Artificial intelligence is often introduced as a technology decision: choose a platform, run a pilot, and encourage adoption. Yet the technology is rarely the hardest part. The more consequential work is deciding where it belongs in the institution, what should change around it and how people will remain accountable for the result.
That calls for a work-led starting point. Before selecting a tool, leaders need a clear view of the tasks, handoffs and judgements that make up a process. Some activities are repetitive and rules-based. Others depend on context, trust, professional experience or careful interpretation. Knowing the difference helps an organization identify useful opportunities without mistaking novelty for value.
Start with the task and its purpose
Map the work as it is actually done, not only as it appears in a procedure manual. Where does time accumulate? Which steps create delay or rework? What information is difficult to locate? Which decisions are routinely escalated, and why? These questions uncover friction that automation may help relieve. They also reveal problems that are better addressed through clearer authority, better data or a simpler process.
A task is not a good candidate merely because it can be automated. Its purpose, risk and connection to other work matter. A draft generated quickly may still need substantial review. A recommendation may be useful as an input, while the decision itself remains with a qualified person. Leaders should be precise about where assistance ends and accountability begins.
Redesign roles with people in view
When part of a role changes, the rest of the role changes too. Employees may spend less time assembling information and more time checking it, explaining options or responding to exceptions. Managers need to set expectations for that work. Teams need access to relevant tools and guidance. Existing performance measures may need to shift from activity completed to quality, judgement and service.
People should be involved before a new way of working is fixed. Their experience can reveal practical constraints, unintended consequences and opportunities that are invisible from a distance. Involving staff is not a substitute for leadership accountability. It is a way to make decisions better informed and implementation more credible.
Build confidence through disciplined learning
Early use should be bounded, observable and connected to a real problem. Define what a useful outcome looks like, what information may be used, who checks the output and when the process must stop. Review actual examples, including errors and near misses. That record helps the organization decide whether to refine, expand or discontinue an application.
Capability is broader than prompt writing or product training. It includes data stewardship, critical evaluation, process design, risk awareness and the confidence to challenge an answer that seems plausible but is wrong. Senior leaders need enough understanding to ask informed questions and make trade-offs visible.
Make adoption an institutional choice
Technology can accelerate work, but it does not choose priorities on an institution's behalf. Leaders still need to decide which outcomes matter, what standards should hold and how benefits will be shared. They also need to account for the capacity required to maintain tools, review performance and adapt governance as practice develops.
The strongest starting question is therefore not “Where can we use AI?” It is “Which work should become better, for whom, and what must remain under human judgement?” Answering that first creates a more useful basis for investment and a more responsible path from experiment to everyday practice.
