The Silicon Valley Product Group has published a new post arguing that AI will accelerate a shift away from middle management in product organizations, leaving a workforce dominated by skilled individual contributor builders.

The piece, titled "Experts Lead Experts," opens by conceding that the long-term impact of AI on knowledge work remains unclear. The thesis SVPG is advancing is concrete: managers and other layers of leadership will become significantly less important, while individual contributors who can ship product carry most of the load.

The framing lands in a market primed for the message. Public companies have publicly cut layers over the past two years, and PM conferences have repeatedly featured panels on the future of product management in AI-first organizations. SVPG, the advisory firm behind several widely-cited product frameworks, remains one of the more influential voices in product practice.

What the post actually claims

The post is short on specifics and reads more like a position statement than a data-driven analysis. It does not cite headcount figures, manager-to-IC ratios, or productivity benchmarks. The argument rests on an assumption: AI tooling compresses the work middle managers traditionally performed — status meetings, status updates, coordination overhead, and the packaging of ambiguous work into routinized tasks.

For product managers, the practical question becomes: what happens to the roles that sit between strategy and execution — the directors, group PMs, and senior PMs whose primary function is to translate fuzzy directives into roadmaps for ICs?

Where the argument holds, and where it strains

The argument is most credible for organizations where managers functioned as meeting schedulers and slide reviewers. In those setups, generative AI tools — from documentation agents to roadmap drafters — can plausibly absorb the coordination layer. Teams of 4-8 ICs that previously needed a dedicated lead PM now operate, in some companies, with a single staff PM acting as a "player-coach" doing real IC work themselves.

The argument strains in three places.

• Performance management and career growth. Most managers do real work even when their calendars look bloated. AI does not currently conduct credible calibration conversations or produce defensible promotion packets.
• Cross-functional coordination outside engineering. Legal, sales, finance, and customer-facing functions sit outside the tooling wave, and someone still has to hold the connective tissue.
• Senior judgment on tradeoffs. AI can model a decision tree; it does not replace accountability for picking a branch.

Historical flattening experiments — most famously the mid-2010s holacracy push at Zappos — generally re-added structural layers within 24-36 months once complexity reasserted itself. The SVPG post does not address that prior evidence.

What PMs should test on Monday

Treat this as a forcing function for an org-design conversation, not a finished prescription.

• Audit your management layer for genuinely distinct work — calibration, career conversations, cross-functional escalation — versus work AI could absorb in the next two quarters.
• Define the player-coach PM role explicitly. Without a written job charter, "we're flattening" usually means senior PMs absorb IC work and middle managers absorb people management, with no net gain.
• Set a six-month review checkpoint. Flattening that improves shipped product at 90 days usually sticks; flattening that improves meeting calm at 90 days usually re-inflates within a year.

The forward view

The "Experts Lead Experts" post is part of a broader 2024-2025 narrative, echoed by senior leaders at major tech firms, that AI will compress knowledge-work hierarchies. The honest read for product leaders: the trend is real at the coordination layer, will partially reverse at the judgment layer, and any specific prediction of "fewer managers in 2030" remains under-evidenced. The product teams that get this right will be the ones that flatten deliberately, instrument the outcome, and leave themselves a defined re-layering path.