The public conversation about AI and jobs has been bizarrely fixated on the wrong targets. Every week there’s another article about whether AI will replace software developers, graphic designers, or writers. These are visible, easily demonstrated capabilities — you can show an AI generating code or creating images in a tweet-sized demo. They make for good headlines.

But talk to anyone actually deploying AI agents in enterprise settings, and they’ll tell you a different story. The roles facing the most immediate, practical displacement aren’t creative or technical roles. They’re coordination roles. The people whose primary function is to gather information from multiple sources, synthesise it, route it to the right people, and follow up to make sure things happen.

In other words: middle management.

What Middle Managers Actually Do

I don’t mean this dismissively. Middle management is often caricatured as pointless overhead, but that’s unfair. Good middle managers perform essential functions.

They translate strategy into execution. A VP says “we need to improve customer retention by 15%.” The middle manager figures out what that actually means in practice — which teams need to do what, in what order, by when.

They aggregate information upward. They collect status updates, spot risks, identify blockers, and present a coherent picture to leadership so decisions can be made without executives needing to understand every detail.

They coordinate across teams. When Marketing needs something from Engineering needs something from Legal, someone has to manage the dependencies, the timelines, and the communication. That’s often a middle manager.

They follow up. A huge portion of middle management work is simply ensuring that things that were decided actually get done. Check-ins, status meetings, reminder emails, escalation when deadlines slip.

These are critically important functions. They’re also almost entirely information-processing functions. And that makes them vulnerable.

Why AI Agents Fit Here

AI agents — autonomous systems that can plan, execute multi-step tasks, interact with tools, and adapt to changing conditions — are getting remarkably good at exactly this kind of work.

Consider what an AI agent can do today, not in theory, but in practice with existing tools.

Information gathering and synthesis. An agent can pull data from your CRM, project management tool, financial system, and customer support platform. It can compile a status report that would take a human manager two hours of checking dashboards and chasing Slack messages. It can do this every morning at 6am without being asked.

Routing and notification. When a customer escalation comes in, an agent can assess severity, pull the customer’s history, identify the relevant team, and route the issue with full context. No human needs to play traffic cop.

Follow-up and accountability. An agent can track every action item from every meeting, check whether they’ve been completed by the deadline, and send targeted reminders to the responsible parties. Not a generic “here’s your weekly reminder” email, but a specific “this deliverable was due yesterday, here’s the context, here’s who’s waiting on it” message.

Cross-team coordination. If an engineering deployment is delayed and that affects a marketing launch date, an agent can detect the dependency, notify the marketing team, suggest alternative timelines, and update the project plan. It doesn’t need a meeting to do this.

This Is Already Happening

The displacement isn’t theoretical. Companies using tools like Notion AI, Asana Intelligence, Monday.com’s AI features, and various custom agent deployments are already reducing coordination overhead.

A mid-size Australian professional services firm I spoke with recently consolidated three project coordinator roles into one, supplemented by an AI agent system that handles status tracking, client updates, and resource allocation alerts. The remaining coordinator focuses on relationship management and complex problem-solving — things the agent can’t do.

An e-commerce company reduced their operations management team from five people to three by deploying agents that handle inventory alerting, supplier communication for routine reorders, and shipping exception management. The three remaining managers handle strategic decisions, supplier negotiations, and edge cases the agents flag for human attention.

These aren’t dramatic, newsworthy layoffs. They’re quiet restructurings. A role opens up through natural attrition and doesn’t get filled because an agent is handling most of what the person did. A team of six becomes a team of four. Nobody holds a press conference.

What This Means for the Future of Work

If AI agents increasingly handle coordination, information synthesis, and follow-up, the roles that remain are the ones that require:

Judgment under ambiguity. When the situation is genuinely novel and the right course of action isn’t clear from historical data, you need a human. Agents can flag the ambiguity, but someone needs to make the call.

Relationship management. Building trust with clients, navigating interpersonal conflicts, motivating underperforming team members, reading body language in a negotiation — these remain profoundly human functions.

Creative problem-solving. Not the kind of “creativity” that means generating images or writing copy, but the strategic creativity of seeing a market opportunity that doesn’t show up in the data, or restructuring a business model in response to a competitive threat.

Ethical reasoning. Deciding when to bend a rule, when a policy doesn’t fit a specific situation, when the numbers say one thing but fairness demands another.

The uncomfortable truth is that these capabilities aren’t evenly distributed across current middle management roles. Some managers spend 80% of their time on coordination and 20% on judgment, relationships, and creative problem-solving. For those roles, the coordination portion is automatable now, and the remaining 20% doesn’t justify a full-time position.

Other managers spend most of their time on the genuinely human elements — they just happen to also do some coordination work. Those roles aren’t going anywhere. They’re probably getting better, because the coordination drudgery gets taken off their plate.

The Honest Conclusion

AI isn’t going to eliminate middle management entirely. That’s a dramatic prediction that doesn’t match how organisations actually adapt. What it’s going to do is reduce the number of people needed for coordination and information-routing functions, while increasing the premium placed on judgment, relationships, and strategic thinking.

The managers who will thrive are the ones who are already good at the human parts of the job. The ones who’ll struggle are those whose primary value has been in keeping the information flowing — because that’s exactly what AI agents are built to do.

It’s less dramatic than “AI replaces all the coders,” but it’s more real. And it’s happening right now.