AI Agents Booked 4 Uber Rides. Skip the Agent-Ready Panic

On its third-quarter call, Uber said AI agents had sent it around 500 Uber Eats orders and 4 rides. The comparison figure was about 320 million trips in Halloween week. That's from Yahoo Finance's write-up of the call, which also covers CEO Dara Khosrowshahi saying he doesn't want Uber depending on "the Googles of the world" again.

Four rides. Uber has a huge brand, simple purchases and customers who come back constantly. If agents were about to take over buying, you'd expect to see it there early. Instead the ride number fits on one hand.

So when small business owners ask us whether they need to get ready for AI shopping agents, our answer for most of them is no, not with real money, not this year.

"Customers will soon buy everything through agents"

Maybe some will, eventually. But "eventually" is doing all the work in that claim. By our math, 500-odd orders and 4 rides against 320 million trips is far below a thousandth of a percent.

One company and one week isn't the last word. We're not saying agents will never matter. We're saying nothing in this data justifies rebuilding your storefront, booking flow or quoting process around them now. If you run an HVAC shop, a dental office or a three-location restaurant, people are still finding you through Google, referrals and phone calls.

The vendor who says you're about to be invisible

You may have gotten the pitch already. AI assistants will soon buy on behalf of consumers, businesses that aren't "agent-ready" will vanish, and for a setup fee plus a monthly fee someone will connect you.

Maybe the vendor means it. It's still insurance against something that, going by Uber's numbers, barely exists yet. A small business has a small experiment budget, and every dollar and hour that goes into a speculative integration doesn't go to something that pays back this quarter. We also suspect the technical standards will keep shifting, so whatever you build now may need redoing. Let bigger companies pay for that lesson.

The cheap part that's worth doing

Some of the "get ready" advice is fine. It just doesn't need a vendor. If someone asks an assistant for a plumber near Logan Square or a caterer who can feed 60, the answer most likely comes from the same public information a person would find. So the prep is dull:

• Hours are correct on Google, Yelp, Apple Maps and your own site, holidays included.
• Services and service area are written as plain text on your website, not buried in a PDF or an image.
• Prices, or honest ranges, are somewhere a person can find them.
• You answer reviews, and the recent ones don't contradict your site.
• The site loads quickly on a phone and the phone number is tappable.

That's an afternoon, maybe a little of a web person's time. It helps with Google and with human customers, and if an agent ever does send you business, you're covered. If none does, you still fixed things that were costing you calls.

Where the money is today

For a small business, AI pays off mostly on the inside right now. Look at where your people lose hours: the shared inbox nobody wants to own, quotes rebuilt from scratch each time, follow-ups that go out only when someone remembers, invoices keyed in by hand, scheduling back-and-forth, and the same dozen customer questions answered again and again.

Those problems are old and the fixes are known. A chatbot answers FAQs and takes booking requests at 9 p.m. when the office is empty. An automation drafts the follow-up email and puts it in front of a human for a quick read. A workflow picks up incoming invoices and files them. None of it depends on how people shop in 2030. It depends on whether your team is buried in repetitive work today, and you probably already know.

These projects do break, usually in boring ways. The source data is messy, the bot has no clean handoff to a person, or nobody owns the workflow after launch. Those are better problems to be solving than whether a robot buyer can see your business.

Two questions for any AI pitch

Vendor, conference talk or a cousin who reads a lot of tech news: ask these before you spend anything.

1. Can they point to a customer behavior happening now, in numbers, at a business like yours? "It's coming" doesn't count. Four rides is the bar to hold people to.
2. What will you measure after 60 days, and what happens if it doesn't move? With no number and no exit, it's a bet, not a purchase.

A good test on top of those: does it save your team time or recover revenue you're losing now, or does it only position you for some future? Fund the first kind. Be stingy with the second.

When to look again

Watch the earnings calls over the next couple of years. When companies like Uber report agent-driven orders as a real share of volume instead of a few hundred, take a look at what the booking and ordering tools in your industry are doing. Our guess is the platforms you already pay for will add support, and you'll switch it on rather than commission a custom project.

Until then, clean up your public listings. Then pick the one internal task that eats the most hours each week and price out automating it. If you want help choosing, our free 30-minute assessment will point at the best candidate. It will also tell you plainly if AI isn't the right answer for you yet.

---