Why you shouldn't fire your managers yet: The McKinsey report on flattening the org

If you fire your Middle Managers today, your GCC collapses tomorrow. Everyone is reading the McKinsey report on "Flattening the Org." Everyone is excited about Agentic AI reducing headcount. Here is the operational reality: AI Agents handle the "Happy Path" (SOPs, standard code, data entry). They do not handle the "Edge Case" (The angry client, the ambiguous requirement, the crisis). If you remove the human layer today... You remove the Safety Valve. When the bot fails (and it will), there is no one left to catch the ball. The Strategic Shift: Don't fire the manager. Repurpose them. Move them from "Task Monitor" to "Exception Handler." Design your org chart for the Crisis, not just the Speed. Is your operation resilient enough to survive a bot failure? #GCC #AgenticAI ##OperationalResilience #OrgDesign ##RiskManagement

That classic line by Warren Bennis comes to mind: “The factory of the future will have only two employees, a man and a dog. The man will be there to feed the dog. The dog will be there to keep the man from touching the equipment.” It’s a reminder that in highly automated environments, human work doesn’t vanish—it shifts from direct execution to oversight, governance, and system design. The real design challenge is ensuring resilience. Repurposing managers into exception handlers and crisis navigators is how you balance efficiency with long‑term stability. Automation handles the routine; people handle the exceptions that protect the business.

I slightly differ. Agentic AI isn’t a traditional bot as it reasons, plans, adapts & goes far beyond “happy paths.” But you’re right: mid-managers shouldn’t be removed, as with right upskilling, they bring what Agentic systems still lack — context, judgment, stakeholder navigation, & turning ambiguity into clarity. In an Agentic world, mid-managers will become the human layer that amplifies intelligence & keeps system resilient.

Agree. People forget that most of the real damage happens in the edge cases, not the routine work. Middle managers are usually the only layer that knows how to absorb chaos when things break. Repositioning them as exception handlers makes far more sense than cutting them in the name of speed.

Interesting perspective. But real GCC operations don’t work like this. Middle managers aren’t a magic safety net — they mostly escalate edge cases, not solve them. AI doesn’t fail because managers are missing; it fails when the org hasn’t redesigned the process or exception logic. The real shift isn’t “save managers.” It’s “fix the system.

Many organizations today remain fixated on short-term targets and immediate revenue gains rather than investing in customer sentiment or operational excellence. In contrast, niche players tend to achieve their goals more sustainably—whether through organic expansion or strategic acquisitions—rather than relying solely on squeezing margins.

Garima Pandey Brilliant point — AI speeds the happy path, but only humans can stabilize the chaos. Middle managers aren’t overhead; they’re the safety net. Do you think companies are ready to redesign roles around exception-handling, not just efficiency?

I just saw that in real-time. I ordered a laptop and had to cancel order due to delay in delivery. However their primary CRM is a WhatsApp bot which had limited responses. Could not connect with a human as their customer service has limited staff. I decided never to order again thus the company lost a client.

I would like to share a frustrating experience of complaining to Meesho about the delivery experience. The call was answered by a bot and there was absolutely no empathy no value for the customer. Definitely automation for Frontline teams would certainly lead to disaster in the event of crisis.

Absolutely spot on Garima Pandey! Middle managers aren’t inefficiencies; they’re the stabilizers of complex operations. AI can handle the routine, but only humans manage ambiguity, escalation, and judgment.

See more comments

To view or add a comment, sign in

Explore content categories