Will AI Drive the Great Flattening?

Pro tip: When the post title is a yes-or-no question, the answer is always ‘no’.

Q: What do managers even do, anyway?

A few days ago a LinkedIn post predicted that dedicated Engineering Managers (EMs) will be extinct by 2027. The post lands in a time of fears about AI-driven job losses, but it cuts deeper because the role of middle management has always been questioned. In this case, the post framed EMs as little more than Jira traffic cops. The real problem is that the essential work of middle management is invisible – when we can’t see the craft, it’s easy to assume it adds no value. Other people’s jobs always look easy from the outside.

A: Middle managers are plumbers for an invisible system

Here's what often gets overlooked: Every product org is a socio‑technical system that produces customer value. A maze of people, tools, processes, hand‑offs, incentives, and feedback loops. When those flows hum, customers cheer. When they clog, fingers point. 

You can watch a manager run a 1‑on‑1 but you can’t see how it subtly re-wires the system. Results – higher quality, faster delivery – arrive months later and look like “good team culture.” The combination of soft activities, delayed results, and ambiguous attribution all add up to a role that has always looked expendable.

System Stewardship 101

W. Edwards Deming (quality pioneer) estimated that 94 % of performance swings come from the system, not the people, while Russ Ackoff (systems theorist) warned that optimizing the parts in isolation can cripple the whole. If the system matters most, someone must be responsible for it.

Distributed Ownership Doesn't Cut It

Autonomy is energizing, but Diffusion of Responsibility (if everybody owns it, nobody owns it) plus Ackoff’s Law (optimize the parts, sub‑optimize the whole) hit hard. Remove the dedicated system steward and every squad chases its own local metric – velocity, cloud costs, bug count – blind to the shared chokepoint. The extra load slams into the real system-wide constraint (e.g. the Friday release gate, the on‑call queue); overall throughput sinks even while each team claims a win.

One Steward at the Top Isn’t Enough

A CEO or CTO can’t observe every flow, so we slice the maze into smaller chunks that a single human can see and navigate end‑to‑end, and make that person accountable. Titles vary – EM, group lead, value‑stream coach – but the purpose is the same: find the bottleneck, trim process bloat, and keep incentives pointed at the North Star.

AI may do that job some day, but all the headline hype aside, today it's not even close.

What AI Actually Changes

Automation, including AI, continues to strip out the chores that once kept whole layers busy – status roll-ups, burndown math, slide-deck assembly. Bots can now draft the retro, escalate cycle-time outliers, and page you when the database sneezes. Yet the hard jobs remain:

• Make sense of the firehose. Dashboards don’t interpret themselves. Someone has to spot when employee turnover spikes two quarters after a culture dip.
• Coach people through the messy bits. A model can suggest learning modules, but it can’t coax a junior through a midnight panic or mediate a roadmap turf war.
• Tend the system itself. With less paperwork in the way, bottlenecks and process bloat become visible sooner – if a steward is watching. The craft shifts from collecting data to acting on it before friction compounds.

Tomorrow’s manager, then, looks less like a traffic cop waving tickets and more like a performance coach armed with a live HUD: fix the flow, grow the people, guard the guardrails. I'll let you in on a secret: That’s what good managers already do today – AI just hands them sharper instruments.

Project Oxygen

But hey, experimenting is always good. Facts over feelings and all that. Fortunately, Google took the plunge for us when they fired every manager in 2002. Their decision lasted for three months, which was how long it took for the founders to became bottlenecks for day-to-day issues like conflict resolution. Later, Google's Project Oxygen showed that teams with strong managers had higher satisfaction and lower turnover – “the smallest incremental increases in manager quality were quite powerful.” Looks like those managers were doing something after all.

How to Tell If Your Org Thinks in Systems

Plenty of teams talk the talk of systems. A three-question gut check cuts through the theatre. Ask these out loud at your next retro – no slides, no hedge words:

1. Can every person on the team name today’s bottleneck? If answers differ, the team is already pulling in different directions.
2. Can they name who owns that bottleneck end-to-end? “Everybody” is code for “nobody.” A single steward needs the charter – and the authority – to fix what they find.
3. Is that steward's effort measured on a flow metric the customer would feel? Throughput, lead time, escaped-defect rate – pick one that spans the value stream. Local KPIs (cloud costs, story points) invite sub-optimization.

Three clear yes answers mean your system has a heartbeat and a caretaker. Miss even one and no amount of LLM-powered automation will save you from the next release-gate pile-up – you need system work more than you need automation.

Bottom Line

Middle‑management bloat is waste that should be culled, but middle‑management craft – tuning the socio‑technical system that turns intent into shippable value – is still a crucial investment. Fire the stewards and invisible plumbing will burst. Keep them, amplify them with AI, and the maze flows faster than ever.

Question for your org: Who owns the current bottleneck, and how will you know when they’ve won?

People, processes, and tools evolve. Principles endure. Soren Brook's mission is to illuminate the enduring principles so that we can get away from the turbulence. Calm minds, clear moves.