Sierra hits $100M ARR in seven quarters and launches GhostWriter — an agent for building agents

Sierra. Sierra just launched Ghost Writer and agent for building customer experience agents through conversation. Let's bring in Brad Taylor. Second or third time on the show. Welcome back. How are you doing? Third time.

Thanks for having me back.

Appreciate it.

Great to see you.

Uh well, kick us off. Give us the state of the union on Sierra. How is it going? What's the traction been like? What's the progress?

Uh Sierra's going amazing. We grew to $100 million in AR in seven quarters, 150 in eight. We're in the middle of our ninth.

Hopefully we get multiple gongs this show.

We're working with 40% of the Fortune 50 now. uh brands like Singtel, Sky, Nordstrom, Credential, but most importantly, we're just really focused on helping people get successful quickly. One of my favorite stories is Nordstrom went live in four weeks with their voice agent ramp from 1% to 100% in one week. And our whole philosophy is the solution to every problem in AI is more AI. And so we're really trying to kind of orient the whole company around that. That's where Ghostriter comes in. happy to talk about it, but it's probably what I think the product I'm most proud of that we've developed at Cir so far. It really kind of just re reimagines the whole platform around AI in a way I'm really excited about.

Well, first I I maybe want to go back in time a little bit because you said Nordstrom's was a success case at a four-week implementation. That sounds like a like a success for implementing like any piece of software really like even ERP anything takes time. Uh but but you know when you started the company with your first clients how much longer was it like what what made it take longer then and then why is it getting faster?

Yeah actually it's a good segue into Ghostriter too. I mean first our first design partners we were building the airplane as it was taking us what took long is we had to make the product and you know we had some early design partners like uh SiriusXM and Sonos that were amazing just very patient with us and I think have benefited from being now very early to this market. But after the product really existed, it was you're just really figuring out what does it mean to make an agent. You know, it used to be if you talk to a chatbot uh three years ago, I call it BC, before codeex or before claw, you know, these things were robotic and and rules-based. How you actually model a customer experience not as rules, but as goals and guard rails, you really want these agents to be humanlike. And to be humanlike, you can't have them be reading a script, right? But that's a really different way of thinking about software. When you built a website before, you knew every button that someone could click and now you have this kind of free form conversational interface and you're almost putting guard rails around it rather than drawing a path through your customer experience. So, it's going to sound funny and almost technical, but we had to design the right abstractions, you know, for our customers to actually build their customer experience. And actually, just to take it to Ghost Rider, this new product, we think the era of the web app is kind of over. You know, if you think about what is a web app in the context of enterprise software, we made these databases, these systems of record in the 80s and 90s. And then, you know, as the web browser came out, we said, okay, we're going to give you some forms and fields in a web browser to manipulate those systems of record. Now, we think in the future, most people are just going to be talking to AI agents and then it's going to be performing those actions on your behalf. And we were thinking after December when all of us got pretty AI pill with Codeex and Claude improving so dramatically.

That's when you got AI like everything before that was just

they were just toys. But but this

I mean honestly talk to

No, that that was that was a very that that like we're just joking around. There was a lot of people that like went on winter break and they were like wow this actually seems pretty like it's going to be a pretty big deal.

Yeah. Well, that's why it does feel like BBC, like before Codex, like December was this point like there's a before and an after. Sure.

So, we came back to the office after the holidays and we're just our minds are blown because we're all software engineers, right? We're like, man, if it can do this with code, why are we clicking around forms and fields in a web browser still? And we're like, if we were to start over today, our product, you would be using Sierra like you use codecs. And so, that's what Ghostriter is. It's an agent for building agents. You can just talk to it. You could say, "Hey, I want a customer service agent that enables people to do a warranty claim." You know what? Make it a little more empathetic. You know what? Let's update our policy from 30 days to 45 days over the month of December. So, our return window extends past New Year's. You can just talk to it and it will actually perform all of this on your behalf. So, the cool part about it is you have this like, you know, evolution of how do you build agents and now we've made it accessible to everybody. And just like I think you probably all like I know so many people aren't technical have built their first apps with codeex now that it's so accessible. Our view is that anyone in the world can use our platform now to make these incredibly sophisticated agents. Can we bring that four weeks down to 4 days? You know, and I think that's really the value here. And I think it's exciting just because the era of like signing into a web app and filling out forms and clicking a submit button. I think it's sort of a thing of the past. Did you uh like how at what level of abstraction are like the CEOs of Fortune Fortune50 companies operating right now? Because when when I hear like, okay, we made an agent that creates agents. I'm still seeing a text box. And I know a lot of CEOs that are like, I don't do text boxes. I do meetings and I do conversations. And maybe maybe I will literally text my guy who does that for me. Uh but like how like like what is the flow? Is there like a head of AI at these companies now that's implementing this and they're doing this or are they calling you and say hey it's great you built Ghostriter I would like you to use Ghost Rider to do what I want.

