When ChatGPT
and other generative artificial intelligence tools emerged – seemingly out
of nowhere – in late 2022 and early 2023, the talk was all about the chat
functionality and using these large language models to find and generate
precise information. In the relatively short amount of time since then, there
has been rapid
development of new capabilities from OpenAI, Google and others, including custom
GPT apps, image and video generation, multimodal functions and more.

This year much
of the buzz has centered on AI agents – a more sophisticated version of a
chatbot, if you will, that can not only answer questions and generate content
but can also make decisions and take actions autonomously.

As Microsoft co-founder
Bill Gates described AI agents on his blog last November, “In the next five years … you
won’t have to use different apps for different tasks. You’ll simply tell your
device, in everyday language, what you want to do. And depending on how much
information you choose to share with it, the software will be able to respond
personally because it will have a rich understanding of your life. In the near
future, anyone who’s online will be able to have a personal assistant powered
by artificial intelligence that’s far beyond today’s technology.”

For e-commerce, a critical piece of the AI agent puzzle is
payments, and just last week a company launched that is working to solve that. Skyfire,
built by former executives from Google, Ripple and more, said its system
is the “world’s first” payment network to allow AI agents to make and receive
payments autonomously. 

Clearly, the momentum is building.

And one of the companies leading work in this space is a
startup that has had a rapid trajectory.

Div Garg and Omar Shaya met in 2022 while they were both
working on graduate degrees at Stanford – Garg a PhD in AI and Shaya an MBA. One year later, their startup, MultiOn, entered the spotlight in part due to
the high
profile investors in its seed round – General Catalyst as the lead
investor, along with Amazon Alexa Fund, Samsung Next, Maven Ventures,
individual investors from OpenAI and others. And according
to The Information, another round of $20 million from many of those same
investors is coming soon.

Travel use cases are a top priority for the company. In a blog
post in January to announce the official launch of MultiOn, the company
wrote, “MultiOn will book your dream travel itinerary in minutes” with a brief
video demonstrating a flight booking.

In November, Garg will provide a demonstration of MultiOn at
The Phocuswright Conference.
In the lead-up to that, we talked to Garg to learn more about MultiOn’s
capabilities. The conversation has been edited for brevity and clarity.

First how do you define an AI agent?

An AI agent is something that can interact with its
environment – this could be something physical like a self-driving car or
something virtual where you have a virtual agent that can take actions on your
computer, like interact with an API. So we define an AI agent as something that
can do decision-making in its given environment and can go do things
autonomously.

And can the agent recall and iterate on past interactions?

It does. We can create a profile about the user, such as “I
like a window seat or an aisle seat when I’m flying” or “I prefer United over
Delta.” The agent can remember this information and also your address, maybe
your passport number, stuff like that, and then can automatically go and fill
it for you. So it can create a profile for you and then automatically act on
that information. 

You have used travel examples several times, on social media
and in blog posts. Why is travel an applicable, maybe an ideal, use case for an
AI agent? 

We’ve talked to a lot of users, and everyone complains about
booking travel because it takes a lot of time – you have to choose from all the
options, spend a lot of time finding the best flight. It’s a process – open your
laptop, find the right things – and that is something that can be fully
automated. You can talk to an agent and say, “Book me a flight to New York” and
the agent knows you enough and can make decisions for you. It can say, “Here
are three options” and then you choose one and then it can go and automatically
pay on your behalf, for example.

So travel is a big one. There’s so much planning, so many
choices and so much friction. Most people also don’t like flight websites,
they’re not the best optimized, it’s kind of a pain to use them, fill out all
the details and stuff. But you can now have this agent and it’s kind of like
your concierge service – you say do this for me and then you don’t have to
worry about anything. So we’re seeing a lot of people really like that.

So the AI agent can complete the booking, including payment?
Tell us more about that. 

We are thinking of partnering with a company like Stripe, so we can have a wallet specifically for the agent where you can put some money. Suppose
you put in $1,000 that the agent can spend, and if you want to book a flight
the agent can go and pay on your behalf, but you also get a sort of 2FA flow [two
factor authentication] where you have to confirm if the agent wants to make a
transaction. We’ve been thinking a lot about how can you make this a safe
experience, especially once there’s a lot of money involved, like for travel. We
don’t want to have unauthorized transactions, and we don’t want to accidentally
book a wrong flight, or to the wrong place, where you might lose money.

