Bret Taylor on Sierra's 400-customer surge and publishing AI agents directly to ChatGPT

Uh we can definitely go over Sierra product updates first, but uh uh do you do you have any immediate reaction to this to to the the timeline because it's crazy right now? The the timeline like are you talking about back? I'm talking about the Yeah, the the X timeline.

This idea of a federal backs stop for AI buildout, the buildout that's going on. Everyone's going back and forth. Do you have a take? I do not have a take. Okay. Uh I I'm sorry for no hot takes on this one. It's like, as you know, we just got off of Sierra Summit. I just been waiting in in uh in the timeline today.

No hot take right now. Sorry for that. Okay. In the trenches with Sierra. Uh let's break it down. What are the biggest updates in your world? Uh so yeah, yesterday we had our first customer conference uh Sierra Summit. It was awesome. I mean, you know, as an entrepreneur, I it was sort of funny.

I had this like before and after post on X and it was like, you know, just a few years ago it was like Clay and me in a whiteboard and now we've got 400 customers in person. Uh but three big product announcements. Um first was some new channels.

So if you think about customer experience, you often think about things like phone and chat. We announced you could publish your agent directly to chat GPT. We think chat GPT is the new front door for the internet.

And if you think about building your customer experience, it's on your first party properties like your mobile app or your website. But it's third party properties like Chat GBT and eventually Gemini and others as well. So the flow there is somebody's looking for they they're having an issue with a product.

They're talking with chat GBT about it and then theoretically it could there there could be some type of popup that would say like do you want to talk with an agent about this specific issue and then that the the individual company would have more control over that interaction. Is that right?

Well, it's sort of it's built on Chat GPT apps which which was announced at developer day uh earlier this month, late last month. And the idea is this. Let's say uh you know you're a home seller and you have a first party agent which you can go to your website and chat with it maybe browse through homes.

Well, a lot of your customers are starting on chat GPT as well. So you have you've done all this investment to make this agent. You have to build it twice uh which parts of your experience you want in chat GPT.

Now with a click of a button you can just publish your agent directly via chat GPT apps and it really means you can have the same experience in your mobile app and your website as you have on chat GBT with this new extensibility model they developed here's the big picture I mean if you just think of any consumer brand uh every new platform brings distribution channels right the uh you know mobile had app store the web had portals and then search in the age of AI it's agents it's chatbt it's Gemini I a lot of our customers are thinking about what do we want our presence to be on our own properties?

What do we want our presence to be when people are starting from chat GPT? We think just the worlds of channels are changing now. It's going from chat and phone and your website to chat and your phone and your website. Chat GPT Gemini agent to agent communication. So we help our our customers build customer experiences.

We're trying to, you know, help them reimagine what that might be in a world where, you know, a person's personal agent might be doing the shopping for them. You know, how do they set themselves up for success there? The complex part is you, you know, there's sort of two sides to that coin.

You know, it's like where do you maybe commoditize your brand? Like what do you want to hold back on your first party experience? What do you want to be present on these third party sites?

So we've really tried to build a platform that's flexible, but we're really excited about it because I think uh you know it is just a brave new world.

You know, people are trying to figure out where demand generation, demand fulfillment comes from in this age of AI and really proud to be one of the first uh first companies to enable it. Yeah. Help me understand a little bit more about the brand side.

I feel like uh like as a consumer there's so many times I'm sure so many people do this in Chad PT.

they're asking about a product and they're just asking the underlying model and they're just assuming that uh if I'm asking about Diet Coke that, you know, that ChatBT has hoovered up all of Reddit on Diet Coke and the actual Diet Coke website and the FAQ page that Diet Coke has been, you know, put up for years.

And so if I ask it how many calories are in there, it's going to get it roughly right. But at a certain point, the other side of the equation wants to have a little bit more control.

And it feels like you can publish that as an app in the chatbt app store, but is there what is the funnel from I just went into a new chat and I don't know that it's in the app store, but and I'm just talking about Diet Coke over here in just a chat.

Is it is at some point chat GPT going to say, hey, you should actually just be talking to the real app. I think that is the big question. You know, I don't think the discovery model is completely known yet. And you know, our it's sort of funny.

our reductive form is like no one really knows where consumer behavior is going to go or even how these discovery models work.

But step one is make an agent because if the only way to use your service is clicking around your website, there's no way to be present in that world because it's just going to be, you know, essentially the index content within the large language model.

So our philosophy now is it's like the early days of the app store or the early days of the internet. I mean, I remember when I was at Google in 2003, the index would update once a month. We called it the Google dance because people's rankings would change.

You fast forward 20 years, SEO and SEM is a super mature industry, right? It's like something that's quantified. We're kind of in the 2003 of this. So, no one really knows, but our whole philosophy is like step one, take your customer experience. That's your website, your mobile app today. Turn it into an agent.

