An employee uses AI to prepare a report in minutes. A manager sees the opportunity to shorten the reporting cycle. Another team starts using the output to make decisions.
Then a discrepancy appears.
Who was responsible for checking the analysis? Did the employee understand the assumptions? Had anyone agreed which decisions could rely on the output?
A useful experiment has become part of an organisational process, but the responsibilities around it have remained informal.
This is where leadership engagement becomes essential. As AI becomes part of everyday work, leaders need to make deliberate choices about how that work will operate.
Accountability needs to develop alongside adoption
Australian research published by ADAPT on 29 October highlights the challenge. Among 100 HR leaders surveyed, more than half reported having no defined frameworks or accountabilities for AI agents making decisions that directly affect employees. Only 3% reported clear, defined decision rights. The sample is small, but the finding raises a practical question: how clearly have organisations allocated responsibility for the AI they are introducing?
An organisation can approve a tool without resolving who may use it for particular tasks, what requires review, or who can stop a process when something goes wrong.
These decisions require business leadership. Technology teams can explain capabilities and controls. Leaders responsible for the work must help determine where those capabilities belong, what good performance looks like, and which judgments remain with people.
Start with the work people actually do
Planning becomes more useful when leaders examine a specific workflow.
Consider a finance team using AI to prepare management commentary. Producing the first draft may become faster. Someone still needs to investigate unusual movements, validate explanations and decide what the executive team needs to understand.
The role changes as the balance of effort shifts. Drafting may take less time, while review and interpretation become more significant.
Anthropic’s internal research, published on 16 October, provides an example of this shift. Employees described spending more time directing agents, exercising judgment and coordinating work as parts of implementation became automated. The study reflects an unusually AI-intensive workplace, so it should not be treated as a prediction for every organisation. It does illustrate why leaders need to examine the work that AI creates as well as the work it takes over.
A practical planning discussion should establish:
• Which tasks will change, and which will remain?
• Who will check outputs and handle exceptions?
• What knowledge will people need to exercise sound judgment?
• How will responsibilities and performance expectations change?
• What time will be available to learn and practise?
These questions make the implications of AI visible before new expectations become embedded.
Engagement means staying close to the experience
Executive sponsorship can establish direction and provide resources. Continued leadership involvement helps reveal what happens during use.
Employees may find that an AI tool works well for familiar tasks but struggles with unusual cases. They may spend time correcting outputs, checking sources or maintaining an existing process while confidence develops.
If leaders only ask whether people are using AI, that additional effort can remain hidden.
More useful conversations explore where the tool helps, where it creates uncertainty and what people are doing to make the output dependable. Leaders then need to act on what they hear—adjusting scope, resolving ownership or making room for practice.
Employees also need to know that raising a problem will influence implementation. Otherwise, the pressure to demonstrate progress can make difficulties harder to discuss.
Plan for capability and capacity together
Training can introduce a tool. Applying it responsibly requires opportunities to test it against real work, compare results and learn when to question its answers.
That takes attention.
If normal delivery expectations remain unchanged, learning and checking may simply be added to the existing workload. The organisation can appear to be progressing while the people supporting that progress become stretched.
Leaders therefore need to agree which work can be delayed, simplified or stopped during the transition. They also need to consider how less experienced employees will develop judgment if AI takes over tasks through which that judgment was previously learned.
Make value visible
A shorter drafting time is useful evidence, but it is only part of the picture.
Leaders should also examine accuracy, review effort, rework, decision quality and the effect on employees. Those measures help establish whether AI is improving the whole workflow.
Leadership engagement gives this learning somewhere to go. It connects experience with decisions about roles, resources and expectations.
The planning can begin with one workflow and a small group of people. Agree the intended benefit, define responsibility, provide space to learn and review what happens.
That is how an AI initiative starts to become a workable way of operating.
Leave a Reply