Field report · Two workshops · One dataset
The middle of
the job is being
hollowed out.
Eighteen senior product leaders. Fifty-four live-survey answers. Four ERRC flipcharts across two perspectives. Read as one: AI is not taking the PM's job. It is taking the middle of it.
Executive summary
Six things the room said
01
The middle of the PM job is being hollowed out.
Tickets, docs, summaries, analysis — all going to AI. The edges grow.
02
Given 20% more time, almost nobody asked for more features.
Discovery. Strategy. Customer contact. Learning.
03
The biggest skill gap is judgment — not prompting.
Knowing when to trust the output, and when to throw it away.
04
The Lead role is changing more than the PM role.
Twice the change surface. Operating-model redesign, not craft.
05
Three tensions the room did not resolve.
Prototype. "Everyone is a PM." Speed vs depth.
06
Fear is on the paper. Openly.
Senior leaders naming the emotional layer in writing.
01
Finding 01 · The shape of change
The middle of the PM job is being hollowed out.
The top clusters on Question 1 — tickets, documentation, summaries, release notes, analysis, first prototypes — all sit in the same structural place. They are the middle work: translating already-understood intent into the artefacts that unblock engineering and stakeholders.
The core is preserved — actually amplified. The connective tissue is going to AI.
Anatomy of the PM job · top of the survey · edges kept · middle released
↑ Upper edge
Customer discovery · problem framing · strategic bets
Kept · amplified
↑ Upper edge
Judgment on whether the output is right
Kept · amplified
◦ Middle
Ticket & story writing
→ to AI
◦ Middle
Documentation & release notes
→ to AI
◦ Middle
Meeting summaries & status comms
→ to AI
◦ Middle
Data & competitive analysis
→ to AI
◦ Middle
First prototypes · roadmap framing
→ to AI
↓ Lower edge
Team coaching · emotional carrying
Kept · amplified
↓ Lower edge
Stakeholder trust · hard-call ownership
Kept · amplified
02
Finding 02 · Where the 20% goes
Given free time back, almost nobody asked for more features.
Eighteen senior leaders were offered a hypothetical twenty percent more time. Only two asked for it to ship more. The inventory of what they would buy back is the inventory of what the current operating model has squeezed out.
If you had 20% more time, where would it go?
Question 2 · mentions across 18 respondents · single question, multiple mentions allowed
The bottleneck this room names is not throughput. It is clarity about what to throw throughput at.
03
Finding 03 · The skill gap
The biggest capability gap is judgment — not prompting.
10
/ 18 participants named some form of it
The single largest cluster in the entire dataset is AI judgment. Not tool-handling. Not prompt libraries. Knowing when to trust the output — and when to throw it away.
What's inside the cluster
›
Understanding the current limits of the tools.
›
Writing context and vision as guardrails, not prompts.
›
Distinguishing where AI helps from where it erodes capability.
›
Judging fast whether a vibe-coded feature is enough — or a full one is needed.
›
Prompting, yes — but only as one of many.
The quiet implication
Tooling is a one-time cost. Judgment under time pressure is a career-long practice. The gap the room names is the harder one.
04
Finding 04 · The change surface
The Product Lead role is changing more than the PM role.
Count the items on each ERRC flipchart and the asymmetry is striking. Nearly twice the change surface on the Lead side. And the content diverges: PMs talk craft. Leads talk operating model.
Flipchart A
PM in the team
~15
items across 4 fields
Agenda DNA
Craft.
Discovery. Critical review of AI output. Deep expertise. Storytelling. How much change users can absorb.
Discovery. Critical review of AI output. Deep expertise. Storytelling. How much change users can absorb.
Change surface
2×
→
Flipchart B
Product Lead
30+
items across 4 fields
Agenda DNA
Operating model.
Team-setup redesign. New collab with design & UX. Experimental leaders. Automation boundaries. Carrying teams.
Team-setup redesign. New collab with design & UX. Experimental leaders. Automation boundaries. Carrying teams.
The Lead role in the AI era spends less time requesting status and more time redrawing the system the PMs work in.
The discourse about AI and product management tends to focus on the PM. The room quietly said the Lead has the harder job.
05
Finding 05 · Unresolved
Three tensions the room did not resolve.
These are not confusions. They are the unfinished work of the field.
01
Prototyping: Reduce — or Raise?
PM flipchart · reduce
Prototyping as a production cost is collapsing — less effort per artefact.
vs
PM flipchart · raise
Judgment about what to build — and discard — matters more than ever.
Reading the full dataset: both. One group drew an arrow between the two fields — a visible disagreement they did not smooth over. A single word now covers two different activities.
02
"Everyone is a PM now" — noise, or signal?
Lead flipchart · reduce
The confusion this phrase creates should be removed as noise.
vs
PM flipchart · raise
Evaluating ideas that can now come from anyone becomes more important.
