Most people who want to start with AI make the same mistake.
They go looking for what AI can do. They watch demos, read articles, maybe try a few tools. And then they try to fit their work into what they saw.
That's backwards.
The better approach starts with you — specifically, with what's slow and frustrating about your actual week. I call it the AI wish list, and it's the exercise I now walk through with every person I coach before we build anything.
The Exercise
Here's all it is.
Block 10-15 minutes. Go through your regular tasks — the stuff you do daily, weekly, the recurring grind. For each one, ask a single question: “I wish AI could do this.”
Write down everything that comes up. Don't filter. Don't judge whether it's technically possible. Just make the list.
When I did this exercise with John, an auditor I was coaching, he rattled off three things in about two minutes.
Audit note-taking. Catching commitments that slipped through client calls. Cross-referencing his email history before meetings with clients.
None of those ideas came from a YouTube video. None of them came from reading about what AI can do in theory. They came from John looking at his actual week through a new lens — and noticing where the friction was.
That list became his roadmap.
Why “I Wish” Works
The framing matters.
“What can AI do?” is a product question. It puts the burden on the technology to impress you. You end up chasing demos of impressive but rare use cases — things that look good in a video but don't connect to your actual work.
“I wish AI could do this” is a pain question. It starts with what's already broken or slow in your day. The gap is real. The motivation is real. And when you eventually build or deploy the solution, it actually gets used because it solves something you genuinely care about.
I've seen this with every client I've worked with. The people who try to “use more AI” without a wish list end up dabbling. They try tools, move on, try others. The people who start from a wish list build one thing, get a real result, and then ask what's next on the list.
The logistics team I worked with recently is a good example. They came in with a long wish list — automating carrier emails, generating quotes, getting market-rate alerts, sending customer ETAs. Good list. But they were ready to build everything at once.
I pushed them to pick one. The single workflow that would create the most obvious ROI if it just worked reliably. Build that first. Get it bulletproof. Then add the next.
They ended up starting with carrier email processing. Once that was running reliably, the rest of the list became much easier to prioritize.
The Filter: Daily Beats Impressive
Once you have your list, one more filter to apply before you pick where to start.
Don't start with the rare impressive stuff.
There's a natural pull toward the flashy use case — the agent that does something complicated, the workflow that sounds like it belongs in a TED talk. Those are often the worst things to build first.
Start with what you do every single day.
The daily friction — email prep, note-taking, scheduling back-and-forth, research for calls, generating the same report every week — is where the compounding ROI lives. A task you do once a quarter isn't worth automating first. A task you do 10 times a day is.
I call this the 80-20 approach to agent building. High-frequency pain creates compounding ROI. A 20-minute daily task that AI handles in two minutes saves you 18 minutes every single day. Over a year, that's weeks of your time back. An “impressive” monthly task that AI handles in 10 minutes instead of 2 hours saves you 22 hours a year — less than one week.
The daily stuff wins. Almost always.
What Happens After the Wish List
Once you've picked your starting point, the path gets clearer.
Build one thing. Get it working. Use it for a few weeks until it's reliable and you trust it. Then go back to the wish list and pick the next item.
This approach accomplishes something important: it builds your judgment. After you've built or deployed a few automations, you get much better at reading your own wish list. You start to see which items are actually AI problems and which ones are process problems in disguise. You get a feel for what's worth the setup time and what isn't.
John, the auditor, started with audit note-taking. After a few weeks of that working, he came back to the email history piece — which by then he had more clarity on how to frame. Each iteration made the next one easier.
The wish list isn't just a starting exercise. It's a running list. I'd encourage you to keep it somewhere accessible and add to it whenever you catch yourself thinking “ugh, I wish I didn't have to do this manually.”
Every one of those moments is an opportunity.
This week's action: Block 10 minutes. Write down five to ten things you do regularly that feel slow, manual, or repetitive. Then pick the one that happens most often. That's where you start.
