AI Agents Are Targeting Middle Management First - Anupam Mittal's Warning to Leaders
Anupam Mittal, founder of People Group and a judge on Shark Tank India, didn't mince words: AI agents aren't coming for coders first. They're coming for middle managers.
His core point is blunt and useful for anyone in management: the traditional "knowledge premium" - knowing the process, who to call, and how to coordinate - is collapsing. AI agents can coordinate, draft, route, report, and follow up across systems faster and cheaper. That puts roles built mainly on orchestration at risk.
Key takeaways from Mittal's note
- "AI is not coming for coders first. It is coming for middle managers."
- The old advantage of seniority-as-process-knowledge is fading to near zero.
- He's invested in companies with 300-1000 crore ARR run by ~50 people using suites of AI agents.
- Titles like "VP of Operations" that don't directly operate or build are endangered.
- The future belongs to "Individual Contributor Plus" - people who build, code, create, or sell, amplified by Gen AI to do the work of entire teams.
- In a high-interest-rate world, overhead gets trimmed - especially roles that primarily "coordinate" without measurable output.
Big Tech is already flattening
Amazon signaled this shift, putting more ownership in the hands of people doing the work rather than layers of managers who want a "fingerprint" on every decision. That came straight from CEO Andy Jassy in an interview with Bloomberg. Source
Google reportedly removed an entire layer in parts of its US ads org, cutting "managers of managers" to speed decisions and align output with revenue. Source
What this means for management
Your value must move from "I coordinate" to "I create measurable outcomes." If your day is pre-meetings for pre-meetings, AI will outpace you. If you build systems, ship improvements, and directly move metrics, you'll be in demand.
The orgs that win will compress layers, empower ICs, and plug AI agents into core workflows - not just as pilots, but as the backbone of operations.
Where AI agents are swallowing coordination
- Status and exec reporting: auto-generated updates from live systems
- Cross-team orchestration: ticket routing, handoffs, and SLAs
- Vendor and ops workflows: intake, validation, approvals, compliance
- Sales ops and CS ops: call summaries, next steps, follow-ups, renewals
- Project hygiene: risk flags, dependency checks, blockers, nudges
How to adapt your org now
- Audit "coordination-only" roles. List every role whose primary output is meetings, follow-ups, or status updates. Attach cost and business impact. Decide: redesign, reskill, or reduce.
- Redesign for IC+. Push authority and budget to the people producing work. Convert manager headcount into senior ICs who ship and own outcomes.
- Instrument everything. Build dashboards tied to revenue, margin, cycle time, and quality. If a role can't show its metric, it will be treated as overhead.
- Deploy an agent stack. Start with high-volume workflows: reporting, customer comms, QA checks, data entry. Integrate with your CRM, ticketing, and data warehouse.
- Change incentives. Reward shipped work, cycle-time reduction, and margin expansion - not meeting count or headcount managed.
- Retrain managers. Move them into ops engineering, automation, data analysis, solutions architecture, or quota-carrying roles.
Career playbook for current middle managers
- Pick a lane with output. Product, growth, sales, ops engineering, data - anything where you can point at shipped work or revenue.
- Get hands-on with Gen AI. Build prompts, automate a report, wire up an agent to your tools. Show a before/after metric, even if small.
- Own a number. Cycle time, conversion, retention, cost per ticket, NPS - make yourself the person who moves a metric monthly.
- Cut meetings in half. Replace status with live dashboards and async updates. Use meetings only for decisions.
Org patterns that work
- Fewer layers, clearer ownership. One owner per objective, with authority and budget.
- AI as a team member. Treat agents as standard capacity in staffing plans.
- Small teams, big scope. Senior ICs with 1-2 analysts/engineers plus agents can replace entire legacy pods.
- Fast feedback loops. Weekly shipping, monthly refactors, quarterly refits of the agent stack.
Simple 30-day action plan
- Week 1: Identify top 5 repetitive coordination workflows. Quantify time and cost.
- Week 2: Pilot AI agents on two workflows. Measure cycle time and error rates.
- Week 3: Publish live dashboards. Replace status meetings with async updates.
- Week 4: Lock org changes: trim layers, convert two roles to IC+, revise incentives.
If you're hiring
- Prioritize ICs who can design systems, automate, and ship without layers of support.
- Screen for prompt fluency, data comfort, and tool integration experience.
- Ask for proof: before/after metrics, dashboards, process maps, shipped automations.
This isn't anti-management. It's pro-output. Keep the leaders who set direction, clear blockers, and engineer systems. Reduce the layers that repackage information.
If you want a structured way to upskill your team for IC+ roles, explore role-specific AI learning paths: Complete AI Training - Courses by Job.
Bottom line: shift from coordination to creation, from meetings to metrics, and from titles to shipped outcomes. The orgs that make this turn now will set the pace for everyone else.
Your membership also unlocks:
