At first, this seemed a bit like a joke: a cloned chat agent named Phil the Voyager is ready to help travelers discover their next destination through a series of "live" chats? Seriously?
Yes, seriously. It's like IBM's Watson, except without much of the fabled pedigree of that particular slice of artificial intelligence. This tech comes from a company called ExClone, Inc. (TLabs here), which exists to create "talking agents via digital cloning of human expertise." Lofty ambitions, for sure. So how does it work?
So far, it's functional for the trip inspiration phase but has some very clear limitations in this early iteration, both in its visual appeal and overall functionality.
Phil always starts the conversation with a few questions to frame out the discussion.
These questions include what sorts of activities the user is looking for, what culture engagement is desired, and other points of interest to the user.
Phil will then pop out some suggestions, which the user can select to learn more. Phil can even pop in some images when the user selects a particular destination for further exploration.
One current limitation that needs to be improved for this tool to be truly useful is the lack of follow-up; Phil just can't offer more information on specific items. For example, let's say that Phil suggests Angel Fire mountain in New Mexico. In his background he mentions a "classical music festival." Upon further query — "tell me more about the classical music festival" — Phil simply repeats his previous statement. That forces the user to Google, drastically reducing the utility of this digital clone.
There clearly seems to be some sort of database that logs each of these individual pieces of information about the chosen destinations. This can then trigger a frustrating overlap of suggestions when Phil can't find a follow-up answer. Here's another example:
Another clear limitation to the technology in its current format is its lack of pricing transparency. For example, Phil is unable to help users price out flights or other travel to the suggested destinations.
Another area that seems to be a miss is Phil's inability to help users sort destinations according to how much an average trip would cost. For example, when choosing beach destinations, Phil will offer several selections. Some of these suggestions could be unfamiliar to the user, who would then want to simply sort according to budget.
Phil has no comment on that, and that's unfortunate.
Give Phil a whirl here, and see if you can stump him with your tough travel questions.
Regardless of its current weaknesses, the technology does show promise as far as being able to offer a new channel during the trip inspiration process. If the feature set expands to actually help users book the trips, there could be another existential threat to travel agents driven by technology.
A booking-enabled virtual chat agent would certainly be useful on the customer service front and could be something deployed by OTAs as an additional customer service channel, one without the costs associated with human agents.