The Fastest-Growing
Job of 2026

In February 2026, Harvard Business Review published “To Thrive in the AI Era, Companies Need Agent Managers” by Suraj Srinivasan and Vivienne Wei. The core argument: as companies deploy AI agents to handle real work — customer service, data analysis, content creation, operations — someone needs to manage those agents. Not build them. Manage them.

Think of it like being a project manager, but instead of managing people, you’re managing AI. You decide what work agents should handle, set them up, monitor their quality, fix what’s broken, and report on impact. HBR predicts this will be a standard job title at AI-first companies within 12–18 months.

Here’s the part that matters most: HBR specifically said that the most effective AI Agent Managers came from roles already accountable for service quality, customer outcomes, and operational judgment. They emphasized “deep domain expertise and lived understanding” over formal AI credentials. In other words — if you already know your job well and you learn how to use AI, you’re exactly who these companies are looking for.

An AI Agent Manager doesn’t write code. They don’t build models. They don’t need a computer science degree. Here’s what they actually do day to day:

Identify what work AI should handle. You look at your team’s workflows and figure out which tasks are repetitive, pattern-based, or time-consuming enough that an AI agent could do them. Not everything should be automated — the skill is knowing which things should be.

Set up and configure AI agents. You write the instructions, define the triggers, choose what data the agent can access, and test it until it works reliably. This is prompt engineering meets project management — you’re designing how the agent works, not coding it from scratch.

Monitor quality and performance. Once agents are running, you track how well they’re performing. Are they handling tasks correctly? What’s the error rate? Where are they escalating to humans? What’s the turnaround time? You catch problems before they become crises.

Manage human handoffs. AI agents can’t handle everything. Knowing when an agent should stop and hand off to a person — and making that transition smooth — is one of the most important parts of the role.

Report on ROI. Leadership wants to know: is this saving us time? Money? Headcount? You measure the impact of every agent and present it in terms the business cares about.

Thing 1

Audit your own workflows for automation opportunities. Open a document and list every task you do in a typical week. Next to each one, mark it as: AI could do this entirely (repetitive, follows a pattern, low-stakes), AI could help with this (needs human judgment but AI could draft, research, or organize), or Human only (requires relationships, creativity, or sensitive judgment). This is literally the first thing an AI Agent Manager does on the job. By doing it for your own role, you’re building the exact muscle the role requires — and you’ll probably find 3–5 hours of work per week that could be automated.

Thing 2

Build and manage your first AI agent. Don’t just use AI — manage it. Set up a Claude Skill with Dispatch (a scheduled task that runs automatically), or build a Notion Custom Agent, or create a Cowork task that runs in the background. The key difference: you’re not just chatting with AI. You’re designing a workflow, deploying it, and monitoring the output. Start simple — a daily morning briefing, a weekly report summarizer, or an inbox triage agent. Run it for a week. Track what it gets right and what it gets wrong. Fix the instructions. That test-deploy-learn loop is the core of the AI Agent Manager role.

Thing 3

Learn to write instructions like a manager, not a user. Most people write prompts like requests: “Can you summarize this?” An AI Agent Manager writes instructions like SOPs: “You are the weekly report agent. Every Friday at 3pm, pull all completed tasks from the Projects database, group them by project, calculate completion percentage, flag anything overdue, and create a summary page with these 5 sections.” Practice rewriting your own prompts this way. Be specific about the trigger, the data source, the output format, the edge cases, and the rules. This is prompt craftsmanship — one of the 5 skills HBR highlighted.

Thing 4

Document one process at work that you could hand to an AI agent. Pick the task from your audit (Thing 1) that’s most clearly “AI could do this entirely.” Write a full process document for it: what triggers it, what inputs it needs, what the steps are, what the output looks like, what “good” looks like, and what should cause it to escalate to a human. Don’t automate it yet. Just document it as if you were handing it to a new employee. This is the exact deliverable an AI Agent Manager creates before deploying an agent — and it’s the kind of thing you can reference in an interview or put in a portfolio.

Thing 5

Start tracking your AI results like a manager would. Create a simple tracker (a spreadsheet, a Notion database, or even a note on your phone) and log every time you use AI for a meaningful task this week. For each one, note: what you asked it to do, how long it would have taken you manually, how long it took with AI, and whether the output was usable as-is or needed editing. At the end of the week, add it up. That’s your personal ROI report. “This week I used AI for 12 tasks that would have taken me 8 hours. It took 2 hours instead. Output was usable as-is 75% of the time.” That’s the kind of data an AI Agent Manager reports to leadership — and it’s the kind of number that makes you impossible to ignore in an interview.

Your AI Agent Manager Starter Kit