5 ways to be great AI agent manager, according to business leaders

akinbostanci/Getty

Microsoft recently said that maximizing AI and agents' performance involves learning about management concerns, including delegating, iterating, prompting, and refining the technology.

The tech giant's research suggested that the requirement for someone to manage these teams of AIs has led to the evolution of a new role, the agent boss, who is responsible for the best performance of these fast-emerging technologies.

Also: 4 questions to ask yourself before betting on AI in your business - and why

So, what might a great agent boss look like? Five business leaders gave us their opinions.

1. Ensure decisions can be trusted

Antony Hausdoerfer, group CIO at auto breakdown specialist The AA, said successful managers will prove agents can deliver value and operate safely.

"I'm very open to AI and want to see how we can use it. Having listened to a lot of conversations around AI, you can apply it to absolutely anything," he said. "However, will that application deliver a real outcome for you, your business, and your customers? Answering that question effectively means you should be focused on how to use agents."

Also: What are AI agents? How to access a team of personalized assistants

When it comes to the safety aspects of using AI, Hausdoerfer told ZDNET that executives must ensure that agentic technologies make good judgments.

"How can you audit those decisions and improve the agent so that it's not creating bias, and is making safe and effective decisions?"

Hausdoerfer referred to his company's operations, suggesting that effective managers will ensure agents can be trusted.

"We're looking after people at the roadside. I need to be sure, before we apply AI to that process, that the technology is ready," he said. "That's a case of being able to audit and review those decisions and say, 'I'm comfortable the AI is making safe decisions.'"

2. Develop a blend of attributes

Bev White, CEO at technology and talent solutions provider Nash Squared, said successful leaders will need to adapt to a new paradigm where AI systems make decisions and take actions autonomously.

"Rather than just responding to prompts, agentic AI will increasingly take the lead and manage whole processes on its own," she said. "This situation raises new questions and requirements for leaders."

White told ZDNET that four attributes will be key. First, adaptability -- successful leaders will understand the unique characteristics of AI systems and how to integrate them into their teams.

Also: 4 ways to turn AI into your business advantage

Second, governance and accountability -- leaders will need to challenge AI assumptions and guide the decisions made by their teams to maintain ethical standards and organizational alignment.

Third, leaders will have a forward-thinking mindset, anticipating the potential impacts of AI on their industry and organization.

"They'll need to 'forward think' about how AI can be leveraged to drive innovation and competitive advantage," said White.

Finally, White said leaders will require empathy, as the human element remains critical to the successful adoption and adaptation of AI.

"Leaders must focus on the well-being of their team members, ensuring that AI is used to augment human capabilities rather than replace them," she said. "Agentic AI will take all of us to new places and new ways of working. Leaders need to lean into this reality, explore what it will mean in practice, and support their teams so that agentic AI helps drive desired outcomes rather than being an end in itself."

3. Embrace constant change

Vivek Bharadwaj, CIO at clothing manufacturer Happy Socks, said effective leaders will be systems thinkers who understand the significance of their role in an age of ever-increasing automation.

"In a brave new world, in which effort is commoditized by these agentic AI systems, the value you provide will be the true differentiator," he said.

Bharadwaj told ZDNET that successful agent leadership will be about professionals moving higher up the value chain to be more in tune with customer requirements.

Also: You've heard about AI killing jobs, but here are 15 news ones AI could create

He said effective managers in this new world will also be comfortable embracing change.

"What I mean by that is whoever is going to be successful needs to be OK that things keep changing," he said. "They need to be the agents of change themselves, rather than being terrified of it. If a train is coming towards you, you better be the person who knows how to drive an engine."

4. Establish the right balance

Rom Kosla, CIO at Hewlett Packard Enterprise (HPE), said managing agents will be one of the most important concerns of our time.

He referred to Nvidia CEO Jensen Huang's comments at CES in January, in which he said that IT will become the HR of AI agents in the future.

Huang suggested IT teams will be tasked with onboarding agents and ensuring they're kept in line, similarly to how HR teams manage employees.

In short, IT professionals will soon be supervising fleets of digital workers -- and that's a suggestion that resonated with Kosla: "There's going to be an interesting concept of, 'How do we manage agents?'"

Also: 5 ways you can plug the widening AI skills gap at your business

He told ZDNET that the relationship between experienced human coders and knowledge-rich agents will develop over time.

"Does everyone know every line of code? You have professionals with domain knowledge, and they become experts. Agents will also play in that space," he said. "I think that question of successful management of agents is something we'll be pondering over time. I also think it's one of those open-ended questions that requires a lot of coffee."

5. Create a federated approach

Manish Jethwa, CTO at Ordnance Survey (OS), the UK's national mapping service, said the management of agents is unlikely to fall under the purview of an individual or team.

"It's going to have to be some type of federated approach," he said. "I don't think there's going to be a single function that's responsible for all the agents. I think the agents will end up being experts within each organization."

Jethwa told ZDNET he expects OS to build agents in the longer term that help the organization access and manage data.

Also: AI won't take your job, but this definitely will

He anticipated a situation where questions are answered by agentic experts in line-of-business functions that draw information from a repository.

"That feels like the right approach, one in which each organization will own its expertise, and they'll draw on a normalization layer at that level," he said. "Because the agents must be tuned to the interface they're connecting to, it's right for the maintenance of those agentic AIs to live close to whoever's building out those capabilities. So, if we're building out APIs that allow access to our data, the agents should live as close to those teams as possible."