And what that means is enable it to be conversational. Enable it to work over the phone. Enable it to work on your own website, your own mobile app.

then experiment with publishing parts of it to chat GPT as a chat GPT app and you know the discovery model how do people activate it with an app mention does it recognize your intent and light it up I think the form factor here just like newsfeed has changed a lot over the years I don't think anyone completely knows yet but you can't wait until your competitors figure it out first right so this is an interesting thing about these new technologies and one of the things we try to do at Sarah is say hey we're going to be a partner to you to help navigate this world even though it's uncertain and I think a lot of people also there's a question of commoditization you know if an agent is doing the purchasing on behalf of a consumer what's the role of my brand what's the role of my first party property it's almost like if you look at the retail world in mobile which retailers had the brand equity to have their own mobile app which were dependent solely on Instagram for the traffic to their retailer it's a very nuanced question and everyone's kind of figuring it out in real time so I don't want to pretend we have all the answers but What's neat about what we've done is we're saying, "Hey, step one, make an agent and prepare yourself for this future and make it super low cost to experiment, which I think if if we're in the 2003 era of search, being able to experiment quickly is worth a lot.

" And that's kind of the value proposition for our customers. Have you ever seen like enterprises adopt a a a product like they're adopting Sierra?

I'm I'm thinking about the post you did uh maybe it was a couple weeks ago at this point or 10 days ago or something where it was like excited to welcome and then you just listed off like 40 companies in the Fortune 500. I've never seen I've never seen a company ever make a a post like that.

It was pretty pretty shocking. Um but uh yeah, I'm cur I'm curious if there's any any other moment in your career that you've seen this this kind of velocity of of adoption. I really haven't.

And man, I can tell you as somebody who started two companies before, I joke with Clay, my co-founder, there's two kinds of stress as an entrepreneur. The existential kind because you're not growing or the holy everything's on fire kind because everything is growing and we're experiencing the latter and it's awesome.

Excuse my [laughter] language. You know, it's like just because you know, sorry like getting bleeped out here, you know, like I was just horn. [laughter] Yeah. Okay, good.

I fig you know it's it is fun and actually here's the thing actually I'll say one other announcement we had yesterday but just to answer your question like the world is changing a lot and right now companies are looking for a trusted adviser to help them guide them on this new path whether it's what demand's going to come from chat GPT what is a conversational version of my interface look like are my competitors moving faster than I am just think about if you run a large telco right now every you know basis point of attrition is worth millions of dollars and if you can get even a small edge these are the moments in technology where you can actually change market share even in in markets that are sort of oified and so you know I I appreciated that that uh that post on X as well it made me smile when I wrote it but on stage yesterday you know we had the rocket mortgage uh executive team talking about how the rocket assist which they built on our platform has improved the conversion rate for refinance financing by 4x.

Uh we had the CTO of Wayfair on stage talking about how it's impacted the returns and warranty claims. We had the chief product officer of SiriusXM talking about the same.

And we launched this product called agent data platform, which I'm happy to talk about more, which is saying, hey, can we move the world of agents from conversational customer service agents into agents that can have long-term relationships, long-term memory, and actually drive sales?

And I think you know why were we able to have all these people show up yesterday is we're looking towards the future I think with a a bolder vision than a lot of our competitors in this space.

We have a really I think consultative approach with our customers to say we're not going to just throw software over the wall and hope you're successful. We're going to help you be successful this technology.

And just like our conversation about chat GBT is we're not going to predict the future perfectly but if you attach your cart to our horse so to speak we're at least going to be march in the right direction.

And I think a lot of companies right now are trying to figure out what they want to be when they grow up as it relates to AI agents. And I'm really proud that we're the partner for some of the best brands in the world. Yeah.

Uh I'm sure you saw or saw clips from the Carpathy interview on Dark Cash saying that it wasn't you know it wasn't the year of agents, it's the decade of agents. Did you what was your reaction to that?

Because I'm sure internally at Sierra it certainly feels like agents are having a moment but I think broadly agents as a category at least consumers have felt a like a little bit let down on on some of the product experiences to date on the consumer side right we've had we've been kind of promised these product experiences that aren't really good enough to use reliably yet but on your side you're obviously seeing companies get you know roll these out get you know tremendous results and so I feel like it's here it's just not so evenly distributed I think that's right.

I think both are kind of true at the same time. You know, there's there's basically AI agents and AGI.

And I think as you think about consumer agents, for those to really work, it's not it maybe isn't complete AGI, but it's close because the amount you need to generalize for an individual like the three of us, if you looked at the home screens of our phones, there's probably very little overlap, right?

to actually have an AI agent that can automate all of our lives, you end in sort of get the union of all the complexity of everyone's personal lives. In contrast, if you think about sort of, you know, Rocket Assist on on rocket.

com, you helping something go through the refinancing uh process, it's a relatively structured agent in the grand scheme of things. You know, it's and so I think where the value right now in the AI market is where you can turn science into engineering.

So, it's companies like Sierra helping with customer engagement, sales, customer service, or Harvey working in legal. And then the moment you say, "We're going to make a computer using agent that does everything for all people, that's a science problem.

" And so, I, you know, Andre Carvati is probably one of the smartest people I've ever met. So, I I think he's probably right in a lot of ways, but I think it'll be uneven.

I think we'll end up with AI agents that make really great math discoveries, but at the same time, you're going to have an agent that can barely use a computer effectively, and you're like, it's like somehow a genius and idiot at the same time. And that's the complexity of this technology.

