The Product Manager Isn't Disappearing. The Job Is Changing

Every few months, someone declares the death of the product manager.

This time, the argument is more interesting. AI can write code, turn rough ideas into prototypes, analyze customer feedback, and increasingly take on work that used to bounce between product, design, and engineering. Gartner now predicts that by 2029, 60% of organizations will adopt smaller software engineering teams at scale.

If teams really shrink to three, four, or five people, something has to give. And the product manager looks like an obvious candidate.

So rather than speculate about which roles AI will eliminate, I wanted to check what companies are actually doing. I looked at 86,361 job postings from roughly 1,600 companies, compared product manager and product engineer hiring, and went through Gartner's tiny-team model and the examples behind it.

I expected to find at least the beginning of a shift away from the traditional PM role. I didn't. What I found instead was a change happening inside the role - and around it.

What Gartner is actually predicting

The prediction comes from a July 2026 Gartner press release: by 2029, 60% of organizations will adopt smaller software engineering teams at scale, up from 15% in 2026. Gartner analysts also ran a webinar on the concept they call "tiny teams," and I went through both.

The headline makes tiny teams sound like a headcount story. Gartner's own explanation is more nuanced: Aliyah Camacho, the Principal Analyst quoted in the release, says tiny teams "are not a cost optimization tactic" and that AI is "fueling the demand for more software engineers, not fewer."

In the U.S., the Bureau of Labor Statistics points the same way: it projects software developer employment to grow 15.8% between 2024 and 2034, with AI adoption fueling job growth in computer occupations. That is consistent with more, smaller teams, though it does not test the tiny-team mechanism itself.

The idea is more teams, each smaller: typically 4-5 people, sometimes 2-3, owning a product or feature end to end, with AI agents doing a growing share of the routine work. The release also carries a warning: Gartner predicts that organizations relying on AI to cut junior software engineering roles will hollow out their own talent pipeline by 2028.

There is also a dependency hidden behind the tiny-team idea: platform engineering. A three-person product team only works if someone else has already made deployment, security, infrastructure, and observability easy to consume - asked directly whether tiny teams can exist without skilled platform teams, the analysts' answer amounted to: build the platform team first.

The adoption math is also modest today:

• In a directional conference poll, about half of attendees reported team sizes of eight or more, and 74% expected AI to shrink their teams.
• The forecast starts from 15% adoption in 2026.
• In the e-book's June 2025 survey, 16% or fewer of software engineering leaders considered their workforce, delivery processes, or architecture ready for AI.

For now, tiny teams look more like a direction of travel than the default way software gets built.

Role compression, in practice

The most concrete part of the webinar was a case study from CBRE, a large commercial real estate company. As told by the analysts, AI made individual tasks faster while overall lead time did not move - work sat idle at role boundaries - so CBRE redesigned roles around the flow of work rather than task ownership.

For readers outside the field, the baseline division of labor: a product manager decides what should be built and why; engineers decide how. That sets up a more useful question than "Will AI replace product managers?" - a company may still need product management while needing fewer people with product manager in their title. If engineers take on more discovery, specification, and product decisions, one PM might support several teams instead of one. That distinction, between product management as work and product manager as a job, is what I wanted to test against the hiring data.

The resulting role map:

• Product owner and software engineer converged into a product engineer, who turns intent into executable specifications and stays responsible for validation.
• Product management and solution architecture merged into a product architect.
• QA became a quality and context engineer, shaping inputs to AI agents and reviewing their outputs.
• The designer moved upstream into research, defining user context and validation criteria.
• The scrum master role became lighter, shared across teams.

The pilot teams were 3-4 versatile engineers - one 3-person team reportedly shipped a self-contained application from concept to production in two sprints - and the convergence was conditional: product owners became product engineers "only when AI tooling was mature enough." That is what role compression concretely means: product judgment moving into engineering seats, and the dedicated coordination roles thinning out.

CBRE is the success story, told by the analyst firm presenting the model; Meta is the other end of the distribution. In August 2026 Reuters analyzed "Project OT," Meta's plan to make its workforce "AI native": an internal playbook mapped the traditional product team of 10-20 people - seven to fourteen engineers, one product manager, a designer, a data scientist, a user researcher, a data engineer - onto pods of 3-5: three or four "builders," one "direction lead" from any function, and specialists pooled across pods. Scenario planning explored shrinking many teams by as much as 60%.

