AI doesn't make your job easier.
Building software has never been cheaper. Doing product management well has never been harder. AI didn’t take the PM job — it exposed the one we should have been doing all along.
There’s a quiet panic running through product orgs right now. PMs see AI writing specs, generating tickets, drafting wireframes, and shipping working code, and the question hanging in the air is the obvious one:
Do I still have a job?
Here’s my answer, the one I’d give any product builder reading this:
AI didn’t take the PM job. It exposed the one we should have been doing all along.
The hard part of product management was never the paperwork. It was understanding what customers actually want, crafting it into something elegant, helping people get value from it, and turning that into value for the company. None of that got easier. You just get there sooner.
AI doesn’t make our job easier. It gets us to the hard part faster.
This piece is sponsored by Atlassian, who hosted me at their Team ‘26 conference a few weeks ago. Their announcements mirror what I’m seeing across product teams right now: the shift from isolated individuals working with single-player AI to teams collaborating with AI and agents. That new way of working needs new tools.
Product Collection is Atlassian’s answer: an end-to-end suite tying customer feedback, discovery, roadmaps, and delivery into one connected system for AI-native product teams.
For decades, the PM’s job was to say “no.”
Software was expensive to build. Engineering time was scarce. The frameworks we learned — RICE, MoSCoW, ICE — were elaborate machinery for saying no on behalf of a system that couldn’t afford yes.
Most team rituals — heavy prioritization, careful sequencing, conservative resourcing — were built for a world where build cost was the binding constraint. Most haven’t been re-examined since that world ended.
Even our most sacred best practice was a workaround for scarcity. The MVP isn’t actually a “best practice” — how many customers are looking for a product that just barely solves their need?
The frameworks we take for granted weren’t designed to help us build the best products. They were designed to help us avoid building the wrong one.
The MVP isn’t actually a “best practice” — how many customers are looking for a product that just barely solves their need?
The PM job, under those constraints, was choreography: moving an idea through a system designed to resist it. Does it move our OKRs? How long? What’s the ROI? The opportunity cost? Most of what we called product management was really an elaborate apparatus for justifying no.
One of the PMs I talked to at Team ‘26 captured what’s changing in a single sentence: “We’re entering an era of customer obsession. We used to have to say no to a lot of things, but now we can say yes.”
Now, “yes” is cheap. That makes the job harder, not easier.
A spec that used to take a week now takes a coffee. A prototype that used to take a sprint now takes an afternoon. Features that took quarters can ship in days.
This was supposed to make our job easier. It didn’t.
Two clocks now run inside every product team:
AI speed governs specs, code, prototypes, and tickets.
Human speed governs customer conversations, onboarding, stakeholder alignment, convincing your CEO, culture. As AI speed accelerates, the bottleneck has moved entirely into the human-speed bucket.
The leaders failing this transition are obsessed with what AI can do and are not thinking enough about the unique value humans bring to the table: raw speed without judgment is its own problem. Look closely at any “high-velocity” team right now and you’ll find:
A graveyard of PRDs nobody read
Prototypes that never got demoed
Features that shipped to crickets
Roadmaps that get re-litigated every six weeks because nothing connects them to what customers actually wanted
That’s motion, not progress.
That leads to products that get more complex but not necessarily more valuable for users who are overwhelmed with new features and constant changes.
And here’s an unexpected twist: complicated software is bad for humans AND bad for agents. The cost of failing to curate is now doubled.
Customers have always had a fixed attention budget — shipping more features they don’t understand has always been a drag on the product. But now there are agents on the other side of your product too, and every undifferentiated feature you ship makes their job harder. More tokens. More wrong turns. More chances for the wrong action.
The bottleneck has moved. It used to be can we build this? Now it’s should we, and how will we know it worked?
The PM job isn’t dead. It changed into something unrecognizable.
The old job was prioritizing: ranking the backlog, saying no to most things, because most things couldn’t be afforded.
The new job is curating: deciding what earns a place in your product, in your customer’s attention, and in your company’s strategy.
Curation is fundamentally different than prioritization.
Curation demands judgment, customer obsession, and conviction. It also demands evidence — because the worst version of the AI era is one where the highest-paid person in the room keeps winning every fight by force of personality. That’s the failure mode most teams are walking into without realizing it.
