Your next team member might be an AI agent
22 August 2026 | 6 minute read
AI is moving from helping people complete tasks to systems that can plan and act. That changes more than productivity. It changes where decisions are made, where accountability sits and what leaders need to lead.
For the past few years, most of us have been getting to know AI as a clever assistant: write this email. Summarise that meeting. Compare these documents. Give me ten ideas on x. Make this sentence sound less like it came from Legal. Useful, but still recognisable as a tool; you ask and it responds.
The next phase most people and organisations are rapidly moving to is using AI to take action rather than merely offer assistance, and for leaders this phase looks very different. In his TED interview The AI Revolution is Underhyped, former Google CEO Eric Schmidt talks about the arrival of non-human intelligence and systems capable of taking on increasingly complex work. His view of where this eventually heads is significant: computers running far more of the business processes humans currently manage. Whether his timeline proves right or not, the leadership question has already arrived.
What happens when AI doesn’t just help your team do the work, but becomes part of how the work gets done?
That’s where agentic AI gets interesting. An AI agent is not simply waiting for you to type the next prompt. It can be given an objective, break that objective into steps, use tools, act on information and adjust what it does based on what happens next. It is a colleague (albeit a non human one) with access to systems, enormous processing capacity and absolutely no instinct for when Betty from Finance is about to lose it.
And before anyone gets distracted by whether we should call technology a “colleague”, that isn’t really my point. The organisational question is what changes when capability that used to sit entirely inside human roles starts sitting partly inside technology, because we aren’t simply automating tasks anymore, we’re starting to distribute action and decision-making differently.
From tools to participants in the work
Imagine an AI procurement agent that identifies potential suppliers, compares terms, prepares recommendations and progresses parts of a negotiation. Or a customer service agent that responds, adapts and resolves issues without every interaction passing through a person. Or an AI planning agent that runs hundreds of scenarios. Some of this exists now in relatively contained forms. Much more is being developed.
The exact capability will change quickly, but I argue the leadership questions underneath it won’t.
Who decides what the agent is allowed to do?
Where must it stop?
What happens when its recommendation is technically efficient but organisationally stupid?
Who notices?
Who intervenes?
And most importantly...who owns the outcome?
That last question is fundamentally important because there’s a temptation in every technology shift to confuse automation with the transfer of accountability, but they are not the same thing. You can automate an action, you can automate analysis and you can increasingly automate parts of a decision process. But the organisation still needs a human answer to, why did we choose to let the system do that?
AI can take on execution. It cannot take away accountability.
This is where I think executive conversations about AI need to mature fairly quickly. We’ve spent a lot of time asking: what can AI do?
It’s an understandable question, but it’s becoming the less interesting one. The better leadership questions are:
What should AI be allowed to do?
Under what conditions?
Where does human judgement still matter?
What evidence would make us intervene?
And who remains accountable when the system acts?
Those are questions about governance, judgement, risk, organisational design and leadership, not tech and systems.
Which is why delegating the entire AI conversation to IT is inadvisable. Your technology team is there to help you understand capability but they cannot decide the organisation’s appetite for consequence. That remains leadership work.
Six things leaders need to get clearer about now
You don’t need to become an AI engineer, you do need to become much clearer about the conditions under which increasingly autonomous technology operates inside your organisation. Here are five places I’d start.
1. Decide where autonomy starts and stops
Most leaders understand delegation between humans becuase you agree on the outcome, clarify authority and decide when someone needs to come back to you. Agentic systems need the same thinking, only more deliberately.
What can the system recommend?
What can it execute?
What requires human approval?
What must never be delegated?
And what triggers escalation?
If nobody has clearly answered those questions, you haven’t created true autonomy or a highly efficient operating system. You will get decisions at speed but they may be incorrect and not fit the unique context of your business.
2. Make accountability painfully clear
Imagine an AI agent makes a poor decision that affects a customer, an employee or a significant commercial outcome. Who owns it? The vendor, technology or the person who configured the agent? Or is it the executive whose function deployed it? If the answer is vague when something goes wrong, your governance was vague before something went wrong. AI makes sloppy accountability easier to expose. Leaders need to decide where ownership sits before the failure.
