Booking.com’s Free Expense Tool is Smart, But Here’s What They Left on the Table

Expense management is a top-tier pain point for small businesses. Everyone who's stuffed a crumpled receipt into a wallet at the airport knows it. But putting a hard number on how widespread the pain is helps separate anecdotes from strategy. Our data tracks 47 distinct problems related to expense management across industries, with an average severity of 3.8 out of 5. That’s not random grumbling—that’s a systemic ache that drives churn.

Enter Jason Lemkin’s article on Booking.com for Business’s free expense tool. Lemkin captures the big lesson: don’t start with AI, start with friction. Find where your customers slow down, point AI at that specific spot, ship the simplest fix, and keep the AI invisible. The product team built the thing in six weeks. The go-to-market team followed the same playbook. The result is a free receipt-to-report pipeline that saves small business owners from manual drudgery.

That sequence is worth studying. But our user data also flags two friction points the article didn’t chase, one inside the product and one adjacent to it. If you’re building your own version—or even if you’re using Booking’s tool—these are worth looking at.

The Blind Spot in Invisible AI

Booking’s expense tool relies on AI to auto-categorize expenses: breakfast vs. lunch, vendor name, location. The AI runs silently under the hood, and the article celebrates that invisibility. “The best AI here is the AI you can’t see,” Lemkin writes.

Our data suggests a caution. Among those 47 expense problems, 23 specifically mention a lack of transparency in AI-driven tools, with an average severity of 3.5/5. Users don’t just want the job done; they want to know why a $42 charge was labeled “dinner” instead of “taxi.” When AI makes invisible decisions about money, trust erodes quickly. An error in categorization can mean a rejected reimbursement or a tax headache. In finance, invisibility can feel like opacity.

That doesn’t mean the article’s thesis is wrong. It does mean that “invisible AI” works best when it’s paired with an optional explanation layer. Let the default be invisible, but give a one-tap summary of how the categorization was derived. A simple “We matched the time of day and vendor name to categorize this as dinner” goes a long way. Booking.com didn’t ship that—yet. For anyone building a competitor, this transparency layer could be a differentiator.

The Overlooked Friction: Approval Bottlenecks

The article focuses on receipt capture and report generation. That’s the pain felt by travelers. But our data reveals another friction point that hits finance teams and employees: approval delays. Expense approval bottlenecks are a separate problem category, and we track 8 problems specifically on this, averaging a severity of 3.9/5. The highest severity comes from mid-sized companies (50–500 employees), where approval chains are longer but processes are still manual.

Booking’s tool generates a clean report in about a minute. But then what? If that report goes to a manager who has to review, approve, and forward it manually, the friction just moved downstream. The employee gets reimbursed slower, the finance team chases paper, and the whole system undermines the speed gain at the front end.

A full solution would route reports through smart approval workflows: auto-approve under a threshold, flag out-of-policy items for human review, and integrate with accounting software to push data straight into the ledger. Booking.com’s free tool doesn’t do this—and for a free product, that’s fine. But for anyone building a paid alternative or supplement, approval bottlenecks are a direct sequel to the receipt problem.

What to Actually Build

Lemkin’s core insight still stands. Copying the output is less useful than copying the sequence. Start with friction, not AI. Ship fast. Keep the technology hidden. Those principles worked for Booking.

But the data suggests two places where the friction hunt could go deeper. First, inside the AI: give users a peek behind the curtain when they need it, especially for financial decisions. Second, beyond the report: automate the approval chain so the last mile doesn’t undo the first mile.

If you’re building in this space, don’t just chase the same friction Booking chased. Look two steps ahead.