to a turbo paddle. Everyone can move faster and harder. But without coordination and clear direction, the boat doesn’t go anywhere.
This post is about leveraging the skills that still matter, the ones that keep your team rowing in the same direction.
What AI Is Actually Good At
First, let’s start with where AI tools are genuinely helpful:
Prototyping
Need a UI flow, some design ideas, or some realistic-seeming demo data? AI-assisted prototypes are quick and cheap (hence the prototype-and-prune approach we’re headed towards).
Refactoring
Cleaning up dead code, renaming for clarity, converting projects between stacks — AI handles the tedious parts with surprising competence.
Drafting
AI tools are great at getting you unstuck. Whether it’s a rough spec, a release note, a blog post, or an onboarding email, it gets you over the blank page problem, and gets something on the page you can react to and iterate on.
Brainstorming
AI is tireless. It’s great at generating a wide range of ideas quickly, and really takes the “no bad ideas” part of brainstorming to heart.
If you’ve worked in an early-stage team where time is tight and resources are even tighter, this kind of leverage can feel like magic.
Where Things Fall Apart
The same tools that make it faster to build can also make it faster to drift from your intended direction.
The challenge is that AI output is only as good as the input you provide. If you have clarity around your users, market, and goals, AI can be a real amplifier. But if you don’t, you may end up with products with tons of features but no product-market fit.
AI tools will generate five polished and confident-sounding ideas even if you haven’t clarified the problem you’re trying to solve.
It doesn’t care if what you’re building is coherent, or if the new feature complements the rest of the product, or if it’s worth building at all.
That part’s still up to you.
The faster the tools become, the more tempting it is to just build. But if your team isn’t aligned, or if you’re chasing output without a strategy, you end up with a product that might look polished but doesn’t achieve its goals.
The Skills That Still Matter
There are a few things AI doesn’t do well (yet), and these are the things that tend to make or break early-stage teams.
Product Intuition
AI can give you a hundred ideas faster than you can gather your team for a sticky-note brainstorm session, but you still need to know which ones are worth exploring, and which ones to set aside. Use it to generate ideas, but have your team react to them.
Try this:
Gut-check with real stakes: Ask, “Would I bet a week of team time on this?” If not, don’t pretend it's top priority.
Play it out: Imagine the feature is already live. How would you explain it in a release note or to a user? If you don’t know how you would explain how it fits in with your product, there’s your answer.
Ask the inverse: “What happens if we don’t build this?” If the answer is “not much,” it probably isn’t worth exploring further.
Alignment
The more output your team can generate, the more costly misalignment becomes. Small gaps get amplified, and without alignment and aggressive pruning, you can easily imagine a team building out and then having to maintain a mess of disjointed features.
Try this:
Start sprints with clear alignment: “This week, we're trying to ____.” Everyone should be able to fill in the same blank.
Prune your backlog together: Take 15 minutes to prune what no longer supports your current direction, especially if it was AI-generated in a moment of “this could be cool.” This also helps hone your team’s shared judgement.
Ask what you're saying no to: Explicitly name one thing you’re not doing this cycle.
Taste
AI doesn’t have taste, at least not yet. If it did, we would all be spared the stream of clearly AI-generated Linkedin posts. It can follow general frameworks, but doesn’t know what feels off-brand, what undermines trust, or what makes a product feel like it’s actually “you.” Building out a team’s shared taste will help you not end up with a product filled with AI-generated-sounding copy.
Try this:
Build a “We don’t ship this” list: Common anti-patterns, awkward phrasing, or off-brand vibes.
Run it through this simple test: “Would I be proud if this were the first thing a new user saw?”
Compare two options out loud: Sometimes taste shows up in contrast. Have your team review two variations and talk about why one feels better, even if both are technically “fine.”
The Opportunity in All This
A core part of the future of product development, according to Julie Zhou’s post, is the “prototype-and-prune loop.” AI makes it easier than ever to explore across features, flows, pitches, designs. But the leverage only compounds if your team knows how to prune effectively.
A misaligned team with AI will move quickly, in all directions at once. More features, more surface area to maintain, more confusion.
But a focused team with alignment, good instincts, and taste?
They’ll move faster than ever — and actually get somewhere.
Final Thoughts
AI capabilities are evolving quickly, and will keep evolving. What feels cutting-edge today will likely be table stakes soon.
But no matter how fast the tools change, some fundamentals stay constant, at least for the foreseeable future: good judgment, alignment, and taste will continue to shape which teams build products that actually resonate.
That has implications for hiring for early-stage startups. Rather than screening for existing skill sets, think about hiring candidates who are curious, eager to learn, and discerning.
You can also actively develop these traits within your team. Taste and judgment aren’t fixed qualities — they’re shaped through shared language, repeated decisions, and reflective conversations over time.
As always, send us an email or leave a comment if there’s anything you’d like me or Jen to write about in the future!
If you haven’t already, it’s worth taking a look at Julie Zhou’s post that Vince Law and I discussed in the fireside chat:
Good points — but this still feels like we're framing AI as “going faster” and “pruning better.”
That’s just upgrading the scissors.
Speed isn’t the point.
Delivering crap 10x faster is still delivering crap — because of the end of the day it's still crap.
We’ve spent a decade hyping "10x PMs" and "10x outputs," and all it proved is that you can scale waste and dysfunction just as easily as you can scale impact.
Moreover, for the past 2 years, I've been showing how Generative AI — and soon agentic systems — aren’t just about turbo-charging output.
They’re about synthesizing, simulating, reframing, and validating before an MVP even exists. Think pretotype (No, that is not a typo).
TL;DR: It’s not about faster boats or stronger oars.
It’s about picking the right damn river —
and if needed, questioning if the river’s even the right way to move.
AI doesn’t just help ship faster.
It helps pick better markets to ship to — or better problems to solve.