But that's why I'm so bullish on the applied AI market. You know, I think I like to think of it as like we're taking this science and turning it into engineering. And, you know, I I'm really proud that we work in healthcare with like Sigma and R1, the revenue cycle management firm. We work with multinational banks.

We work in the mortgage industry and it's because we're taking all this science and we're whittling it down. Say here's what's shovel ready right now today. And I think the reason why we've had such traction in the market is that's really valuable. You know, there's a lot of science projects out there.

We have a term for it at Sarah called AI tourism. And we're like, no, AI tourism is not useful. It's just, you know, peacocking for your board that you have an AI strategy. What you want are the things that can actually drive sales and cost savings today. And that's what we're trying to deliver at Sierra.

How is the role of uh and and and responsibility set of customer experience teams changing already from the C companies that you guys are working with? I mean, historically, it'd be like maybe you start out as like an individual like CX rep or agent. You're like handling tickets.

Eventually, you're managing a team of of reps and then you're managing a team of managers that are managing reps and you just go up from there. Do do you see it evolving already uh or or is it or is it more you know an individual rep is managing tickets themselves that get elevated and then kind of overseeing the work.

How is that how is that kind of evolving today? One of the most fun and interesting things is with a couple of our customers the team that was managing their contact centers changed their job titles to AI architects.

And so you go from teams who essentially have operating roles managing teams of people to teams who are responsible both for those teams of people but also the AI agent that works with them uh in concert.

So some of the AI agents answering questions automatically uh you know some is actually managing the team that when the AI agent transfers goes to them. But it's a real organic system.

One of the other features we launched yesterday was called live assist and it's saying can you take the agent you made for your consumer and put it on the desktop of the people in your call center as well but have a really seamless conversation you know because there's not going to be things like AI is agents are great they're not great at everything but can you have like a single system of intelligence that helps through that whole process so it's dramatically changing the role particularly of the teams managing customer experience because it's gone from just managing people and processes to managing this piece piece of software.

And the thing I mean you all will understand this intuitively is when you take something analog like a phone call to a person and you make it digital with an AI agent all the past two decades of digital technology are available to you. You can run AB tests. You have real time analytics. It can self-improve.

Those are all things you know you can't run an AB I mean you could if you have a 12,000 person call center running an AB test is like a huge operational burden. But now you can do it with a click of a button.

And the interesting thing is it's it's stuff we've been doing on web pages for the past 20 years, but now you can do it on phone calls.

And so it's actually for our product managers like a lot of our inspiration is actually saying like what have we been doing in the world of digital technology for the past two decades that we can just bring to this previously analog channel which I think is a really exciting and oddly simplistic way of looking at the technology but it's kind of how we think about it.

It's awesome. Yeah. Do you care about Would it would it upset you if the models didn't improve at all for the next 5 years?

It would upset me particularly as a chair of OpenAI [laughter] but uh uh but you know I think I get where you're going with the question which is like how much breakthrough technology do we need for Sierra to make it applicable and I think right now actually the agents work exceptionally well.

I'll tell you two areas that I think could really improve a lot. native voicetovoice models can get a lot better. Um, they sound really interesting today, but a lot of them are prone to hallucination and, you know, aren't great at rules adherence.

Uh, uh, and I think right now a lot of applied AI, you might be in the world of speech to text and text to speech because you need better reasoning, better adherence. And I think that will change.

And I'm really excited about in general video and voice feel like we're still in the earlyish endings relative to you know the core reasoning models. Um and then the the other thing is I think I always bring this up. We're like in the 1997 era of making agents.

Like if you look I found this article for Sierra Summit about the uh creating websites in 1997 and there was this wired article. I'll send it to y'all. And it was basically about banks spending $23 million to add transactional support to their websites, like adding a login form basically.

And then you fast forward to like the late two 2010s and you know Kylie Jenner starts a multi-billion dollar cosmetics line with as I read seven full-time staff and you're like that's kind of so I think we're kind of in the 1997 year of building agents where it's still like way too hard and you end up putting a lot of engineering around what is a very intelligent you know set of models just to make it work well.

And I think, you know, I always imagine like what do you need to have a sevenperson team create a multi-billion dollar business on agents? And I think we just have a lot of product and technology work left to do. But for an applied company like ours, it's the models are actually pretty great right now.

I would be disappointed if they didn't get better, but I don't see like a major technical barrier to progress right now, which I think is really exciting. Um, you know, as a as a business person. Yep. Makes a lot of sense. Amazing to get uh get the update. Congrats on the first summit and uh come back on again soon.

We'll talk to you soon. Thanks. Have a great rest of your day. Before we bring in our next guest, let me tell you about Google AI Studio. The create an AI powered app faster than ever. Gemini understands the capabilities you need and automatically wires up the right models and APIs for you. Get started at AI.

studio/build. Uh our next guest is the CEO of Roblox, Dave Bazooki. How you doing, Dave? Good to see you guys. Good to see you again. See you again twice in one week. in one week. U we we didn't stream our last conversation. I'm very excited that we