Managing Agents Like People: Why the Old Rules of Leadership Still Apply in the Age of AI

In the not-so-distant past, IT teams worried about uptime, infrastructure, and whether someone had accidentally pushed to production on a Friday afternoon. Today, there’s a new “colleague” in the mix: the agent.

Agents, AI-driven systems capable of executing tasks, making decisions, and interacting with other systems—are rapidly becoming embedded in modern IT environments. They triage tickets, write code, manage workflows, and sometimes even explain themselves with unnerving confidence. And while they don’t need coffee breaks or complain about meetings, they introduce a new kind of challenge: you are no longer just managing systems, you are managing behaviour.

Here’s the surprising part: the most effective way to manage agents looks a lot like managing people.

Not because agents are people, but because once you give something autonomy, direction, and responsibility, the same principles of clarity, leadership, and oversight apply. Organizations that recognize this early are seeing better outcomes, fewer surprises, and far less “why did it do that?” moments.

Managing agents well comes down to three core pillars: strategic thinking, leadership, and management. Ignore any one of them, and things can go sideways quickly, albeit very efficiently.

1. Strategic Thinking: If You Don’t Know What You Want, Your Agents Won’t Either

This starts with strategy, and this is where most things go wrong.  When managing people, unclear goals lead to confusion, duplicated effort, and the occasional existential crisis. With agents, unclear goals lead to something arguably worse: precise, scalable misalignment. Agents are exceptionally good at doing exactly what you ask… not necessarily what you meant. That’s why strategic thinking isn’t just a high-level exercise, it’s operationally critical. Before deploying an agent, you need to define:

• What outcome are we trying to achieve?

• How will we measure success?

• What constraints or boundaries must the agent respect?

• What does failure look like, and how should it be handled?

Consider a simple example: deploying an agent to “reduce support backlog.” Without further clarity, the agent might prioritize speed over quality, closing tickets prematurely or providing shallow responses that create more work later. Technically, it’s reducing backlog. Practically, it’s creating chaos.

This is no different from hiring a new team member and saying, “Just improve things.” You might get initiative, but not necessarily alignment.

Good strategy translates into clear, testable objectives. Instead of vague ambitions, you define something like:

• Reduce average ticket resolution time by 20% without lowering customer satisfaction scores

• Automate Tier 1 queries while escalating edge cases with full context

• Maintain auditability for all decisions made by the agent

Specificity creates direction. Another important aspect of strategic thinking is scope control. Not every problem needs an agent, and not every agent should operate everywhere. Over-deployment is a real risk. Just because an agent can do something doesn’t mean it should.

Treat agents like you would people early in their tenure: give them a clearly defined role, a known set of responsibilities, and a safe environment to operate in. Expansion comes later—after trust is earned. In short, strategy is about setting the destination. If you get this wrong, no amount of clever engineering will save you.

2. Leadership: Turning Strategy Into Action (Without Losing the Plot)

Once strategy is clear, the next challenge is execution, and this is where leadership comes in. If strategy answers what and why, leadership answers how.

With human teams, leadership involves communication, prioritization, and translating big-picture goals into achievable steps. With agents, the same principle applies—but instead of conversations and team meetings, you’re working through system design, prompts, workflows, and feedback loops.

Leadership in the context of agents involves:

• Breaking down strategic goals into actionable tasks

• Designing workflows that guide agent behavior

• Defining intermediate objectives and checkpoints

• Creating alignment between multiple agents (and humans)

An agent without leadership is like a highly capable employee with no manager. They might be productive, but not necessarily in the right direction.

Agents tend to perform better when complex objectives are broken into smaller, well-defined steps. Rather than asking an agent to “handle onboarding,” you might design a sequence:

1. Gather user information

2. Validate inputs

3. Generate account setup instructions

4. Flag anomalies for human review

This isn’t micromanagement, it’s structure that enables consistent outcomes.

Leadership also requires defining checkpoints. Strategic goals are often long-term, but agents operate in the short term. You need intermediate measures to ensure progress is on track:

• Daily accuracy thresholds

• Weekly performance metrics

• Error rate limits that trigger intervention

Leadership also involves alignment across systems. In many environments, multiple agents interact with each other and with human teams. If one agent is optimized for speed and another for accuracy, conflicts can emerge quickly.

This is where orchestration becomes critical. Something needs to ensure that all parts of the system are working toward the same goals, not optimizing themselves into dysfunction. Strong leadership ensures that agents augment people rather than disrupt them.

3. Management: Trust, But Instrument Everything

If strategy sets the direction and leadership drives execution, management answers the ongoing question: Are we actually doing what we said we would do?

With people, management involves performance reviews, KPIs, and the occasional “quick chat.” With agents, it’s all about measurement, monitoring, and feedback. The good news: agents generate data constantly. The risk: without the right visibility, you can still miss the point.

Effective management of agents focuses on three areas:

3.1. Tracking Against Objectives

Management ensures you are progressing toward your defined goals. This means monitoring:

• Task success rates

• Accuracy and error rates

• Time to completion

• Impact on downstream systems or teams

The key is alignment. Your metrics should reflect your strategy.

3.2. Observability and Transparency

Understanding why an agent did something is as important as what it did.

This requires:

• Logs of actions and decisions

• Context for inputs and outputs

• Traceability across workflows

This isn’t just for debugging—it’s essential for trust.

3.3. Feedback and Continuous Improvement

Agents, like people, improve through feedback.

This includes:

• Automated evaluation systems

• Human review of edge cases

• Adjustments to workflows and constraints

Feedback loops need to be fast and actionable. Otherwise, small issues become systemic.

At this stage, the focus shifts from performance to behaviour. Are agents following intended processes? Are they respecting constraints? Are they escalating when they should?

Why the Analogy Holds

Agents are not people. But they are autonomous actors within a system. And once you have autonomy, you need:

• Clear goals (strategy)

• Structured execution (leadership)

• Ongoing oversight (management)

The difference is that agents amplify outcomes. A well-managed agent scales productivity. A poorly managed one scales mistakes just as quickly.

The Future Is Managed, Not Just Built

There’s a temptation to treat agents as purely technical artifacts, things to build, deploy, and optimize. But that view is incomplete.

Agents are part of your operational fabric. They influence outcomes, interact with people, and shape how work gets done.

Managing them effectively isn’t just an engineering challenge, it’s a leadership one.

The organizations that will succeed won’t necessarily be the ones with the most advanced agents. They’ll be the ones that know exactly what those agents should achieve, how they should operate, and how to ensure they stay aligned over time.

In other words, they manage agents the same way they manage people: with clarity, structure, and accountability. Minus the need for annual leave.

We would love to talk to you more about your current documentation validation process and how our award-winning FATCA and CRS Validation platform may add value to your organisation.

For more information on how our fully automated FATCA and CRS Validation platform can add value to your business, get in touch or request a demo to see it in action.