As we enter age of agentic AI everyone becomes a manager
More from this theme
Recent articles
As generative AI evolves into agentic AI, the skills we need to focus on are changing. If the last few years have been dominated by prompting and evaluating outputs, the next phase is likely to be dominated by goal setting and delegation.
Before we go on, let’s start by unpicking what I actually mean by agentic AI.
I’m talking about AI tools that work towards a goal without the user giving them the exact steps to get there. To do that, these tools may have access to your files, a web browser and even other agents.
If generative AI is largely about producing content, then agentic AI is about actions.
This doesn’t make the things we currently focus on with AI literacy less important. We still need a broad understanding of how these systems work, and of ethical and legal concerns.
To this we can add a new set of skills: the ability to clearly set goals and define outcomes, delegate effectively and remain accountable for work we did not do ourselves.
Goal setting in an agentic world
Let’s start by looking at goal setting. Agentic AI shifts the emphasis from specifying how a task should be done to being clear about the outcome you want to achieve.
Are we good at this? I’d argue, often not.
For every great example, there are thousands of weaker ones. Goals that are so vague they are closer to aspirations than outcomes.
“Our goal is to transform lives” is well intentioned, maybe, but unclear about what success would look like, for whom, and over what timeframe.
An AI agent is likely to take a goal at face value. It might ask for clarification if it has been developed to do so, but it may equally just work towards the goal.
So to make the most of agentic AI, we need to get much better at this, much earlier in our careers. We need to build it into our curriculum.
Delegation as a lived skill
The second skill is delegation, and this is one that really hit home. Whilst delegation might be normal for senior staff, it can be quite unrelatable to many of us, especially those just starting out in their careers.
There is, of course, always a risk of anthropomorphising AI and drawing too close a parallel between delegating to people and delegating to code.
But it is still worth reflecting on what good delegation involves and how it relates to AI.
It means making deliberate choices about what not to delegate, being clear about constraints and authority, and knowing how to monitor progress without micromanaging.
It also means recognising when things are drifting off course and being willing to step in when needed. Above all, it means understanding that accountability remains even when tasks are delegated.
So how do we normally learn to delegate? As with goal setting, it is usually from a mix of experience and development over time.
Implications for curriculum and assessment
For both skills, if AI has even half the impact on entry-level jobs that is often predicted, learning by observing more experienced colleagues will not be enough.
We cannot assume that students will ‘pick up’ delegation once they become managers, because in an agentic world, they are managers of systems from day one.
By the time people encounter these expectations in the workplace, it may already be too late.
Alongside critical thinking and digital literacy, goal setting and delegation need to be explicitly woven into our curriculum and assessment frameworks.
If we want our learners to lead in the age of AI, we must teach them not just how to use the tools, but how to take responsibility for the outcomes they produce.
If prompting was the skill of the generative AI era, delegation and accountability will be the defining skills of the agentic era.
