Ted and Daniel were interviewed by the FT about how AI agents could help with our travel arrangements.
“Ted Lappas, chief technology officer of Conscium, a start-up developing AI agent verification services … envisages a world in which we each run our own “digital twin” owning all our data and operating in our own interests. ‘I want to see a future where the consumer builds the twin,’ he tells me.”
Read the full article:
Te Anau in New Zealand is about as far removed from the tragic turmoil in the Middle East as it is possible to imagine. As the gateway to the South Island’s Fiordland, the tranquil town is all about stunning scenery, leisurely cruises and hipster vibes.
But even at the bottom of the world my wife and I were frantically following the US-Israeli attack on Iran, which erupted while we were staying in Te Anau. No news addict can disconnect from such an event. Besides, we were focused on a more humdrum concern: would our return Etihad flight to London via Abu Dhabi take off the following week? No.
Like many thousands of air passengers around the world, we found ourselves temporarily stranded. Our tortuous efforts to return home highlighted the fragility of global travel networks. But they also exposed the flaws of the current flight booking system and the potential for developing better agentic AI solutions.
It was easy enough to find alternative flights via comparison sites and online travel agencies, such as Skyscanner, Expedia or Booking.com. The trouble was that prices had surged and some agencies failed to deliver the deals they offered. On one occasion, I thought we had secured a two-leg return via Singapore on Expedia. But the site buffered for hours before timing us out.
In such situations, you can feel like the world’s worst mug for failing to find the least bad solution (no doubt FT readers can suggest better ideas).
How much smoother might the process have been if we could have employed our own personal AI agent to help find the optimal flights, secure the best prices, apply for the relevant visas and book everything with our approval? Agentic AI services are being developed to do just that, although they will take time to operate securely and reliably given the complexities.
Over the past few months, OpenAI, Anthropic and Google have all launched commercial agentic protocols that could provide the underlying infrastructure for such services. Google has already highlighted travel as a promising area for the agentic AI future.
At an industry conference this month, James Byers, Google’s travel product lead, said the company was working with industry partners to develop end-to-end agentic travel services, promising to give users “superpowers” they never had before.
Google already operates a travel search engine and could, with user permission, pull in data from its other services. To be effective, an AI agent would have to access confidential information, such as location, bank account, passwords and email. “Trust and transparency about these decisions are paramount,” Byers acknowledged.
One other trust challenge for companies such as Google, though, will be to convince users that their AI agents are working purely for their benefit — and not surreptitiously accepting deals with the company’s preferred commercial partners.
After abandoning the online agencies, I went straight to Air Canada’s website and managed to buy two one-way economy tickets to London via Vancouver and Toronto (with an overnight stay) on three jam-packed aeroplanes for £1,770 apiece. We finally made it home and did at least receive a refund on our original Etihad tickets.
But a good example of how the interests of paying partners can trump those of regular users emerged when I applied for an electronic travel authorisation to enter Canada. The top-ranked (sponsored) link on Google search promised to help sort my application for C$99. The official Canadian government visa site, ranked a lot lower down, speedily completed the process for just C$7.
Many users will be reluctant to increase their dependency on the Big Tech companies. But a radically different future may be possible given how fast adaptable open source AI models are evolving and how individual users are increasingly building AI agents for themselves. One day it may be possible for everyone to create their own personal AI travel agent, if the glaring security concerns can be overcome.
That’s certainly the vision of Ted Lappas, chief technology officer of Conscium, a start-up developing AI agent verification services. He envisages a world in which we each run our own “digital twin” owning all our data and operating in our own interests. “I want to see a future where the consumer builds the twin,” he tells me.
If that really is the future of agentic AI, then bring it on. Not just the world’s travellers will be grateful.
