Imagine you’re working with an AI agent on a customer proposal. It retrieves the figures you need, updates the presentation and prepares the follow-up. But when you review the final offer, you discover that someone else on your team asked the agent to change the terms. And it did.

Who gave it permission to do that? The colleague who gave the instruction? You, because you’re responsible for the proposal? The manager who owns the customer relationship? Or the person whose account connects the agent to the company’s systems?

Organizations have spent decades developing processes to answer questions like these for their human workforce. Hiring establishes who can join and at what cost. Onboarding provides context. Roles define responsibilities. Managers set expectations, review performance and decide when someone is ready for more responsibility. But what happens when the work is increasingly performed by people and AI agents together?

Cisco’s recent WebexOne announcements illustrate why this question is becoming urgent. The company described agents that employees could invite into team conversations to retrieve information and act across connected applications. As agents become accessible through the tools people already use, they increasingly participate in everyday work. Technology controls may determine what they can access, but organizations still need to decide what they are authorized to do and who is responsible.

Applying familiar workforce management processes to AI agents may seem like a logical next step. But the idea has already generated controversy. In July 2024, Lattice announced plans to introduce digital worker employee records , bringing AI agents into organizational charts with goals and managers. Following a backlash , the company abandoned the plan three days later.

The reaction is understandable. Giving agents employee records can appear to blur the distinction between people and technology, particularly amid concerns about AI replacing human workers. But in a conversation on The Future of Less Work podcast, Lattice CEO Sarah Franklin made a practical observation: “There’s very real known processes we have for bringing entities into an organization. They are already documented, already in place.”

Those processes were developed to manage people. Agents are fundamentally different from human employees, yet they increasingly contribute to the same work and outcomes. For CHROs and business leaders, the question is which of these established processes need to evolve to manage the combined workforce, creating an AI agent lifecycle that spans everything from deciding when an agent is needed to managing its work, authority and eventual exit.

Rethink Hiring As AI Agent Approval

A manager cannot hire unlimited people simply because there is useful work to do. Someone must approve the need, budget and responsibility. Creating an AI agent can feel much more casual.

Consider two teams independently creating agents to follow up with the same customers. Each might appear productive while duplicating work, generating costs and sending inconsistent messages.

The equivalent of hiring approval should establish whether a new agent is needed, what it will cost and which human owns the result. Workforce planning must also consider when to add people, agents or a combination of both.

Rethink AI Agent Onboarding As Context And Boundaries

A new salesperson gets more than access to a customer database. Onboarding establishes the company’s products, commitments, standards and expectations. An agent communicating with customers needs that context too.

An agent arranging replacement products, for example, must know which customers qualify, what it can promise and when an exception requires human judgment. Onboarding should establish who provides that context, who tests the agent’s decisions and who is responsible for correcting its behavior. Connecting an agent to information is only part of preparing it to work.

Rethink Roles And Reporting For Human-AI Teams

An organizational chart clarifies responsibilities and decision-making authority. A combined workforce needs similar clarity about the relationships between people, agents and work.

The person who builds an agent, the employee who uses it and the manager accountable for its results may be different people. Consider an agent receiving instructions from multiple employees while operating under policies owned by another department. When those instructions conflict, whose direction takes precedence?

This also raises questions about how decision rights and authority need to evolve as AI expands employee capabilities . An agent may be capable of carrying out a request without being authorized to do so. Drafting a customer offer, approving its terms and sending it create different responsibilities. Organizations need to define who can instruct agents, who can authorize their actions and which humans remain accountable.

Rethink Performance Management For Human-AI Teams

How do you measure the performance of a team made up of people and AI agents? Measuring them separately can miss what happens between them: an employee correcting an answer, an agent escalating too late or work returning for revision.

Franklin offered an example of an agent instructed to close customer-service cases within two minutes. It could meet that target simply by closing cases, regardless of whether customers’ problems were resolved. A human employee would likely understand the intended outcome. With an agent, expectations may need to be made explicit.

Human-AI team performance management must look beyond individual activity to evaluate the business results people and agents produce together. It should reveal where work breaks down, when human intervention adds value and whether feedback improves subsequent outcomes.

Rethink Transfers As AI Agent Ownership

When an employee changes roles, what happens to the agents they use?

A personal assistant organized around the employee’s working habits might move with them. An agent managing the old team’s customers should remain with that responsibility.

Organizations therefore need to distinguish between agents attached to individuals , agents attached to roles and agents supporting shared processes. Transfers should trigger decisions about ownership, outstanding work and permissions. Otherwise, agents may continue accessing information or acting on behalf of someone who no longer holds the relevant responsibility

Managing A Combined Human And AI Workforce

The question of whether people and agents belong in the same workforce management system is only part of a much larger organizational challenge.

For decades, HR has managed people while IT has managed technology. That separation made sense when technology was primarily a tool people used to do their jobs. But AI agents are increasingly performing work alongside employees, making decisions, interacting with customers and contributing to business outcomes. The boundaries between managing people and managing technology are becoming less clear.

This has significant implications for workforce planning. Organizations need to understand their combined human and AI capacity when deciding what capabilities they need, where to invest and how to organize work. They need to see what people and agents can do separately, what they accomplish together and where additional capacity or capabilities are needed.

It also requires HR, IT and business leaders to rethink how their processes connect. Decisions about deploying agents, assigning responsibilities, managing performance and allocating resources can no longer happen independently across functions. Organizations need a shared view of the entire workforce to make those decisions effectively.

As the workforce becomes a combination of people and AI agents, organizations need management processes that reflect that reality. Otherwise, they risk making decisions about workforce capacity, investment and performance without understanding the full resources available to them.