If software can summarize meetings, track projects, analyze performance, and increasingly execute the work itself, why would companies still need this many managers? 🤷♀️
Fair question. I just think it starts from the wrong definition of management.
Management was never really about moving tasks between people or writing status reports. It's about creating clarity, making decisions, allocating resources, developing people, and taking responsibility for outcomes. AI is getting very good at the first layer of that. What it can't absorb is the responsibility for deciding what an organization should do, and who's accountable when it goes wrong.
Which creates a paradox: the more autonomous these systems get, the more valuable human ownership becomes.
From assistants to coworkers
Most workplace AI today just makes individuals faster: drafting emails, summarizing meetings, crunching a spreadsheet. The bigger shift is AI moving from assisting tasks to running whole workflows on its own.
Microsoft calls this the rise of the "agentic organization." Its 2026 Work Trend Index, based on 20,000 knowledge workers across 10 markets, found the number of active agents in the Microsoft 365 ecosystem grew roughly 15x year over year 🤯
An agent can now monitor a process, make bounded decisions, trigger actions, and report results, with no human in the loop until something actually needs a call. At that point, "make sure everyone's doing their tasks" stops being the job, because the system already does that. What's left for the manager is context, direction, constraints, accountability.
I've seen this play out on our own team already: nobody's chasing status updates anymore, but someone still has to decide whether the automated recommendation is actually the right call for the customer.
The accountability gap is already here
KPMG's Global AI Pulse Q2 2026 surveyed over 2,100 organizations and found only 24% name the CEO or exec committee as ultimately accountable for AI-informed decisions. 29% point to a specific C-suite exec. The rest is scattered across teams with no clear owner.
BCG's 2026 AI at Work survey, with almost 12.000 respondents across a dozen markets, found half of companies lack clear governance for managing human-and-AI teams, with accountability ranked as a top-3 concern going forward.
Two different research houses, same gap. If an agent recommends something, a human signs off, another system executes it, and a customer gets hurt, "the AI" isn't an answer. There has to be a human owner: who approves, who can pull the plug, who's on the hook when it fails.
Those are management questions, not technical ones. No org chart update fixes them on its own.
The "middle-management apocalypse", now with real numbers
There's a legitimate case that some management roles are exposed. If your main job is collecting and relaying status updates, automation eats a lot of that, and honestly, it probably should.
This is no longer a slow-burn forecast. Gartner projects that by the end of 2026, one in five organizations will use AI to flatten their hierarchy, cutting more than half their middle-management layer.
But cutting that kind of manager isn't the same as cutting management. The line is between coordination and judgment. A manager relaying status is easy to replace. A manager deciding which of three competing priorities the company should actually chase is not.
There's a catch, though: most managers aren't positioned to make that case yet. A 2026 proficiency study found 54% of workers rated themselves proficient with AI, but only 10% actually tested as proficient, and managers scored barely better than the people they manage. It's worth sitting with that gap for a second before assuming you're on the safe side of it.
Fewer people, more systems
The manager of the future may not run a bigger team, just a more complicated one: humans, AI agents, automated workflows, and software constantly generating recommendations. You can't supervise every action manually in a setup like that.
So the job shifts to designing the environment those actions happen in. Which decisions need human sign-off? What can an agent do unsupervised? What gets audited? What happens when two automated systems disagree? 🙃
That's not really people management anymore. It's systems design. And it reframes the core shift from task ownership to outcome ownership: the analysis can be automated, but deciding whether it's right, and what to do about it, stays human.
The real skill: knowing what not to automate
Most of the current conversation is "can AI do this?" The sharper question is:
Should we automate this?
Not every decision benefits from autonomy. Some are high-stakes, hard to reverse, or too ambiguous to reduce to a clean objective. A rule I keep coming back to:
Automate execution before you automate accountability.
Let AI prep the analysis and run the routine work. For anything consequential, keep one clearly identified person with the authority and the responsibility to make the final call.
So what's the manager actually for?
AI won't eliminate management. It'll expose which parts of the job were valuable and which were overhead wearing a manager's title.
Less coordination, more judgment. Less reporting, more deciding. Less task ownership, more outcome ownership.
The question worth asking isn't what AI can do. It's who owns what happens when it does it.
That person isn't going anywhere.
Top comments (1)
The line “automate execution before you automate accountability” really hits the point.
I think the biggest change will be that managers spend less time supervising activity and more time designing the boundaries around autonomous systems. What can run freely, what needs approval, what gets logged, and what happens when the system makes the wrong call.
There’s also an engineering parallel here: giving an agent more autonomy without defining failure handling and ownership is basically shipping a system without an incident process.
Fewer managers might make sense. Fewer accountable decision-makers probably doesn’t.