Essay 04
What your people should be doing now
Your team keeps asking what AI means for them, and "do not worry about it" is not an answer. They want to know what to actually do.
The useful answer is specific. It is not "take a course" or "learn to write prompts." It is a handful of habits that make a person more valuable as the work changes, and that also happen to be what makes AI work inside a company.
Here is what to ask of your people, and what to give them so they can do it.
Learn the tools on their own real tasks
General training fades quickly. What sticks is using a tool on the work someone does every week: drafting the reply they write over and over, summarizing the long thread they have to read anyway, pulling details out of the documents they already handle.
Ask each person to pick one task of their own and try it with an assistant for a couple of weeks. Not to prove anything. To learn where it helps and where it does not.
What to give them: access under business terms, a short plain list of what company information can and cannot go into the tool, and time. Learning on top of a full workload does not happen.
Own and check the AI step in their work
When AI takes a step in someone's process, that person still owns the result. AI drafts, people approve. AI sorts, people handle what does not fit. AI summarizes, people decide.
That means checking the output regularly, not only when something looks wrong. The person who does the work knows what right looks like, which makes them the right one to catch what the tool gets wrong.
Ask people to keep a short note of the mistakes they catch. That list is useful to whoever owns the process, and it makes the checking visible instead of invisible.
Lean into judgment
The parts of a job that depend on judgment are where AI is weakest: deciding whether to make an exception, reading a situation that does not fit the pattern, knowing when a question needs a phone call instead of a reply. Encourage people to spend the time they get back there.
Make it explicit. If someone used to spend Monday morning rebuilding a report, ask them what they should do with Monday morning now. Their answer is usually better than yours.
Lean into customer relationships
Customers notice when they reach someone who knows them. The time saved on paperwork can go to the calls, visits, and follow through that keep a customer coming back. Those relationships are hard to automate and easy to lose.
Become the person who handles exceptions
Every process has cases that do not fit: the unusual order, the complaint with history behind it, the request that touches three departments. AI handles the normal cases far better than the odd ones. People who become known for handling exceptions well get harder to replace, not easier.
Care about quality
When output gets faster, quality is what slips. Someone has to notice when the drafts start sounding generic, when a summary missed the one detail that mattered, when a figure does not match the source. Ask people to treat quality as part of the job, and say so out loud when they catch something.
What not to ask
Do not ask everyone to become an AI expert. Do not run a contest for who uses it most. And do not ask people to find ways to automate their own work and then cut the ones who do. That ends the effort faster than any technical problem.
What leaders should do
Go first. Use the tools on your own work and talk about what did and did not help. Make it normal to say the tool got something wrong. In reviews, ask about judgment and quality, not only speed.
On Monday
Ask each manager to sit with their team and have every person name two things: one weekly task they will try with an assistant, and one part of their job that depends on judgment and deserves more of their time. Write both down. Check back in two weeks.