So we’re looking at how can you build trust … how can you have almost like a
shared credit card program, like if you have a credit card for your child. So
it’s like having a shared card with the agent, and the agent can spend something
but then you get notified and you can approve or disapprove. We are internally
testing it out – and we are seeing positive responses from the users that have
tested it.

I would think trust and security are critically
important. Can you explain more of what you are exploring there?

We are thinking about biometrics as an example. There’s a
lot of things happening with passkeys these days, so we are also looking into
that. Or even can we link it to your Apple ID. And then we’re also thinking how
can we build some sort of digital ID, especially for the agent. Suppose your
agent is going and doing things, it can’t have a fingerprint about you, so if
it’s communicating with a website can it say, “This is Div’s agent or this is
Mitra’s agent,” so the website knows whose agent this is. So can you
communicate an identity to websites … and agents can interact with one another.

One question we hear asked quite a bit is how will discovery
work in an AI agent-powered world – and how can travel sellers become visible?

We’ve been thinking about that. Kind of an agent SEO kind of
thing. Can we make use of existing systems so the agent can also focus on that.
It’s also possible we can have some sort of auction-based models, or partnership
models. This is still early. We’re trying to figure out what makes sense, and
we’re trying to make it as streamlined as possible. So if you have different
providers, can we make it something that’s not a biased model.

It could be similar to an SEO kind of system where you are
able to expose some keywords that the agent can hone on and give more priority
based on that.

How is MultiOn being used today? 

We have seen a variety of form factors. A lot of people have
been using it for a lot of e-commerce. Flights has been a big one, shopping,
people have been using it for event invites, communication, LinkedIn outreach.
So we’ve seen a variety of use cases. Travel is one that keeps popping up as a
big use case when we have asked users, and so that is something we are also
starting to focus a lot on. We are also thinking of launching a mobile app so
you can use the agent from your phone. You could use the app as you are
traveling to book flights, to find places to visit – the agent can help you do
a lot of this and make the bookings automatically.

Are you primarily focused on direct to consumer or also
enterprise level?

Right now it’s direct to consumer, but we are looking at
partnerships with a lot of different brands. So we have Amazon Alexa as one of our investors, so working
with them, can we unlock some of our capabilities through Alexa, so if you have
an Amazon Echo device you can use our agents with that?

That’s one thing we’re doing. But we’re also considering partnerships with
potentially [brands such as] Expedia, United in the future. Using our agents in
a partnership, we can facilitate things much, much easier compared to what’s
possible right now with websites.

Can you explain how MultiOn could integrate with a travel supplier
or online travel agency?

You could have a one-click booking experience. You could
have MultiOn built in on the website and you press a button and the booking automatically
happens in the background and then you don’t have to worry about anything. … What
happens today is you go to United, for example, and you have to put in all your
information, but if we can remove all of that part and you press a button and …
the agent can finish the whole task.

And this type of integration is in development now?

Yes. We are going to start with flights, but it’s definitely
possible we can go do this for hotels, for cars and stuff like that, so it can definitely
expand. But we see flights are the major pain point right now.

What is your time frame? 

I’d say by October we might have this live.

Finally, what advice do you have for travel industry leaders
regarding agents?

About agents I like to say it’s like the internet before the
internet exploded. Before the internet, everything was very slow, and you had
to have physical stores. But then everything went digital and much faster. I
think the same thing will happen with agents.

With these agent capabilities you will have a lot faster
transactions. Right now if people want to book a flight it’s just so
cumbersome. You have this mental thing like I have to go and spend so much time
coordinating and finding this thing. But if that can happen faster, the volume
of flights [and] the volume of travel will increase a lot.

People will think, “OK, I can go and do this and won’t spend
more than a minute.” And I think the folks who take advantage of this, they can
create more value. So for folks working in the travel industry, if they use
agents and make their processes streamlined and faster, they can get a leg up
on their competition, and they can kind of move toward these new capabilities.

I think in the end the technology always improves over time and then you need
to be the first to capitalize on the new technology. And I think agents will be
one of the biggest technology waves that’s going to happen maybe in the next
three years.