When it comes to the work that can be carried out by AI agents, we often focus on repetitive and routine tasks. The types of tasks that machines can get really, really good at if they perform them again and again.
Examples include writing and scheduling social media posts, solving routine customer service issues, or labelling and categorizing unstructured data.
In case you’re not up to speed, agents can be thought of as generative AI chatbots like ChatGPT but with virtual arms, legs and the ability to take action on our behalf.
A slightly more technical definition is that they are large language models (like ChatGPT’s GPT-5 engine) capable of connecting with other tools, including web browsers, to carry out more complex and sophisticated tasks, with minimal human intervention.
But it would be wrong to think that agentic AI is only useful for low-level, tactical decision-making on the shop floor; What time should you send an email newsletter? Or which piece of machinery needs maintenance work?
In fact, it’s increasingly being used by business leadership to make strategic or management decisions. One McKinsey study recently found that 53 percent of C-suite executives and 44 percent of middle managers use GenAI at work, putting them well ahead of the total overall figure.
So here, I’ll take a look at some of the ways that agents are already making their way into business management functions. I’ll also provide some tips for anyone who wants to make sure they’re prepared for the convergence of agents and business leadership.
AI Leadership?
No one will expect the average CEO to be ready to hand over control of their company to a machine any time soon. But there are many theoretical and a growing number of real-world use cases for agents in leadership and high-level decision-making.
One of the most visible is through integration with the decision support systems often used to augment human decision-making by ensuring we have the information and guidance we need.
For example, luxury goods manufacturer LVMH recently announced its decision to build a platform that will use agents to monitor and surface signals so decision-makers can react more quickly.
Asset managers BlackRock have developed their own agentic platform, called Asimov. Asimov can work through the night, gathering research data in real time, monitoring market activity and scanning company filings, ready to present an actionable report to executives in the morning.
And Citi Group has talked about its plans to divert some of its $12 billion tech budget into agentic AI, which could be used to connect different business AI initiatives, improving strategic oversight.
Enterprise software vendors like Microsoft and Salesforce are increasingly building agentic functionality into their tools, to assist decision-makers with strategic advice, or deploying their own agents.
This means we’re likely to see agents being used in exciting new ways as adoption grows among leadership in different industries.
For example, in healthcare, agents could be used to monitor and oversee an increasingly sprawling connected health infrastructure. This helps leaders understand what’s generating value, what’s being underused, and what’s just a waste of resources.
And in manufacturing, agents can track production operations, supply chain effectiveness and energy efficiency, creating opportunities to make strategic plays.
While agents aren’t yet taking the wheel on business decision-making, they’re increasingly in the passenger seat, ready to help us navigate or warn about trouble ahead.
So if you’re a leader or a decision-maker, what steps should you be taking to make sure you aren’t left behind?
What Leaders And Managers Should Do Now
Business decision-makers who’ve already had experience with deploying generative AI will be at an advantage here. But agentic AI requires a different approach, more suited to the long-term strategic planning tasks agents are built for.
Agents are more big-picture and goal-oriented, and your thinking should be too. While ChatGPT can draft you an email, an agentic platform like Manus or Operator can create an email marketing strategy and build the infrastructure to deliver it.
At least, in theory, it can, but it has to be noted that early adopters of those tools frequently report that they don’t always get it right just yet.
From a general usage point of view, it’s important to remember that AI agents are not simply evolved AI chatbots. They’re not really designed to chat or knock ideas back and forward about the best way to create something, which is something ChatGPT is good at.
Agentic AI can be expensive because the “thinking” chews through a lot of tokens. So it’s best to go in with a clear idea of what you want and instruct the agent explicitly on how you would like it to achieve it.
One of the great strengths of agents is that they can use digital tools just like we can, from using a web browser to shop online to controlling complex industrial machinery. Preparation for this might involve ensuring that proprietary and legacy systems can be persuaded to play nicely with agents. Or making a commitment towards open-source technology, where standardization is common and popular tools are likely to be quickly patched to work with an agentic ecosystem.
Next tip: to fulfil their full potential, agents will need access to fast streams of high-value, timely data via well-documented APIs. Organizations that don’t have their house in order here are likely to miss opportunities as they attempt to catch up.
And, absolutely critically, leaders must take steps to ensure they understand the principles of secure, responsible and ethical AI use. As we take steps to give AI the power to make more decisions on our behalf, the need for guardrails, oversight and accountability is more important than ever.
Autonomous Organizations And The Future Of Humans In Leadership
Will we get to a point where machines are so good at predicting the right course of action that humans just aren’t needed? Could we even become a liability?
Well, maybe one day. It’s pretty hard to know where this technology will be in two or three years, let alone ten years.
But one thing we do know is that human brains are still immensely more complex than the best artificial brains. That additional complexity allows us to understand nuanced situations, navigate complex interpersonal relationships and make complex, long-term, strategic plans in ways that computers probably won’t be able to do for some time.
This means that human leaders probably aren’t ready to delegate their powers entirely just yet. Instead, they will look to agents to help them become better leaders and more effective decision-makers.