The playbook shows where compression starts on paper: the PM seat is the first thing that becomes optional - direction is a hat anyone can wear.

The most aggressive version did not survive the year. As Reuters tells it, employee sentiment fell from 74% to 55% favorable, and internal data undercut the premise: code changes to internal platforms were up 220% year over year - a surge Reuters attributes to employees' AI use - but changes reaching users only 36%, while major technical and security incidents spiked 40% and firefighting time 70%. Hours before the first wave of layoffs in May, Zuckerberg called off the second wave planned for November, and in July he conceded that AI agent technology had not accelerated as quickly as he had anticipated. The plan assumed a capability that, by its author's own admission, was not there yet.

Still on the roster

Here is the part that surprised me, given how often I read that AI will eliminate the product manager: across Gartner's press release and webinar, the illustrative roster still includes one.

• The release says tiny teams require roles "such as" a product manager, a UX/agent experience designer, and at least one AI-native software engineer.
• The webinar names a product engineer, a product manager focused on vision and roadmap, and a product designer.

The rosters differ slightly, but the PM is on both.

The stated reason for the product engineer role is more interesting. The analysts argue that the traditional ratio of one product manager to eight or ten engineers "isn't good enough" - with spec-driven development, you need more product thinking mixed into the team, not less. So the interesting part may not be engineers replacing PMs. It may be engineers being expected to do more of the work we currently classify as product management.

McKinsey sketches the same merge - the product manager and developer roles "could eventually merge into a product developer" - while noting it may make as much sense "to maintain or enlarge teams" as to shrink them. Both point toward role convergence, but neither makes a strong case that teams will necessarily become smaller.

The CBRE case sits oddly next to that roster, though: there, product management merged into a product architect, product ownership moved to product engineers, and no dedicated PM seat was in sight.

And neither artifact resolves the staffing question that follows: if every tiny team keeps a dedicated PM, demand for PMs should grow with the number of teams; if product engineers absorb the day-to-day product work while one PM spans several teams, it may not. To me this is the most important open question in the whole material.

What the job-postings data shows

Gartner can tell us what organizations may look like in 2029. Job postings give a much narrower signal, but one we can observe now: are companies already changing who they hire? I ran an analysis on job postings collected between late June and early September 2026 - 86,361 postings from 488 sources, covering roughly 1,600 companies.

The source universe is deliberately skewed toward product roles, so these counts are not market totals. Sources were also added in waves during that period. I therefore use the full dataset to describe scale, and a fixed panel of company job boards for month-to-month comparisons.

For the classic product manager role (individual-contributor titles, spam excluded), across the only two fully comparable months I found no meaningful shift - and that absence of movement is the finding:

• Raw counts, mainly for scale: 792 PM postings in July from 263 companies, 829 in August from 274.
• In the fixed panel, I found no detectable change in PM's share of intake from July to August: 4.24% to 4.42% (absolute counts fell 567 to 510 alongside a broader hiring dip).
• Seniority is top-heavy: senior titles are 41% of PM postings, and junior, associate, and intern together are 2.5% (a further 34% of titles carry no level marker at all; explicit mid-level "II" titles are 1.2%). In this dataset, product manager hiring is heavily skewed toward senior roles.

For the "product engineer" title - the poster child of role compression, used by Linear, Intercom, OpenAI, Replit, and others - the role is real but small: 144 postings from 61 companies in the whole dataset. In August alone, PM postings outnumbered product engineer postings roughly 26 to one on the same sources (829 vs 32).

One important caveat: this is a title-level proxy. My classifier does not inspect responsibilities, and "product engineer" may describe a conventional software engineer working close to product rather than Gartner-style role compression. Many of these titles also predate the tiny-teams conversation - Linear and Intercom have used "product engineer" for years - so the title often reflects a company's naming convention, not the change Gartner describes. Job-posting titles can show whether companies are renaming roles; they cannot show what is changing inside existing teams.

The +33% that wasn't

Raw counts show product engineer postings up 33% month over month - 24 in July, 32 in August, distinct companies up 29% - against a mature PM benchmark growing 4.7% in the same raw data. The increase does not survive a sample control. In a fixed panel of 358 reliable company boards, with each posting counted on the date it first appeared in my data and each board's pre-existing backlog excluded, product engineer postings went from 17 in July to 16 in August. Flat.

