What I Know This Week

Work smarter

A prompt is not a use case

A prompt is a one-off instruction. A use case is a task you repeat, tied to a goal, that runs the same way every time. Most people stall at the prompt stage. They ask a vague question, get a generic answer and decide AI is overhyped. The jump that matters is from trying AI now and then to having two or three routines you actually use.

A workshop I sat in on gave a simple way to find them. Look at your week through four lenses: Creator (first drafts), Analyst (patterns in your numbers), Strategist (a thinking partner that pushes back) and Researcher (pulling sources together). Then put each idea on an impact versus effort grid and start in the high impact, low effort corner, which SixSigma.us calls the low-hanging fruit. Try it with ten tasks you repeat every week, and pick one.

Work smarter

Let AI direct your attention

A daily brief is a scheduled AI task that reads your email, calendar and task tools, then tells you what matters today, what can wait and your first move. Its real value is taking the "where do I start" decision off your plate before the day starts.

The setup advice was refreshingly simple. Connect one easy source first, run the prompt, then fix it by talking to it: half the length, drop the meetings, add a first move. Once it reads right, schedule it. Claude runs scheduled tasks hourly, daily, on weekdays or weekly on paid plans, and standard tasks keep running when your laptop is asleep. A short "about me" file with your goals and priorities makes every brief sharper. I run my own version every morning for AI news.

Money and risk

Agents need an ID badge and a budget

More AI plans now bill by usage instead of by seat, and the reporting is catching up. Claude's Team and Enterprise analytics can export per-user, per-model spend. In a spend analytics session this week, the useful measure was spend per person compared against skill level. High spend from a beginner is worth a look, since it can mean a premium model is doing simple work.

The same week, Google Cloud introduced its Gemini agent, and the governance details are the part worth copying. Every agent gets its own identity with least-privilege permissions. Every action is written to an audit trail attributed to the agent, and a spend cap pauses the agent when a project hits its limit. You don't need Google to adopt the principle. Any AI that acts for your business should have its own login, a spending limit and a log.

Work smarter

AI won't give your week back on its own

In 1865, William Stanley Jevons noticed that more efficient steam engines didn't cut England's coal use. They increased it, because cheaper power found more uses. That's now called the Jevons paradox, and AI is doing the same thing to work.

An eight-month study of about 200 employees, published in HBR in February, found AI made people faster but busier, with more task switching, more checking of AI output and more open tasks. The authors call the research in progress, and it covers one tech company, so treat it as an early signal. It still matches what an engineer said in an interview this week. Keep the build small, and decide what the saved time is for before you automate. If the answer is "more of the same," expect to feel busier, not freer.

Built this week

One link that never changes

I built a /workshops page for my first local workshop series, "AI for the Work You Already Do." Instead of guessing a date and a venue, the page runs a waitlist that asks people whether they want in person or online and what time works. The date gets picked from that data.

The part I like most is the QR code. It's printed on my new business card and points to one permanent page. Workshops come and go on that page, and the card never needs reprinting. If you hand out anything printed, point it at a page you control, not at a single event.

Your move

There are three ways to use this issue. Reach out if you want help, input or a second opinion on any of it. Try one thing this week, even if it's just the ten-task list. Or send this to someone who keeps saying AI hasn't saved them any time.