A recent article called “Next-Gen Expense Management” by JoAnn DeLuna in BTNposited that automation and artificial intelligence could “all but eliminate expense reports and approvals.” The idea is intriguing because expense reporting (especially done manually) can be a dreaded task for employees. However, the notion that expense management will be (or can be) completely eliminated by artificially intelligence isn’t supported by the examples DeLuna cites. That’s because none involve artificial intelligence.
In this article, several system capabilities were attributed to the power of artificial intelligence when, in fact, those capabilities were simply automation. Steven Mandelbaum, who according to the article “has been on the leading edge of this concept,” has used an “email-like inbox” of individual transactions of “post-trip expense reporting” provided by employees.
In fact, we noticed that employees were doing a lot of the work throughout this process: travelers provided the data through individual transaction reports and then they were doing the organizing of these reports by “choos[ing] a purpose for each expense from a drop-down menu.” Finally, managers were manually reviewing those individual transactions based on which employee submitted which data. Mandelbaum’s contribution was to expense reporting design--separating expenses by transaction rather than by report. It didn’t reflect in any way artificial intelligence unless you want to suggest that Mandelbaum is a robot (which of course we don’t!).
Throughout this process, humans were busy doing complex interpretive tasks. The software was restricting choices and applying rules, but never rose to the level of artificial intelligence in any of the referenced systems.
We, of course, agree with the article’s point that computers can be much better than human approvers at applying clear-cut rules to expense reports. They do a great job of matching transactions and presenting instances, for example, where two employees are submitting transactions for the same meal. They are also far more trustworthy when adding columns of numbers. Neither example is artificial intelligence.
Frankly, the automation discussed in the article is not a sign of wondrous things to come or even an advance in the industry’s programming sophistication. Many of us have been doing some pretty cool things for many years. The change is in the packaging. The article is really about how players in our space are dressing up small incremental enhancements to their products as breakthrough technology. It isn't breakthrough technology, but good try guys! The problem remains that meaningful differentiation is increasingly harder to come by.
You want real AI in business travel management? Think about eliminating most business travel. Why would you need traveling salesmen, for example? To talk to the customer’s AI procurement system? You get where this is going.
DATABASICS is relied upon by leading organizations representing all the major sectors of the global economy: financial services, healthcare, manufacturing, research, retail, engineering, non-profits/NGOs, technology, federal contractors, and other sectors.