Judging by a slew of recent corporate announcements, your next “co-worker” might be an artificial intelligence agent – doing the work of an assistant, job scheduler, morning debriefer, learning coach and more.
JPMorgan Chase, the largest U.S. bank, describes a clear vision for a new world of omnipresent AI agents: “Every employee will have their own personalized AI assistant; every process is powered by AI agents, and every client experience has an AI concierge.”
In brick-and-mortar retail, Walmart is already implementing its vision around agents, which involves support of customers, in-store employees and other business areas, with supervisor agents assigning tasks to subagents much like managers oversee employees.
What these and many other large organizations realize is that agents don’t just answer questions, like an AI-powered search engine or simple chatbot. They complete real work by planning tasks, taking actions and checking results to achieve a goal.
But there’s a problem. Companies in industries ranging from finance and tech to logistics and legal are rapidly embracing the promise of AI agents. But the flesh-and-blood workers they’re meant to assist – and sometimes replace – are struggling to adapt, hurting morale and productivity in the process.
The result is a growing climate of fear about AI job insecurity. FOBO – fear of becoming obsolete – is now a thing. A recent survey by consultancy KPMG found that 52% of workers report they are concerned that AI could eventually take their jobs. And some are fighting back. In another survey, nearly one-third said they are actively sabotaging their company’s AI strategy.
To make matters worse, some of these AI agents are going rogue, deleting data or executing other unintended actions.
My research on AI and agent capabilities, value and risk, as well as emerging studies of the cognitive implications of AI, the future of work and the role of AI in workplace inequality, suggest two key lessons for anyone navigating this new AI agent reality:
First, learn how the agents you’re working with operate, including what they do well, where they fall short and how to catch mistakes.
Second, lean into your fundamentally human strengths. These are things agents can’t replicate. Doing so can also help you sustain your own health and well-being.
Rise of the AI agents
AI agents began entering the workforce in 2025 – mainly in tech, finance and customer service – as the next stage of the generative AI revolution. But in 2026, the AI-powered automators are increasingly being deployed in other areas, such as legal and compliance, supply chain management, research and development, healthcare services and retail.
One example is global transportation giant FedEx, which is planning an entire AI agent workforce for its logistic network. The company plans to create “manager agents,” “audit agents” and “worker agents” to create a trail of accountability for their actions, according to The Wall Street Journal.
The theme of agents working together is echoed by North American food service company Gordon Food Service, which is using cross-team agents to reshape its product sourcing strategy.
As companies rapidly adopt the latest AI systems, agents are taking on increasingly diverse roles in the workplace that leverage their ability to do autonomous work.
It’s all driven by economics, with 88% of early corporate adopters reporting a return on investment on at least one use of an AI agent, according to a Google survey of senior business leaders.
Retail giant Amazon says customers who engage with its Rufus AI agent while shopping are 60% more likely to make a purchase compared with those who don’t get help from Rufus. Overall, Rufus is expected to generate over $10 billion in additional annual sales for Amazon, compared with a baseline without Rufus.
Anticipating gains from agents, global consultancy McKinsey already has 25,000 agents doing various tasks. It plans to have as many AI agents as human workers by 2027.
And new model releases, such as Anthropic’s Mythos, will expand what is possible and likely accelerate agent deployment.
Agents still can’t do it alone
A 2023 study I conducted found that AI has the capacity to simulate human capabilities such as cognition, decision-making, creativity and collaboration with people and other agents. The simulation is imperfect, however, so agents need support.
For example, agents are resourceful and relentless. They try repeatedly until they get results, without the loss of motivation humans sometimes experience. But they can also behave unpredictably, even taking harmful actions, such as deleting emails or conducting a smear campaign. And research illustrates that agents can be easily tricked into bad behavior, such as overcorrecting when told not to do something, being persuaded by human appeals to urgency and simple manipulation tricks.
Agents can be quirky too, using odd emojis in formal business writing or responding cynically when you just want the facts. And unlike humans, agents lack emotion, self-awareness or intent. When they fail, such as by pursuing misaligned goals, it’s not personal any more than an espresso machine can break out of spite.
In short, agents can act in unpredictable ways, and it’s difficult to know when that will happen. Managing this uncertainty is a new task for employees, and it comes at a time when many employers have heightened expectations of productivity gains from adopting AI agents.
The bottom line is that if you understand how your agent behaves, you’ll be able to be more productive working with them, be in a better position to avoid risks and be more valued yourself.
Get to know your agent
So what do you do if your boss suddenly tells you you’ll be working with an AI agent from now on?
Just like you would with a new human colleague, the first thing you should do is get to know it. In essence, you need to learn how to collaborate with agents effectively, how to evaluate their performance, what makes them tick and associated ethical implications.
And then jump right in. Give the agent a task and observe how it responds. Try different approaches and pay attention to its output quality, behavior and style. Focus on three essentials:
-
Clarity of intent: Define exactly what to do in clear instructions, what information your agent needs, the role your agent is performing, limitations on behavior, and what success looks like for the task.
-
Evaluate results against clear criteria.
-
Guide the agent during the task by answering questions as they arise.
Overall, this will help you effectively use and critically evaluate AI agents.
Leaning into your humanness
Your agent deskmate still needs you – and your humanness – to be effective.
There’s a growing understanding that in a world of ubiquitous agents, certain skills, such as analyzing information, are becoming less important, while other skills, such as interpersonal and communication skills, are increasing in importance. Think about it: No AI agent can read a room like a human can and pick up on the vibe.
But making the switch will not be easy. It will require psychological work.
Researcher Tomas Chamorro-Premuzic suggests connecting with others as only humans can do, and unlocking your curiosity, while your agent handles the drudgery. In other words, focus on qualities that AI agents don’t have, such as the ability to pick up on nonverbal cues, deliver a pitch with a human touch, manage conflict and build relationships. These skills are the glue that holds human teams together.
AI agents are likely to become a significant part of the workplace. But how and how fast that will happen is unknown.
To make the best of it, learn how to work effectively with agents and embrace your own humanness. This way, you’ll be in a better position to make informed decisions about how to interact with humans and agents alike.
