The Agent Manager Era

Why your next hire might report to an AI — and why you need to learn to manage both.

---

Building AI agents is easy now. Managing them at scale is the hard part nobody prepared you for.

Salesforce just reported Q4 earnings that should make every executive pay attention. Not for the revenue beat, but for what they're now measuring: Agentic Work Units (AWUs) — a metric tracking work completed autonomously by AI systems. The company is moving beyond chatbots toward multi-agent systems where specialized AI handles sales, finance, and supply chain tasks with minimal human intervention.

This isn't experimental technology. This is how enterprise work is being restructured in 2026.

The Shift Nobody Announced

Here's what changed while you were still debating chatbot pilots:

Low-code platforms made agent creation trivial. Microsoft Copilot Studio, Zapier, and dozens of similar tools mean anyone in your organization can spin up an AI agent in an afternoon. The technical barrier that once protected companies from runaway automation is gone.

The differentiation moved downstream. Building agents is no longer the competitive advantage. Scaling them sustainably is. Companies treating agent creation as the finish line are already drowning in fragmentation.

A new management layer emerged. EY's global chief innovation officer Joe Depa predicts this will define 2026: "You're going to have to orchestrate the agents, train them, audit them, retire them as new agents come in. New skills are going to be developed."

Read that again. Orchestrate. Train. Audit. Retire. These aren't technology tasks. These are management tasks.

What Agent Management Actually Looks Like

Traditional management involves setting direction, allocating resources, monitoring performance, and developing talent. Agent management isn't that different — but the execution is radically new.

Orchestration replaces delegation. When you delegate to a human, they figure out the "how." With agents, you're responsible for ensuring multiple AI systems coordinate across workflows, hand off tasks correctly, and don't create conflicting outputs. This is systems thinking applied to digital workers.

Training becomes configuration. You're not teaching agents through mentorship or feedback conversations. You're defining guardrails, selecting appropriate models for specific tasks, and embedding governance into their operations. The skill is precision, not patience.

Auditing replaces reviews. Performance reviews happen annually. Agent audits need to happen continuously. As agents gain the ability to act across systems — approving expenses, drafting contracts, responding to customers — oversight becomes a live operation, not a periodic check-in.

Retirement replaces offboarding. Humans leave for better opportunities. Agents become obsolete when better models emerge. You need processes to phase out underperforming or outdated agents while maintaining workflow continuity.

The Skills Gap Nobody's Addressing

Most leadership development programs are still teaching pre-agent skills. They're preparing executives to lead teams of humans who happen to use AI tools.

That's yesterday's job.

Tomorrow's executive needs to lead hybrid organizations where humans and AI agents collaborate as peers. Where some "reports" are digital workers operating 24/7 across time zones. Where accountability flows through both human judgment and algorithmic guardrails.

Three capabilities matter most:

1. Systems orchestration. Understanding how workflows connect, where handoffs fail, and how to design processes that work regardless of who — or what — is executing them. This is architecture, not management.

2. Governance design. Every agent operating in your organization needs boundaries. What can it approve? When must it escalate? What data can it access? Designing these constraints requires understanding both the technology and your risk tolerance.

3. Model flexibility. Different tasks require different AI capabilities. Knowing when to use a powerful reasoning model versus a fast execution model versus a specialized domain model is the new resource allocation skill.

The Accountability Problem

Here's what keeps CIOs awake: when an agent makes a mistake, who's responsible?

The answer is still "you." Agents don't absorb accountability. They execute within parameters you define, using permissions you grant, operating within workflows you approve.

This is why governance isn't optional. Technologies like Model Context Protocol (MCP) and structured workflows help embed oversight into agent operations. But the design choices are still human decisions.

Companies scaling agents without governance infrastructure are building liability, not leverage.

What To Do Monday

Inventory your agents. Most organizations already have more AI agents operating than leadership realizes. Shadow IT didn't disappear — it evolved. Find out what's actually running before you can manage it.

Define ownership. Every agent needs a human accountable for its performance, permissions, and retirement. If no one owns an agent, it shouldn't exist.

Build your orchestration layer. Whether through dedicated platforms or internal protocols, you need visibility into how agents interact across systems. Fragmented automation creates more problems than it solves.

Start developing the skill. The executives who learn to manage hybrid human-AI organizations in 2026 will lead them in 2027. The learning curve is steep, but the early movers have an advantage.

The Bottom Line

We've spent three years asking whether AI will replace managers. Wrong question.

The right question: will managers learn to manage AI?

Adaptability, as Depa puts it, is the new job security. Jobs are changing fast. The executives who treat agent management as a core competency — not a technical detail to delegate — will thrive in what's emerging.

The ones who don't will find themselves managing increasingly small human teams while their competitors scale with digital workforces.

The agent manager era has arrived. The only question is whether you're ready for it.

---

Tommy Kenny is the founder of Digital Executive Insight and author of Pragmatic Disruption. He advises executives on navigating AI transformation without the hype.