What if a group of people planning a trip could explain their objectives in plain terms to an AI-powered platform, and then within moments have a comprehensive itinerary that worked for everyone, ready to purchase with one click? No more going back and forth on WhatsApp trying to agree how, where, what, how much, and when. At TenFive we’re using AI to solve the types of problems that groups of people have when they need to come together to make complex plans.
A group of people planning a trip is a great example of the kind of problem TenFive solves. The collective consensus on what makes a trip work emerges from the expectations and requirements of each participant, and these in turn are identified through a process of turn-taking, cooperatively negotiating about what’s possible. Planning a journey by committee will typically involve several rounds of checking up on variations of candidate transport, accommodation, and activity options – and if someone in the group is too slow, options change, which can be frustrating.
Imagine if the back-and-forth involved in planning a trip could be handled by an AI. Current single AI models, while they can be fantastic at sifting through data and communicating about what they find, are not up to the task of understanding how to work through the logic of a set of potentially conflicting objectives. Even state-of-the-art “reasoning” models come up with outputs that can be plain wrong: they look reasonable but they don’t actually meet the criteria given, and they sometimes even output non-existent travel options. This is because these models have actually just been trained to mimic the way humans use language to express ideas and objectives, not to actually have ideas and objectives themselves. For the same reason, it’s impossible to understand what went wrong; we just get something unsatisfying coming out of a black box.
At TenFive we’ve developed a technique which embraces the inherently linguistic aspect of generative AI, and we’re especially interested in one of the things language does best: communication about goals between multiple parties. Rather than try to smash through complex problems and fuzzy logic with huge amounts of opaque computation, our framework is grounded in the idea that a set of cooperating agents that don’t necessarily need to each be doing anything too complicated can communicate using language to work towards an emergent solution. Because reasoning happens between multiple agents rather than inside individual agents, the pathway towards a solution is plainly visible, and so up for further discussion if desired.
Group travel – friends planning an adventure, colleagues travelling together for business, a large extended family working out the details of a reunion – is a great example of the type of problem-solving our system is designed to handle by deploying not one AI but a network of AI agents working together towards the goal of a fully scoped out trip. Some agents can be associated with the requirements of an individual (“I really want a hotel with a good gym, and I hate flying early in the morning”), and other agents can be responsible for a specific aspect of the trip (a hotel agent can negotiate with a flight agent to discover optimal options for amenities versus flight times, based on everyone’s preferences). Agents can also be equipped with policies regarding peripheral aspects of travel, like insurance or environmental impact targets. All of this can be set up through a simple user-friendly interface combining familiar fields for indicating things like date ranges with plain natural language inputs phrased in users’ own specific terms.
We’re looking for a partner in the form of a pilot customer, willing to team up with us to provide the seminal AI-enabled application for a new kind of travel service. We think the product we’re developing would be a great fit for a large-scale, technologically savvy travel platform provider with an established customer base and the drive to provide top-notch user experience.
Please click here if you’d like to schedule a chat with us.