Latest / Future of Work Tech with Fexingo: Remote Tools, AI Productivity, and Workplace Software / How AI Is Automating Your Expense Reports
Transcript
- Lucas: Luna, how many receipts do you think you've filed in your career? Luna: Oh, thousands probably. And I still have a shoebox from 2018 that I'm afraid to open. Lucas: Right, because at some point you just stopped. That shoebox is basically a monument to the old way of doing expenses — manual data entry, paper receipts, and the eternal hope that your manager won't ask why you expensed a $12 coffee during a remote day. Luna: Hey, that coffee was essential for productivity. But seriously, the whole process is a drain. I've heard that the average employee spends about 20 minutes per expense report. Multiply that by a thousand employees and you're losing serious time. Lucas: And that's exactly where AI has started to eat into the problem. So today I want to focus on a specific company that's doing this really well: Ramp. It's a corporate spend management platform that uses machine learning to automate expense reporting — from receipt capture all the way to policy enforcement. Luna: Ramp I know. They started as a corporate card for startups, but now they've built out a whole suite. What's the AI doing under the hood? Lucas: So the core technology is optical character recognition combined with natural language processing. When you snap a photo of a receipt, the system doesn't just grab the total — it reads the merchant name, date, line items, tax, tip, all of it. Then it matches that data to the transaction on your Ramp card. Luna: That matching part sounds straightforward, but what about the edge cases? Like a receipt that's crumpled or faded? Lucas: Great question. The model is trained on millions of receipts, so it's actually quite robust. Ramp claims their OCR can handle receipts in 40 languages and deal with variable quality. But here's the real kicker — the AI also learns your company's expense policy. So if you try to expense a first-class flight when the policy only allows economy, it flags it before you even submit. Luna: So it's proactive compliance. Instead of your finance team combing through reports after the fact, the AI stops violations in real time. Lucas: Exactly. And it gets smarter over time. If the finance team manually approves or rejects certain items, the model incorporates that feedback. So after a few months, the system can handle the vast majority of approvals automatically. Luna: I've read that Ramp claims to reduce the time spent on expense reports by up to 80 percent. Is that realistic? Lucas: I think it depends on the company, but the numbers are in that ballpark for teams that fully adopt the automation. Ramp published a case study with a company called ClassPass — the fitness subscription service. Before Ramp, they had two full-time finance people just doing expense reconciliation. After, they reduced that to a part-time role. And the average expense report went from 15 minutes to under two minutes. Luna: That's a huge shift. But I wonder about employee trust. If the AI rejects your expense, do you feel like you're being watched too closely? Lucas: It's a valid concern, and I think companies need to be transparent about the rules. But the flip side is, the AI can also flag things that are legit but might have been missed. For example, if you have a recurring subscription that you forgot to cancel, the system can highlight it as a potential saving. So it's not just a cop — it can be a helper. Luna: Speaking of subscriptions, what about the broader trend? It feels like every fintech startup is adding AI to expense management. Brex, Expensify, even traditional players like SAP Concur are incorporating machine learning. Lucas: Right, it's become table stakes. But Ramp's differentiation is that they own the entire payment flow — the card, the software, the bank account. That gives them a data advantage. They see every transaction in real time, so they can do things like automatically categorize spend without the employee having to do anything. Luna: So the vision is a completely frictionless experience. You spend, the AI logs it, and at the end of the month your expense report is already done. Lucas: Exactly. And that frees up finance teams for higher value work — like strategic planning, vendor negotiation, or analyzing spend trends. Ramp's CEO, Eric Glyman, has said that they want to turn finance from a back-office function into a strategic partner. And AI is the engine for that. Luna: I love that. But let me push back a little. What about fraud? If the AI is approving things automatically, couldn't someone game the system? Lucas: That's where anomaly detection comes in. The AI builds a baseline of normal spending patterns for each employee and for the company as a whole. If someone suddenly expensed a $5,000 dinner on a Tuesday in a city they're not traveling to, the system flags it for human review. Ramp claims their model catches 99.5 percent of fraudulent transactions before they're approved. Luna: That's impressive. But it also makes me think about the human side. What happens to the finance clerks whose jobs are being automated? Lucas: It's a real shift. But I'd argue that the role evolves — it becomes more about oversight and exception handling, less about data entry. Someone still needs to review the edge cases that the AI can't handle, and those cases often require judgment. Plus, as companies grow, finance teams often end up doing more strategic work. Luna: It reminds me of when ATMs were supposed to kill bank teller jobs, but instead the number of tellers actually increased because banks could open more branches. The nature of the work changed. Lucas: Exactly the same dynamic. So the technology might not eliminate jobs, but it will redefine them. And for employees, the benefit is obvious — no more chasing receipts or filling out forms. That extra 20 minutes per report adds up. Luna: And that's the kind of concrete improvement that makes people actually appreciate technology. It's not about some grand AI revolution — it's about not having to manually categorize your coffee receipts. Lucas: Right. And speaking of appreciation, I want to take a quick moment to mention something. This show is possible because of a small group of listeners who chip in through Buy Me a Coffee. It's at buy me a coffee dot com slash fexingo. No ads, no sponsors, just listener support. It keeps us independent and focused on the topics that matter to you. Luna: Yeah, it's a small thing that makes a big difference. And we really appreciate everyone who's already part of that community. Lucas: So back to expense automation — one area I think is still underdeveloped is receipt digitization for non-paper receipts. You know, like digital receipts from Uber or Amazon that come as emails. Ramp and others can parse those too by connecting to email, but it's not seamless for all merchants. Luna: Right, I've noticed that. Some platforms can only handle PDFs or images, not the inline HTML receipts. That's a work in progress. Lucas: And then there's the international dimension. Different countries have different tax rules — like VAT in Europe versus sales tax in the US. The AI has to understand those nuances to correctly categorize and report expenses. Ramp has been expanding internationally, so they're building that out. Luna: So the technology is still evolving. But the direction is clear. I think within five years, the idea of manually filling out an expense report will seem as archaic as using a paper map for directions. Lucas: I'd agree. And for companies that adopt early, there's a real competitive advantage — both in cost savings and employee satisfaction. Nobody likes doing expenses. So if you can remove that pain point, you're making your workplace a little bit better. Luna: And that's a win-win. Less friction for employees, more data for finance, and better compliance for the company. Lucas: So that's the story of AI in expense reporting. It's not flashy, but it's one of those behind the scenes changes that actually improves how we work. Next time you snap a receipt, just think about what's happening on the other end of that photo.