An AI Payoff You Can Count: Build a No-Show Reminder, Start to Finish
A TechCrunch piece got picked up on Hacker News this week. Apollo's chief economist, Torsten Slok, looked at the S&P 500 minus seven big tech names, and profit margins and revenue per worker look flat to him. His conclusion: AI has delivered "zero financial payoff to corporate America outside of tech."
The thread pushed back. One top comment says the note only looks at company-wide revenue, so it can't tell you what any actual AI project paid back. Another says that when AI is the right tool it's a silver bullet, and when it isn't, you spend more time steering it than you'd spend doing the work yourself.
I think both are right, and for a small business the lesson is simple. Don't point AI at "productivity." Point it at one problem with a number on it. If people book time with you, the cleanest one is no-shows. An empty chair is lost money, you can count it, and most people who miss a booking just forgot.
So let's build a booking reminder on DontCode, then check whether it actually paid off.
1. Booking page, plus a baseline
Create the app by telling Insight about your shop: what you offer, your hours, and that customers book a time. The first build is your landing page, and on your first app it's free. Then ask for the booking form as your next feature.
Two details matter. Ask for an email field, because the reminder goes out by email. And ask for a "showed up" checkbox on each booking. Bookings land in a table on the Database page, which works like a spreadsheet: when someone walks in, double-click the cell and press Apply.
Run it like that for two weeks before you add reminders. That's your baseline. Skip it and you'll never know whether the next step did anything.
2. Ask for the reminder
Type it into the Insight chat the way you'd say it: "The evening before each booking, email the customer their time, what they booked, our address and a button to change or cancel."
That's the booking reminder recipe, so Insight already knows the shape. It asks a few things first: when reminders should go out, whether people can cancel or only change (and up to how late), and anything worth mentioning, like parking. Then you get a build card with a credit estimate. Press Build it.
3. What runs every night
At 6 PM in your time zone, the job finds tomorrow's bookings that are still on and emails each customer. The change or cancel button works without signing in, and a cancellation frees the slot right away so someone else can take it. Nobody gets the same reminder twice.
Want 8 PM instead? Open Scheduled jobs, pick the new time, Save. It's a scheduled job, and you never see cron syntax.
4. Did it pay off?
After two weeks with reminders on, open Emails, then Activity, and make sure the reminders were delivered. Then go to the Database page, use Export CSV on your bookings table, and count no-shows against your baseline.
If the number dropped, that's a real payoff, and you can count it in chairs. If it didn't, you found out cheaply. Maybe your no-shows are really late cancellations. Ask Insight to send the reminder two days ahead, or to move the cancel deadline earlier, and measure again.
Slok wrote that "in business, intelligence is mostly judgment." I agree. Picking the one problem worth fixing is the judgment part. The reminder is just plumbing, and plumbing should take an afternoon.
If people book time with you, give it a try on DontCode. And if no-shows aren't your problem, the shop recipes have a few other numbers worth going after.