Booking.com for Business Shipped a Free Expense Tool in 6 Weeks. Don’t Copy the Tool, Copy the Sequence.

Booking.com for Business Built a Free Expense Product in 6 Weeks. The Lesson Isn’t About Expenses.

Booking.com for Business has been coming to SaaStr for 4 years. Joshua Wood, who runs the business, opened their AI Annual session with a line that runs against every booth on the floor:

“In four or five years, no one is going to be talking about AI.”

Not because it goes away. Because it disappears into everything. Wood lived through the dot-com boom, where every company suddenly announced “we have a website,” then a few years later “we have an app.” Today every company announces “we have AI.” Hotels stopped advertising “free HBO and free Wi-Fi” the moment it became table stakes. AI is on the exact same path. In a few years it won’t be a feature anyone markets. It’ll just be how products get built.

That framing matters because it changes the question you ask. Wood and Nadine Blokker, who runs Booking.com for Business sales and marketing, spent their session on a more useful one.

Stop Asking “What Can AI Do?” Start Asking “Where Is the Friction?”

Most teams start with the capability. They build the cool demo, the flashy interface, the booth banner that says AI in five places. Booking went through that phase too, including the part where they got their own internal models to give up company secrets they weren’t supposed to have. Then they moved past it.

The question that actually drove product was different: where do customers hit friction? The specific moments where a user thinks “why do I have to do this manually?” or “why does this take so long?”

Blokker was direct about why those moments matter. In B2B, friction is where churn lives. It’s where customers slow down, stall out, and leave for something else. Hunt for friction and you find the work that actually moves retention.

The Friction in Business Travel Isn’t Booking the Trip

Here’s where it got concrete. Booking went looking for the biggest pain point in business travel and it wasn’t the part they already solve. With a decent tool, booking and managing a trip is fine.

The friction is the admin that wraps around it. Expense management.

Everyone in the room recognized the picture Blokker painted. You’re at the airport, you buy a coffee, you stuff the receipt in your pocket where it may or may not survive. At the end of the month you’re staring at a $42 charge on your card trying to reconstruct whether it was lunch or a taxi three weeks ago. Then you manually enter it, categorize it, maybe itemize it in a spreadsheet, and hand it to your finance person, who now gets to chase missing receipts and missing tax.

Small companies feel this hardest. They don’t have the budget for an enterprise expense tool, so the whole thing stays manual.

They Didn’t Rebuild Concur. That Was the Point.

The instinct for a big company is to give small customers a smaller version of what the big company uses. Booking explicitly rejected that. Wood pointed out that Booking itself runs expenses through Concur because it’s a large, complex organization. Their customers are small businesses. They don’t need the heavy process, and giving it to them would have recreated the friction, not removed it.

So they built the simplest possible version, and they shipped it in 6 weeks with a small team.

Here’s how it works. Snap a photo of a receipt. OCR pulls the text, which is nothing new and they were the first to say so. The interesting part is what they layer on top: when the photo was taken, where it was taken, who the vendor was. From that, the system auto-categorizes the expense, and not just “meal” but breakfast versus lunch versus dinner, because that distinction matters depending on the company’s policy. A human reviews and confirms, and that confirmation feeds back to tell them whether the model is working. Then it generates a clean report in about a minute.

For context, Wood noted his own Concur expenses take a lot longer than a minute.

AI Built the Product, Not Just Lived Inside It

The 6-week timeline came from somewhere. Blokker walked through how AI ran through the whole build, not just the feature.

The product and engineering team used AI-assisted coding, test generation, and automatic bug detection to move fast without giving up quality. That’s how a small team shipped in weeks instead of months.

Go-to-market ran the same way. AI to segment customers, generate the messaging, handle design, and personalize at the individual level so the message matched the customer’s reality. Then AI on the back end to analyze performance and pull customer feedback, which fed straight back to product to find the next piece of friction to remove.

The loop is the real product. Build with AI, sell with AI, learn with AI, hand it back to product, repeat.

The Best AI Here Is the AI You Can’t See

The most useful takeaway was the quietest one. Wood pointed out what the expense demo deliberately wasn’t: there was no big flashy “we have AI everywhere” moment. The AI sits silently behind the scenes and just makes the process better, easier, and faster.

That invisibility is also why the product is free. Booking optimized the cost out of the process instead of charging for it. The result is millions of customers on the platform and thousands already using the expense tool within weeks of launch.

What to Actually Take Back to Your Team

The headline number is real: a free, working expense product shipped in 6 weeks by a small team. But copying the output isn’t the lesson. Copying the sequence is.

Don’t start with the AI. Start with the friction. Find the place your customers go manual, slow, or quiet, because that’s where they’re about to leave. Then point AI at that specific moment, build the simplest thing that removes it, and keep the AI out of the customer’s face. In a few years no one will be impressed that you used AI. They’ll only notice whether the friction is gone.