Your Team Has AI Tools. Your Workflow Is Still Broken.
Having Notion AI and a meeting recorder doesn't mean you have an AI-enabled PM workflow — it means you've bolted new tools onto the same broken process.
Your team probably has Notion AI. A meeting recorder of some kind. Maybe a feedback synthesis tool someone found at a conference last year. And yet, if I asked you to walk me through how a user insight moves from a customer call to a prioritization decision, I'd bet the answer still involves someone spending half a day reading notes, writing a summary in Slack, and hoping the right stakeholders see it.
That's not an AI PM workflow. That's a 2023 workflow with a few subscriptions added on top.
The teams I've worked with that are genuinely faster aren't the ones with the most tools. They're the ones that stopped asking "what tool should we use?" and started asking "which parts of our workflow are just humans doing low-judgment work that a machine could do?" Those are different questions. Most teams are still asking the first one.
"The biggest issue I see in product is not that teams lack the right tools. It's that they lack the right system."
— Melissa Perri, Escaping the Build Trap
Five layers, and most teams have touched one
There are five distinct layers where AI can actually change how a PM team operates: meeting intelligence, feedback synthesis, prioritization input, stakeholder communication, and execution tracking. Each is a different kind of work. Each has a different risk profile for automation.
My experience is that most teams have automated one, maybe two of these, and they did it ad hoc. Someone added a meeting recorder. Someone else set up a Slack integration that summarizes tickets. Nobody sat down and mapped the full workflow before buying anything. So now you have partial automation with no shared system of record, and the humans are still bridging the gaps manually, just in different places.
Melissa Perri put it clearly in *Escaping the Build Trap*: "The biggest issue I see in product is not that teams lack the right tools. It's that they lack the right system." She was writing about process, not AI, but the point lands exactly here. A tool is not a system. Five tools with no operating model is chaos with better UI.
Buying tools and redesigning workflows are different projects
This is the uncomfortable part. Buying a tool takes an afternoon. Redesigning a workflow takes real work, usually requires someone with the authority to enforce a new operating model, and means some people will have to change how they spend their time.
Most teams don't have that conversation. They approve a subscription and assume adoption will follow. It doesn't. What follows is six months of the tool being used by two enthusiastic PMs while everyone else keeps their old habits, and then a Slack message asking whether anyone is actually using it.
The 28% salary premium research showing up in PM hiring data for AI-skilled candidates isn't measuring tool familiarity. It's measuring workflow redesign fluency. Knowing that Fathom exists is not a skill. Knowing that you should automate meeting intelligence before you automate prioritization inputs, because one is low-risk and the other requires trust in data quality you don't have yet, that's a skill. Sequencing matters.
The GDPR constraint nobody in the generic roundups mentions
If you're running a product team in Europe, there's a layer of this conversation that US-focused AI PM content skips entirely. A significant share of the AI tools being recommended in PM communities are hosted on US infrastructure with no EU data residency option. That matters the moment you're feeding them customer feedback, internal documentation, or anything that touches personal data.
I've seen teams in Stockholm and Amsterdam spend weeks integrating a feedback synthesis tool, then have legal flag it during a routine audit and pull the whole thing. The workflow work was real. The outcome was zero.
Before you redesign around any tool that touches customer data, check the data residency story. This rules out more tools than people expect, and it means the shortlist for a European product team looks different than what you'll find in a general AI tools roundup.
The audit your team actually needs
Take one PM's current week and map it layer by layer. Meeting intelligence, feedback synthesis, prioritization input, stakeholder communication, execution tracking. For each layer, ask one question: where is a human spending time on something that requires no real judgment?
That list is your backlog. Sequence it starting from the highest-friction, lowest-risk layer first. In most teams I've seen, that's meeting intelligence. It's high volume, low stakes if the AI gets something slightly wrong, and the time savings are immediate and visible.
Start there. Get one layer working properly, with a shared operating model the whole team uses, before you touch the next one.
The goal isn't more tools. The goal is fewer humans bridging gaps that shouldn't exist. Those are different projects. It's time to run the second one.
Fredrik Göth is a CPO and product leadership consultant working with product teams across Europe.
References
- Melissa Perri — Escaping the Build Trap (2018)
Ready to try it yourself?
Sign up free and start connecting strategy to impact today.
Related reading
- Your Analytics Tool Is Not Your Source of TruthMost product teams have good analytics but no real system of record for why decisions get made. This article examines the gap between insight and roadmap and what to do about it.
- Why AI Roadmap Features Don't Actually Improve PrioritisationAI roadmap prioritization tools generate ranked lists quickly, but without strategic context, constraints, or trade-off reasoning, they're doing autocomplete — not real prioritisation. Here's what actually matters.
- AI Product Discovery Still Fails Without an Operating ModelAI tools for product discovery are multiplying, but most teams still face a synthesis-to-decision gap. Learn what an operating model for AI product discovery actually requires.