April 23, 2024

Here AI go again

Like most of us, Scott Wylie at TripStax hates doing his travel expenses, so he's been working on a way to make it easier with the help of AI

A lot has been written about how artificial intelligence will totally change life, the universe and everything. We’ll have to see: our three dogs, Bow, Ozzie and Reggie, will still want walking whatever happens: and, if their master has any say in the matter, that walk will still end at an alehouse –or at least it will when my New Year resolution to stay off the sauce for three months finally ends on March 31st.

But I do believe AI truly will prove transformational. Sometimes, though, it can be hard to understand what that change might look like, so I thought I would give you a real-life example of a prototype we have been testing.

It’s a project I have a vested interest in making succeed because I hope it will help end a chore I absolutely hate: submitting my dreaded expenses. My boss Jack ordered me at the beginning of 2024 to start getting my monthly claims done on time. Not a pleasant thought if, like me, you consider expense filing the work of the devil.

The way most expense reporting works today is that if travellers book a flight, hotel or whatever through their regular online booking tool or travel management company, the transaction is automatically captured and flows into their expense report. But if the traveller books, say, direct on Ryanair’s website, or on a car-ride app, then they often have to add the booking manually as a line item to their expense claim.

There are solutions which capture these “off-piste” bookings, by automatically extracting the information from the email reservation confirmation. But, from a techie’s point of view, this process is a nightmare to get right. It relies on those emails conforming to a regular template structure where details like the airline name, origin and the destination are always located in the same place. If any details move position on the email even marginally, then identification can fail.

Another expense report challenge is batching together different line items (flight, hotel, dinner and so on) to provide visibility of the whole trip. If you can create an entire trip record, that helps understand travel costs much better, and the merged record can then be passed to other systems. But, again, if the traveller isn’t to do this manually, you need very sophisticated technology to sift through information like dates, cost centres and email addresses from lots of different sources to make reliable matches of the right line items for the right person.

The pilot we have worked on uses AI to do all of it. Where AI makes a difference is that it deploys large language models to seek and recognise certain types of data irrespective of where and how they are located in the document it’s scanning. That flexibility allows it to be much more flexible than the older tech it is replacing. What’s more, AI keeps learning, so the success rate just keeps climbing.

The pilot has gone even better than we dared hope and really brought home to me how much potential AI has. There’s plenty of innovation just around the corner.

Talking of AI potential, I recently heard that it’s being used to improve the taste of vegan meat with ‘smart ingredient pairing’ techniques – it’s the new flavour matchmaker apparently. Call me a cynic, but I think I’ll stick with my dry-smoked sizzling rashers for now, thank you.

Scott Wylie is Chief Technology Officer at tripstax.com