Amex and Microsoft turn to AI to make expense reports less horrible

According to Amex, the initial solution will leverage machine learning and AI to automate expense reporting and approvals.

This goes beyond simply learning how to classify certain expenses, as many of today’s tools already do. Instead, the new system will implement an AI-powered decision engine that understands the company’s own travel and expense (T&E) policy and how it applies to submitted expenses. It will use that understanding along with other factors — like the employee’s purchase and payment history — to categorize and assign a risk score to individual transactions.

To make this work, the employee will be prompted to snap a photo of their receipt after paying with their Corporate Amex card. The system will then apply one of three risk scores: red, yellow or green, based on whether the expense is recommended for automatic approval or not, or if it needs another look. This information is passed along to the company’s expense management system with the receipt details attached to automatically generate reports for managers and auditors to use in their own decision-making. Amex says the AI is something it built in-house — it’s not leveraging Microsoft’s partnership with OpenAI, as Bing is, but is using Microsoft Cloud.

Over time, the expense management system will get smarter thanks to machine learning. As more expenses flow through the system, it will improve its algorithms around what sort of expenses can be automatically approved.

Microsoft will be the solution’s first tester and will integrate it with its own internal expense system later in the year, American Express notes. Over time, it will roll out to more Amex Corporate clients and add support for more expense management tools.

While it’s not exactly ChatGPT for expense reports, if successful, the solution could save time and reduce headaches around corporate expense management.

Imagine, instead of spending hours manually categorizing expenses, uploading receipts and justifying the charges, corporate employees would only have to focus on the outliers that actually required further explanation.

 

Of course, it remains to be seen if the solution is capable of actually accomplishing this goal, as described, or if corporations will even utilize the tech when it becomes available.

Amex also isn’t the only company looking to AI to improve the tedious processes around expense report management. Just days ago, Palo Alto-based business-travel software maker TripActions rebranded to Navan and announced it would integrate ChatGPT into its platform for similar reasons. The company said its new system would learn a user’s preferred airlines, hotels and restaurants to build itineraries and will streamline expense reporting through its own receipt-scanning tool, among other things.

There’s reason, of course, to be skeptical of these forthcoming AI solutions.

As The Atlantic recently pointed out, new technologies meant to reduce employee labor inevitably just create “new types of work” for people to do instead. And ChatGPT’s way of confidently providing the wrong answer suggests some of that extra work may involve dealing with false positives and negatives.

 

Plus, other companies have been leveraging AI for some time, like SAP Concur. And many employees would argue that Concur isn’t exactly a user-friendly system.

The timing of the announcement is suspect, as well. Amex is likely hoping to ride on the wave of interest in the Microsoft-OpenAI deal and the ChatGPT-powered Bing to get more eyeballs on its far less exciting use case. (Unless, of course, expense management thrills you!)

Still, AI is on its way to the broader financial services and corporate travel industry, not just expense management. In that respect, Amex isn’t necessarily getting out ahead of the market, it’s just keeping up with where it’s headed.