The company bought artificial intelligence. Now accounting would like to know what it does.

This was probably inevitable. After the demonstrations, the subscriptions, the executive announcements, and the solemn declarations that everything had changed, someone was eventually going to ask whether the machinery had improved anything.

Anthropic has introduced a set of tools designed to help companies answer that question. Its new enterprise reports analyze samples of teams' Claude conversations, identify the work being done, connect that work with spending, and flag recurring problems. Employee surveys add another perspective, while shared team budgets help companies manage costs.

There is something almost comforting about this development. Even artificial intelligence eventually has to explain an expense report.

For people looking for work, however, the more useful story is what this suggests about the next conversation with an employer.

"I know how to use AI" may get you through the opening question. You should have something ready for the follow-up.

What have you accomplished with it?

A hiring manager can do considerably more with that answer.

Consider two candidates. One lists ChatGPT, Claude, and several other tools under "Technical Skills." The other explains how she used AI to organize customer feedback, checked its categories against the original comments, and helped her team identify a recurring service problem.

The second candidate has given the interviewer something to investigate. What was the problem? How did she verify the analysis? What happened afterward?

Now they are talking about work.

That is where you want the conversation to go.


A great deal of AI discussion remains trapped at the level of demonstrations. Look how quickly it wrote this. Look how much it produced. Look, it made a presentation while we were getting coffee.

Fine. Was the presentation accurate? Did anyone need it? Did it help somebody make a decision?

An employer has reason to care about the time between receiving an assignment and delivering something useful. Generating a first draft in thirty seconds is an interesting intermediate event. If correcting it takes three hours, the stopwatch deserves another look.

This creates an opportunity for candidates who can explain the entire process, including the parts that required human judgment.

Suppose you used AI to help prepare a monthly report. You supplied the source material, developed a repeatable set of instructions, checked the figures, corrected misleading summaries, and delivered the report sooner. That is a useful example.

Perhaps you have reliable records showing that preparation fell from four hours to two. Use them. If you never measured it, describe the improvement honestly. A specific account of what changed will serve you better than a magnificent percentage you invented while updating your résumé.


The same principle applies across occupations. A salesperson might explain how AI helped prepare account research before a call. An operations coordinator might show how it helped turn scattered instructions into a procedure people could follow. A marketing candidate might demonstrate how it helped organize competing messages for testing.

In each case, the tool belongs inside a story about a task, a decision, and a result.

If you are unemployed, you can still build such a story.

Choose a small project related to the work you want. Use public information or material you have permission to use. Prepare a competitor comparison, improve a confusing set of instructions, or analyze a modest collection of customer reviews. Keep the scope manageable enough that you can check the work thoroughly.

Then preserve the evidence: your source material, the finished product, and a short explanation of your process.

You should be able to explain where AI helped, where it went wrong, and what you changed. An example of catching a mistake may tell an employer more about your judgment than a flawless-looking document with an unexplained origin.


There is also a management issue worth watching. Anthropic says its reports are intended to inform adoption and investment, and are not for evaluating individual performance or making employment decisions. That distinction is explicit in its announcement.

Companies will need discipline as they interpret this kind of information. A busy AI account can reflect productive work, repeated failed attempts, or an employee generating material that everyone else now has to read.

Counting activity is easier than assessing its value. Anyone who has endured a meeting about the number of meetings should recognize the problem.


For job seekers, the practical response is to develop a few examples you can defend. Know what you were trying to accomplish. Explain your choices. Show the finished work when you can, while respecting confidentiality. Be candid about the limits.

You do not need to audition as an AI evangelist. You need to help an employer picture you doing useful work.

Eventually, someone across the desk will ask what all those tools on your résumé helped you accomplish.

Have a good answer ready.