Same phenomenon, two postures. Leads want less noise. PMs want to work with richer input. Both are probably right. Neither has worked out yet how they fit together.
03
Speed versus depth.
Across the room
Prototype faster · ship experiments faster · iterate faster.
vs
Same room
More time for deep discovery · reflection · strategic thinking · deep domain.
Not contradictory wishes — but not yet reconciled in any operating model anyone in the room described. The closest answer: the Lead's Raise column — "create time for learning and cross-functional competencies." Create is doing a lot of work in that sentence.
06
Finding 06 · On the paper
The emotional layer is named. Openly.
Rare in a senior leadership room. Across both perspectives, items explicitly about emotion showed up — on the page, not just in the corridor.
Emotion · as written on the flipcharts and survey
A community that knows the pace is unsustainable — and is quietly asking whether AI will give them room to breathe, or finish the work of burning them out.
Lead ERRC · Eliminate
"Fear of roles — both one's own and others'."
Lead ERRC · Reduce
"The teams' fear of AI, and the hurdles to using it."
Q3 · cluster 6
"Let go of traditional best practices."
Q3 · cluster 6
"Disrupt yourself before you are disrupted."
Q2 · outlier
"Work ten percent less."
Q3 · outlier
"Take breaks."
That senior leaders named this — on paper, in a workshop — is itself a finding.
What isn't in the data
The silences are as loud as the mentions.
Absent · 01
Nobody asked for better roadmap tools.
The long-running debates of the last decade of product management are simply not present.
Absent · 02
Nobody asked for better OKR frameworks.
Twenty percent more time does not go to "ship more of the roadmap" — it goes to understanding whether the roadmap is the right one.
Absent · 03
Nobody asked for better Jira templates.
The vocabulary of ticket-management is missing entirely.
Absent · 04
Nobody named "stakeholder management" as a missing skill.
Alignment appeared as an activity in Q2, but not as a capability gap in Q3. The gaps are sense-making and judgment, not influence mechanics.
Absent · 05
Nobody celebrated AI as a feature-multiplication machine.
When AI-as-productivity-gain appears, it's framed as "time given back" — and that time goes to deeper work, not faster work.
Absent · 06
The vocabulary of "velocity" is almost absent.
The room has collectively moved past the productivity argument.
A composite picture
The 2027 product leader, as this room draws them.
01
Frames problems more carefully.
Because AI can solve any well-framed problem cheaply.
02
Judges AI output under time pressure.
The cost of producing bad features is trivial. Shipping them is not.
03
Carries their team through the structural rewiring.
Change management, not status reporting.
04
More opinionated. More in contact with customers.
More willing to throw prototypes away.
05
More honest about what they don't yet know.
Not waiting for permission. Not pretending there's a playbook.
06
Quite plainly, tired.
Aware the current operating model will not survive if AI is just added on top.
Synthesis · The metaphor that makes all six findings one
From Product Manager to
Product Curator.
When execution is expensive, the scarce skill is delivery. When execution is cheap, the scarce skill is selection. That is the quiet thread running through everything the room wrote down.
When the cost of producing a feature collapses, the scarce role is no longer the one who builds. It is the one who chooses — which problems, which outputs, which experiments — and has the taste to stand behind the selection.
How the metaphor reads each finding
Finding 01 · hollowed middle
The middle is the printing press. The curator was never the printer.
Finding 02 · where the 20% goes
Discovery and customer contact are how a curator builds the eye.
Finding 03 · judgment gap
The curator's whole skill is judgment — not making, not prompting.
Finding 04 · change surface
The Lead curates curators — the conditions under which taste can develop.
Finding 05 · tensions
Speed vs depth, noise vs signal — the daily arithmetic of any curator.
Finding 06 · the emotional layer
Curation is personal. It takes nerve to leave things out.
The eulogy in this room was not for the product manager. It was for the manager-as-allocator-of-scarce-build-capacity. What emerges in its place is a curator of attention, judgment, and narrative — in a world where anyone can build, but not everyone can choose well.
The shift · Manager → Curator
Where this leaves the three questions the workshop opened with
What the room actually answered.
Question 01 · Role
Where does PM end and Engineering begin?
The PM role becomes more strategic, more judgment-driven, more human. The Lead role becomes more about operating-model design than product oversight. The eulogy for "PM as spec writer" is already being read in this room.
Converged
Question 02 · Team
How do you restructure teams?
Less settled. The Lead flipcharts point toward new team setups, new collaboration with design & UX, deliberate scarcity, automation models. No one claimed to have the answer. This is the open front.
Open
Question 03 · Skills
What new skills do leaders and teams need?
The room converged on two: AI judgment (not tooling) and strategic outcome-thinking. Both are hard to teach. Both are harder to hire for. Both will likely separate the leaders who adapt from those who do not.
Converged