Consultants used to fire up their computers every day to hunt through company databases, scan industry reports for relevant benchmarks, analyze competitor pricing strategies and craft multi-page reports complete with citations. Now, at some firms, AI agents are handling the key parts of that workflow. These workers don’t realize it, but they’ve already become AI managers.
The shift is happening subtly, but it’s happening. Workers are learning to prompt agents, navigate AI capabilities, understand failure modes and hand off complex tasks to AI. And if they haven’t started yet, they probably will: A new study from IDC and Salesforce found that 72 percent of CEOs think most employees will have an AI agent reporting to them within five years. This isn’t about using a new kind of software tool — it’s about directing intelligent systems that can reason, search, analyze and create.
Soon, the most valuable employees won’t just know how to use AI; they’ll know how to manage it. And that requires a fundamentally different skill set than anything we’ve taught in the workplace before.
What Skills Do Tech Workers Need in the Agentic AI Age?
The role of knowledge workers is shifting from data entry and manual analysis to AI agent management. They are evolving into conductors of intelligent systems, responsible for strategic oversight, quality assurance and workflow orchestration. This requires a fundamentally different skill set than traditional software use, focusing on clear communication and delegation to intelligent but imperfect AI systems.
The Agent Manager Is Emerging
Traditional software follows predictable patterns. Click here, get that result. Input data, receive output. AI agents operate more like remote employees: capable but requiring direction, prone to occasional mistakes and needing quality assurance checks.
Consider what happens when a consultant asks an agent to research market trends for electric vehicles. The agent must decide which sources to prioritize, how recent the data should be, whether to include regulatory factors and how to structure the analysis. Each decision point requires judgment, so the human’s role shifts from data entry to strategic oversight.
I work directly with companies to build custom agent workflows, and I’ve found that in the most successful deployments, users think like managers, not just prompt writers. They understand that getting accurate results from an agent requires the same skills as managing a junior analyst: clear expectations, quality checks and iterative feedback.
Beyond Prompt Engineering
Companies have invested significant effort in training workers to craft effective AI prompts, as better instructions yield better outputs. But managing agents requires distinct competencies that traditional tech skills and basic AI training don't cover. Here are the key skills that “agent managers” need:
Delegation Strategy
Workers need to know which tasks are well-suited for which agents and when to break complex tasks into smaller components versus letting an agent handle everything end-to-end. A financial analyst might direct one agent to gather earnings data while another analyzes sentiment from earnings calls then synthesizes the results.
Quality Assurance Methodology
Unlike debugging code, evaluating agent output requires domain expertise and contextual judgment. When an agent cites more than 100 sources in a research report, the human manager needs to assess whether more citations indicate thoroughness or information overload.
Orchestration Across Multiple Agents
As workflows become more complex, managers will coordinate teams of specialized agents. The core skill here is workflow architecture: designing how one agent’s output becomes another’s input, mapping dependencies so agents execute in proper sequence and building checkpoints to prevent errors from compounding downstream. This also requires managers to understand where human intervention is most valuable and how to use AI’s capabilities while accounting for agents’ limitations.
The AI Management Training Gap
Here’s the problem: Nowhere in corporate training programs will you find “AI Agent Management 101.” Business schools teach project management and software skills, but not, at least right now, how to delegate to systems that can think but can’t always be trusted.
The companies succeeding with AI agents are creating their own training programs. They’re teaching employees to write detailed agent prompts the same way they’d brief a new hire. They’re establishing quality benchmarks and review processes. Most importantly, they’re helping workers understand that managing agents isn’t about perfect automation. Instead, it’s about fostering productive collaboration with intelligent but imperfect systems.
Training programs also help identify “resident experts” within the organization who excel at and enjoy learning agent management techniques. Leaders should provide these experts with resources and support to help others get comfortable with new AI skills.
And companies that provide AI agents should recognize a business opportunity: Their customers’ success depends not just on the technology, but on developing agent management capabilities. Forward-thinking AI companies are offering role-specific AI training to clients, structured around actual workflows.
For instance, a sales training might walk reps through delegating prospect research to an agent, then teach them to evaluate whether the agent surfaced the right decision-makers and pain points before using that intelligence for outreach. Product managers learn different skills, including how to prompt agents to synthesize user feedback themes, verify the agent didn’t miss edge cases and identify when human judgment is needed to prioritize features.
Preparing for the Agentic Workplace
The transition is already underway, but luckily, most workers can start building these skills immediately. Here’s where to start.
Practice Constraint-Based Thinking
Stop asking “What can this AI do?” Instead, ask “What specific task needs to be completed, with what quality standards and how will I verify the results?” This mirrors how effective managers think about delegation.
Develop Evaluation Frameworks
Start establishing personal benchmarks for AI output quality. If you regularly use AI for writing, create criteria for tone, accuracy and completeness. Apply the same systematic thinking to research, analysis or any repetitive workflow.
Experiment With Multi-Step Workflows
Try breaking larger tasks into agent-manageable components. Instead of asking for a complete competitive analysis, first request market size data, then competitor profiles, then strategic implications. Notice how managing the handoffs changes your approach. Maybe you find yourself giving the AI agent more guidance because you have more reference material to work with, or you could find ways to adapt the report halfway through the process.
Get Comfortable With Ambiguity
Agents won’t always produce identical results for identical prompts. Learning to work with intelligent but non-deterministic systems is a skill that transfers across all AI tools. When outputs vary, you need consistent evaluation criteria to judge quality — whether you’re assessing research depth, writing tone or data accuracy. AI’s variability makes having clear benchmarks even more critical since you can’t rely on identical outputs as proof of quality.
The Future Is Management, Not Replacement
The narrative around AI replacing jobs misses a more nuanced reality. In most knowledge work, AI is creating a new layer of management responsibility for both managers and ICs. Workers aren’t becoming obsolete. Instead, they’re evolving into conductors of intelligent systems.
The professionals who thrive in this transition will be those who recognize that managing AI agents requires the same fundamental skills as managing people: Clear communication, quality oversight and strategic thinking about how to break down complex problems.
Companies that invest in developing these capabilities now will have a significant advantage as agentic AI becomes standard practice. Those that don’t risk finding themselves with a workforce trying to solve tomorrow’s problems with yesterday’s skill sets.
The age of AI agent management has already begun; the only question is whether your organization is ready to capitalize on it.
