In second grade, I was given two desks because I couldn't keep my things organized enough in one. I was the only kid in the class with two desks. Looking back, my parents probably should've gotten me help. Instead, the teachers just gave me two desks.
My whole life, I've been finding crutches for my inability to complete executive function tasks. Now, with AI, I've outsourced as much of my executive function as I can, because, well, I'm crappy at getting things done, but I'm pretty good at figuring out what needs to get done.
This quarter I've been leaning into every place AI can take over tasks that take me and my team longer than necessary. Here's what we're using the most.
1. A chief of staff that never sleeps
I built a set of Claude skills that run on a schedule, checking my calendar, email, Slack, and Linear so I don't have to. It preps me for meetings, flags what needs a reply, and catches the stuff that would otherwise fall through the cracks between my desks.
2. Product photography without the photoshoot
We're using FASHN to put clothing on AI-generated models instead of booking a shoot every time a client needs new product imagery. What used to be a multi-week production is now something we can turn around same day.
3. Bullet points at catalog scale
We generate PDP summary bullets across catalogs running into the thousands of SKUs, a job that used to cost a copywriter weeks per client. Now it's a pass we run and review. We're testing them at the top of the product page, and seeing real CVR gains.
4. Closing the relevance gap automatically
Throughline, the product we've been building at Mobile1st, dynamically matches landing page content to the ad that actually drove the click. Instead of every visitor landing on the same generic page regardless of which ad brought them there, the page adjusts to match.
5. Building bigger experiments in less time
Using Opal and Claude, we're able to build more complex, multi-change experiments in the time it used to take to ship one simple variant, and have the AI verify the implementation actually works before it goes to real traffic.
6. Ad creative that copies what's already working
We're testing ways to generate ad variations modeled on our own top performers, so instead of guessing at the next creative, we're iterating off what's already proven to convert. Still hunting for the right tool here. If you've found one that actually works, tell me.
7. A daily check that catches broken tests before they cost us
A scheduled check pulls our GA4 experiment data every day and flags in Slack the moment a live test's event counts drop or stop firing. Before this, a broken test could quietly burn two weeks of traffic before anyone noticed.
I got two desks in second grade because I couldn't manage one. This quarter, AI has basically become desk number three. What did I miss?