3. Demand enough transparency to challenge the outcome
Again, you do not need to understand every technical detail behind an AI -driven system, but “the AI said so” cannot be used as an excuse either. If a system influences material decisions, the right person in your organisation needs to understand the basis well enough to question the output.
What information was it using?
What might it have missed?
What assumptions are embedded in the process?
Where might bias or error appear?
When should a human override it?
The point is you need enough quality information to maintain the ability to exercise judgement.
4. Get much better at scenario thinking
If AI dramatically increases the speed your organisation can analyse information, model possibilities and execute work, the traditional planning rhythm starts to look out of sync becuase your rigid plans become stale faster. Get more comfortable holding a direction while continually reading new information.
What are we seeing?
What has changed?
What assumption no longer holds?
What would cause us to change course?
Your leadership advantage will be as the person who can distinguish a useful signal from noise quickly, without lurching towards every shiny new possibility.
5. Decide what deserves to remain human
This might be the most important conversation of all. The question shouldn’t only be: can we automate this? But should we automate it? And what would we lose if we did?
Some work has value beyond efficiency: a difficult performance conversation, understanding why an employee’s resistance is actually telling you something useful, building trust after a difficult decision, judging when the technically correct answer is the wrong human one and creating meaning around change.
Those moments can look like big time investments but they are human moments and where your leadership is most needed.
As AI takes more administrative, analytical and transactional work, you have a unique opportunity to spend more time on judgement, relationships, thinking and conversation (spoiler: those were supposed to be your priorities already). Use the time well.
6. Design the handoffs between humans and agents
The risk isn’t only that an AI agent makes a bad decision, it’s that nobody is quite sure where the agent’s work ends and the human’s begins. As agents become part of everyday workflows, you will need to design the relationship between human and machine much more deliberately.
Who sets the intent?
What can the agent do independently?
Where does human judgement need to stay close?
When does it need to hand work back to a person?
Who owns the outcome when the work has moved between both?
What information does the human need to make sense of what the agent has done?
And when something unusual happens, who gets involved?
Poorly designed handoffs create very familiar organisational problems: duplication, missed assumptions, fuzzy accountability and people spending their time checking the system instead of benefiting from it.
The goal isn’t to have humans hovering over AI and nor is it to remove humans from the process wherever possible; it’s to work out where each is genuinely better. Let the agent handle speed, scale, pattern recognition and repetitive execution and keep humans close to context, judgement, relationships, trade-offs and an understanding of consequences. And then make the handoff between the two explicit. Because once humans and agents are working inside the same workflow, ambiguity can be very expensive.
Your org chart is starting to change
It’s a given that roles will change significantly, some work will disappear and new work will emerge; but resist reducing this to a headcount conversation. The bigger shift is in how capability is assembled inside your organisation.
For most of our working lives, the model was fairly simple: work was allocated to people, and technology helped them do it, and now that boundary is becoming less tidy. More work will be carried out across combinations of people, software, algorithms and autonomous agents. And when that happens, the shape of the organisation changes even if the boxes on the org chart stay exactly where they are.
Teams may need fewer people doing certain tasks, but more people exercising judgement around them. Managers may spend less time overseeing activity and more time setting parameters, interpreting exceptions and deciding where human intervention matters. Some roles may become broader. Others more specialised.
And accountability will need to be even more focused on outcomes and value rather than task completion
What work belongs where now?
This is where your leadership is about to get more demanding. The more autonomous the technology becomes, the more explicit you will need to be about authority, accountability, risk and judgement.
The real leadership question
Schmidt has described increasingly capable AI as one of the most consequential technological shifts humanity has faced and he may prove right about the scale.
The question you should be asking yourself, if you aren’t already is: what do I need to change about how I lead now execution is no longer entirely human? That question takes us well beyond productivity; it changes where knowledge is created, how decisions move through an organisation, what managers are there to manage and where responsibility ultimately sits.
One of the biggest mistakes you can be making now is focusing too heavily on what AI can take over and failing to ask what deserves more human attention as a result.
What requires better judgement?
Where do we need more restraint?
What conversations should leaders finally have time to have?
If intelligence becomes abundant, what becomes scarce?
AI may change who, or what, does the work but as the leader you still have to decide what good work looks like, where the boundaries sit and what you are ultimately prepared to stand behind.