The eight-posting raw increase decomposes cleanly: four postings came from boards outside the fixed panel (added mid-period or too unreliable to qualify), and the rest is a posted-date artifact - within the panel, employer-reported dates suggested growth from 17 to 21 while first-seen counts fell from 17 to 16. Reposts and refreshed posting dates created the appearance of growth.

This changed my interpretation of the data. The raw numbers initially looked like evidence that the product engineer role was taking off. Once I controlled for which company boards were actually comparable between months, that signal disappeared.

For now, I can show that the role exists. I cannot show that companies are hiring it more often. The qualitative signals still point in Gartner's direction - the title already appears at senior, staff, and principal levels, and the count of distinct companies posting it ticked up slightly even in the fixed panel. But with a dozen-odd postings per month, my panel does not yet show a clear signal either way.

What the descriptions say

The descriptions themselves give a partial answer to what these postings emphasize. I ran a phrase scan across all 144 product engineer postings and, as a baseline, across the 4,036 product manager postings. Caveats: phrase matching misses synonyms and boilerplate inflates themes; weighting companies equally instead of postings keeps the direction of every gap but moves some numbers (product-and-design collaboration 74% to 56%, AI 92% to 84%); and the corpora differ in company mix, sector, and region, so this describes the language of the postings, not a transfer of duties, and the AI shares are upper bounds.

Four themes stand out (product engineer share vs product manager share):

• AI-related language: 92% vs 52% (within the product engineer postings: AI agents or "agentic" 52%, LLMs 34%; 38% name a specific coding tool, Claude Code alone 25%)
• End-to-end or ownership language: 88% vs 67% (literal "end to end" or "from idea to production": 73% vs 43%, some of it boilerplate)
• Product sense or a product mindset: 46% vs 16%, and working closely with product and design: 74% vs 50%
• Talking to customers or users: 16% vs 26% - with "customer-facing" excluded (it usually describes the product, not conversations), PM ads mention feedback, interviews, and research more often

The chart below adds the mirror image: the roles share the end-to-end, AI, and collaboration language and part ways on strategy versus operations.

The stack items in the chart separate the roles most sharply and confirm the title describes an engineering job, not a rebranded PM.

AI language, then, is not what makes the role distinct - it is already in half of the product manager postings. A separate scan of 3,064 product-manager postings in Qarera's 360,336-posting dataset put AI at 37%, below my 52%; the gap is a useful reminder that source mix and keyword definitions move the level even when they agree on the direction. The difference is specificity: naming a concrete coding tool runs 38% against 6%, though the scan cannot tell engineers who use AI to work from engineers who build AI features.

Which tools, concretely: Anthropic, Cursor, and OpenAI lead in both roles, roughly five to ten times apart depending on the tool. The PM numbers are genuinely that low - even counting v0, Lovable, Replit, or Figma AI, only 11% of product manager postings name any specific tool; most of the remaining matched AI language is generic ("AI tools", "generative AI", "LLMs").

The rest of the named toolbox splits even more sharply:

Roadmap, stakeholders, prioritization, KPIs - coordination language is the backbone of the PM ad and marginal in the product engineer ones.

The discovery vocabulary is thin in both roles - and the literal "talk or speak to users or customers" is actually rarer in product manager postings (1.5%) than in product engineer ones (6.9%).

I would be careful about interpreting the lack of discovery language. Job descriptions describe outputs and responsibilities more clearly than the actual practice of talking to users, so the near-empty lanes say little about how either role spends its time. The absence of roadmap, stakeholder, and prioritization language in product engineer postings is more informative: these jobs are not being advertised as coordination roles.

As a sanity check, I read forty postings end to end, twenty per role, and fixed the dictionary where it missed real phrasings. That reading surfaced one more thing: two of the twenty PM postings ask the product manager to build and ship prototypes with AI tooling themselves. Role compression, from the PM side.

So does the PM role change?