At Atlassian Team ‘26, I met Tanguy Crusson — Atlassian veteran and founding product leader for Jira Product Discovery). He described the breakthrough for teams running Atlassian’s new AI-powered discovery process this way:
“Most of what product teams were basing their decisions on was disintermediated information — one line from sales, a CSM summary. We’re trying to turn that around. We go straight to the source. Your CEO comes up and says, ‘I’ve got a brilliant idea, it’s a new shiny object,’ and often teams end up scrambling for weeks to respond. The breakthrough comes from the moment where they can put that idea in and compare it more objectively against everything else on the roadmap.”
That’s the curator’s job in a sentence. Triangulate across analytics, customer feedback, and intuition — and make a call based on informed conviction, not gut instinct.
A counterintuitive consequence: in the AI era, qualitative work matters more, not less. Customer attention is the binding constraint, and you can’t A/B test your way to understanding it. That means fewer experiments where we treat customers like bacteria in a Petri dish and more conversations where we actually listen to them.
Customer attention is the binding constraint,
and you can’t A/B test your way to understanding it.
Curation needs a system.
If curation is the new job, where does it actually happen?
For most teams, the answer is: nowhere coherent. Feedback lives in one tool. Prioritization in another. Delivery in a third. The PM ends up as the human API between systems that don’t talk to each other. That’s where context splinters, making it harder to make good decisions.
The deeper problem is that almost every AI tool we’ve adopted so far has been single-player. Each PM builds up their own context inside ChatGPT or Claude. Their own prompts. Their own custom GPTs. Brilliant for individual productivity. Useless to the team. We ended up with a generation of 10x individuals trapped inside teams that didn’t get any faster.
That’s the shift I think product leaders are underestimating: we’re moving from single-player AI to multi-player AI. From individual context to shared context. From 10x individuals to 10x teams.
Mike Cannon-Brookes, CEO of Atlassian, discussed this shift at Team ‘26:
“Raw intelligence is now a commodity. You can buy smarts by the token. Models cannot be your differentiator. The differentiator is your context. If your AI doesn’t know what choice you made in 2024, it can’t help you win in 2026.”
That’s the right frame. In the AI Strategy course I built with Reforge, we teach the same underlying principle. The moat isn’t the model. It’s the proprietary context — and increasingly, the shared, team-level context — wrapped around it.
What stood out to me at Team ‘26 isn’t really about any one company’s product stack. It’s that almost every product team is walking into the curation era carrying the same three gaps — and the teams that close them are going to compound while the teams that don’t get buried in their own output.
The signal gap. Customer reality lives in sales calls, support tickets, Slack threads, app reviews, and analytics dashboards — and almost none of it reaches the people deciding what to build. Voice of the customer stops being a one-line summary from sales or a 200-slide deck prepared twice a year. As Sachin Rekhi puts it, it becomes a continuous feedback river the team can constantly pull from.
The evidence gap. When the highest-paid person in the room drops a shiny object, most teams have no way to compare it objectively against everything else on the roadmap — and what customers really want. Conviction requires receipts. Without them, the loudest voice keeps winning.
The continuity gap. The “why” rarely survives the trip from discovery to delivery. It gets stripped out at every handoff, like a game of telephone with the customer waiting anxiously on the other end. Teams that close this gap treat shared context as a shared artifact, not an ephemeral prompt.
Atlassian’s Product Collection is the most credible attempt I’ve seen at closing those gaps inside one connected system. But the gaps are the point, not the product. They exist whether you’re on Atlassian, Linear, Productboard, or a stack of spreadsheets — and closing them is the new job.
The bar moved. Set yours accordingly.
The idea I keep coming back to from Team ‘26 is:
“A feature is not done when it ships. It’s done when customers get value from it.”
That’s the bar in the AI era. Not “did you think it and ship it.” But “did the customer get value from it — and did that value accrue to your company?”
The teams that win this next decade won’t be the ones with the smartest model. They’ll be the ones that operate on shared, durable, accessible context — and use it to curate ruthlessly.
The PM job isn’t dying, but it is changing. We can outsource the parts we used to dread — the ticket-writing, the spec-formatting, the apparatus of saying no. What’s left is judgment, customer obsession, and the hard, human work of figuring out what should actually exist in the world.
The job we should have been doing all along.
Sponsored by Atlassian. All opinions, framing, and takes are my own.