Key takeaways
Agentic AI changes leadership because AI can increasingly participate in the execution of work rather than simply support individual tasks.
AI agents can be given objectives and increasingly plan, act and adapt across multi-step workflows.
Leaders need explicit boundaries around what AI systems can recommend, execute and decide.
Accountability does not disappear when an action is automated. Organisations still need clear human ownership of outcomes.
Leaders need enough transparency to question AI-influenced decisions rather than accepting outputs because “the system said so”.
Faster analysis makes scenario thinking and judgement more important, particularly when conditions and assumptions change quickly.
Not every task that can be automated necessarily should be. Leaders need to decide where human judgement, trust and relationships create value.
Design clear handoffs between machines and humans to avoid costly ambiguity.
As technology becomes more autonomous, leadership needs to become more explicit, not less.
Leading when the answer isn’t obvious
My work with senior leaders increasingly sits in this territory: making sound decisions when the information is incomplete, clarifying accountability as work changes and leading complex transformation without losing trust or momentum.
AI may accelerate the work, but it doesn’t remove the need for human judgement. If anything, it raises the standard.
Explore Executive Coaching & Leadership Consulting →Leading well through complex uncertainty
Frequently Asked Questions
What is agentic AI?
Agentic AI refers to AI systems designed to pursue objectives across multiple steps rather than responding to a single prompt. Depending on their design and permissions, AI agents can plan tasks, use tools, take actions, assess outcomes and adjust their next steps with varying levels of human involvement.
How is an AI agent different from generative AI such as ChatGPT?
Generative AI is commonly used to create or analyse content in response to a user request. AI agents can go further by working towards an objective across a sequence of actions, potentially interacting with other software or systems along the way. The important distinction for leaders is the movement from generating an answer towards taking action.
What does agentic AI mean for executive leadership?
Agentic AI introduces questions about authority, accountability, governance and judgement. Executives need to decide what systems can do autonomously, where human approval remains necessary, how AI-driven actions can be questioned and who is accountable for the resulting outcomes.
Can accountability be delegated to an AI system?
No. An organisation may delegate tasks or automate actions, but leadership accountability still needs to sit with people. Leaders need clear ownership of decisions about how AI is deployed, what authority systems receive and how risks or failures are handled.
What leadership capabilities become more important as AI becomes more autonomous?
Judgement, scenario thinking, risk discernment, ethical decision-making, communication and the ability to build trust become increasingly important. Leaders also need to distinguish between work that benefits from automation and work where human context, relationships or judgement remain essential.
Should leaders understand how AI systems work?
Leaders do not need to become technical specialists, but they need sufficient understanding to question assumptions, challenge outputs, recognise limitations and make informed decisions about where AI systems should and should not operate.
Keep exploring
Eric Schmidt, The AI Revolution is Underhyped, TED2025
AI is changing what organisations can automate, but many of the harder questions remain thoroughly human. Judgement, accountability and the ability to lead when the answer is not obvious become more important as execution speeds up.
Why Competent Leaders Accidentally Create Decision Bottlenecks
What one GM losing 11 hours a week to approvals reveals about trust, judgement and the unintended consequences of leadership behaviour. Read about decision bottlenecks →
When Compassion Costs Honesty
Why good intentions can sometimes make leadership harder, particularly when judgement and accountability collide. Read about compassion and accountability →
You’re Paying For It. You Just Don’t Know How Much Yet
How seemingly small organisational friction quietly consumes leadership capacity and slows execution. Read about leadership friction →
About Louise
Louise Zawada is an executive coach, change strategist and leadership mentor based in Perth, Western Australia.
She works with senior leaders and executive teams navigating complex organisational change, helping them close the gap between strategy and execution by strengthening executive judgement, reducing leadership friction and improving the quality of conversations that drive performance.
Her work spans mining and resources, government, infrastructure and corporate organisations, where she coaches leaders to make better decisions under pressure, build trust through uncertainty and lead change with greater confidence and clarity.
Louise is the creator of the Leadership Friction framework and writes regularly on executive judgement, organisational legibility and the behavioural evidence that determines whether strategy becomes action.
If you're leading significant change and need a trusted thinking partner, connect with Louise or book a conversation.