My reading after all of this, as conclusions rather than a story:

• Product judgment is no longer described as something only product managers need. Nearly half of product engineer postings in my dataset mention product sense or a product mindset (46% vs 16% in PM postings), while roadmap, stakeholder, KPI, and prioritization language stays concentrated in PM postings. That does not show product engineers replacing product managers - it does suggest companies increasingly expect engineers to make decisions that would once have been treated as exclusively "product" work. A snapshot of two corpora, not a measured trend.
• The shift is not visible in hiring titles yet. In my fixed panel, product engineer postings went from 17 in July to 16 in August, and the product manager's share of intake did not fall: 4.24% to 4.42%. LinkedIn's 2026 U.S. workforce report finds the same split at a broader scale: AI has not yet dramatically changed hiring patterns, but it is already changing the skills employers ask for.
• The role seems to be changing from the inside instead. In an ICSE 2026 study of 885 Microsoft product managers, 62% of individual contributors reported using GenAI daily, yet the study's central conclusion was that accountability should remain human - and Microsoft is an unusually AI-forward environment, so that rate should not be read as market-wide. In my own audit, two of the twenty PM postings I read end to end ask the product manager to build and ship prototypes with AI tooling themselves.
• If role compression creates a hiring problem first, it will show up at the entry level. Senior titles are 41% of PM postings in my dataset, while junior, associate, and intern together are 2.5%. That does not prove AI caused the imbalance, but it leaves relatively little room for further compression at the bottom - and Gartner predicts that organizations relying on AI to cut junior roles will hollow out their own talent pipeline by 2028. Marty Cagan of Silicon Valley Product Group has argued a sharper version of this for years: the administrative, backlog-managing variant of the role - what he calls "product management theater" - is the easiest to cut.
• Adoption is earlier than the headlines suggest. Gartner's own numbers say 15% adoption today and over a year to scale a pilot, and one webinar analyst described AI-native software engineering as moving toward, if not already in, the trough of disillusionment. Meta is the counterexample on speed: as Reuters describes it, the company explored much more aggressive team compression, then backed away from part of the plan within months as the expected productivity gains failed to materialize.

Two months of comparable hiring data cannot validate or disprove a forecast for 2029. At best, this gives me a baseline - and right now that baseline is surprisingly boring: product manager hiring is not collapsing, product engineer hiring is not exploding, and companies are still advertising the two as distinct jobs.

The more interesting change is happening inside those jobs. Engineers are being asked for more product judgment; some PMs are being asked to prototype and build. Tiny teams may accelerate that overlap, but the hiring data does not yet show the organizational rewrite implied by the headlines. That is the part I want to keep measuring.

If you work in a product organization, I am curious about one thing: compared with a year ago, which decisions are now being made by a different role?

Thanks for reading!

Methodology in brief

• Dataset: 86,361 job postings collected between late June and early September 2026 from 488 sources.
• Deduplication and hygiene: postings are deduplicated across sources by company, title, and location; spam postings are excluded, and role counts use individual-contributor titles only.
• Classification is title-level: "product manager" counts individual-contributor PM titles, "product engineer" counts postings with that title. Neither inspects responsibilities.
• Month-to-month comparisons use a fixed panel of 358 company boards present and reliable across the whole period, with portals and aggregators excluded, each posting counted on the date it first appeared in my data, and each board's pre-existing backlog excluded.
• The phrase scan runs identical pattern dictionaries over both corpora (144 product engineer and 4,036 product manager postings).

Sources

• Gartner press release, "Gartner Predicts 60% of Organizations Will Adopt Smaller Software Engineering Teams by 2029", July 2026.
• Gartner webinar, "Tiny Teams: New Organizational Structures for AI-Native Software Engineering" (Keith Holloway, Aliyah Camacho), accessed September 2026.
• Gartner complimentary e-book, "Software Engineering 2030: The Impact of AI", 2025.
• U.S. Bureau of Labor Statistics, "Artificial intelligence, information technology, and employment, 2024-34", 2026.
• McKinsey, "The gen AI skills revolution: Rethinking your talent strategy".
• Reuters analysis, "Mark Zuckerberg had a bold plan to replace Meta staff with AI. Here's how it imploded" (Katie Paul, with Jeff Horwitz), August 2026.
• Marty Cagan, Silicon Valley Product Group, "Product Management Theater".
• LinkedIn Economic Graph, "Skills and AI: The U.S. Workforce Imperative", 2026.
• Mara Ulloa et al., "Product Manager Practices for Delegating Work to Generative AI", ICSE-SEIP 2026.
• Qarera, "The Most In-Demand Skills of 2026" (360,336 postings, Dec 2025 - Jun 2026; open dataset on Zenodo).
• Author's own analysis of 86,361 job postings collected June-September 2026.