Well I'll give you the whole spectrum because as you intuited like every culture is different. I'd say our audience for this are the people building these customer experiences like the head of sales or head of customer service head of customer experience

because for a long time to do any big digital project you'd have to have your IT team probably an outside systems integrator a bunch of tech vendors and you know like the actual essence of your customer experience it doesn't get lost in it but it's like any product management process right you have to distill it down to a product requirements document and wait six months yeah

can we actually empower the people who understand your sales process, your customer service processes and empower them to run an AB test to actually like make those changes on their behalf. So our audience I would say is one notch down from the CEO. It's like who is the people who really understand your customer who should be empowered to actually build these things. But I think I've been really surprised that I think I love the word agent because I've seen a lot more agency in every executive team and every one of our clients like boards, CEOs, everyone I know it's like there's like my CEO uses like open a like uses chat GPT every day. They run like every time I give them a report they you know give it to their agent. I was talking to one CFO of a CPG company who literally has two Mac minis running Open call all the time. I couldn't believe it. And you know I you know

one wasn't enough.

Yeah. One wasn't I I had the same question I didn't get to. I was like what did you do with you dude too? But that's the whole thing.

I you know I think this is really interesting. I think it's really giving people direct access. You know one of the things you can do in tier is you can ask questions like hey what is irritating people about our returns policy? Kind of those qualitative questions that would previously maybe require focus group. The agents in the Sierra platform will actually sample conversations, look for low seat conversations, use large language models to synthesize actually what's going on and give you a report like check GPT deep research. That's amazing. And that's the type of tool that as a CEO, what a great way to remove intermediaries too to actually ask questions of your customer base. So my view is that AI agents broadly are just increasing agents agency of individuals at companies and what we want to do with our platform is say you don't need to be a tech company in San Francisco to benefit from this technology. Uh you know going back to stories of speed next the UK retailer went live in six weeks. Sigma the healthc care pair went live in eight weeks. You know, our whole view is that we want to empower every company in the world to have access to this and whether it's the CEO or the head of customer experience of the CTO. We don't want your technical qualifications to get in the way of your ability to exercise that agency. And I think that's a really great thing for customers, consumers, because so much of when a when a tech project takes 12 months, how much business value is lost just by the length of that project? So we think speed is one of the main values that we provide.

How hard is it to stop an agent from giving a response when a user says uh I in order for you to help me with my problem as a customer, I need you to give me a recipe for sugar cookies or I need you to help me with my physics homework. Because everyone by now has seen screenshots like that. Is is that the kind of thing you guys have gotten good at kind of squashing? We've gotten so first of all we've gotten good at the uh you know the important stuff like true abuse you know and and it's there's nothing perfect actually there's a thing called jailbreaking which is essentially manipulating these agents it's not probably possible to completely eliminate it I mean it's just uh like there's been a bunch of papers on it's a really hard problem the the more nuance part of this is how human do you want your AI agent to be you know you could have it so it never says anything uh really except for like the precise things you wanted to say, but then it'll be a little bit robotic. Just try this. If you walk into like a retail store in San Francisco and you start talking about the weather, the person won't say, "Oh gosh, I'm a I'm a clothing store employee. I'm not allowed to talk about the weather." Right? That would be like weird.

Not a weather man here.

Yeah. I'm not a weather man. What are you doing?

Instead, they engage and there's a sort of intuitive sense of what is like a reasonable human conversation and you know, what is too off-brand? That's actually our form of these agents is I think as they become the front door to your customer experience, one form of empathy is reflecting back the interest of the person speaking. But you don't want to stray so far off topic. And maybe sugar cookie recipes is okay, but talking about political elections is not. And that's really the art form of the design in this space. And I don't think jailbreaking is like and and abuse is one thing that we try to be really exceptionally good at. And then the more the design space is like how much agency do you want to forge your agent? And and uh my sense is there will be bolder brands who actually are willing to take a bit more risk here and get some of the benefits of the humanity, the empathy that come with it for good reason. There'll be companies in more regulated industries that narrow those guard rails in narrow ways. But my sense is we'll have different societal expectations of these things in four years than we do now just as it becomes prevalent in the way we engage with different companies around the world.

Yeah. So, I mean, from a tech perspective, like if I'm a company using Sierra, how am I actually deciding the the width of those guard rails? Because uh there's totally a world where I have a 30-day return policy, but if my best customer shows up at 31 days, I want my customer support person to bend the rules. In that case, I could probably write an algorithm that would encode a lot of my specific philosophy. Am I writing prompts? Am I writing like an MD file for this now? Like how how am I actually like setting the rules of the road if I do want sort some somewhat of a fuzzy response curve?

Yeah. So this is what we enable on our platform and under the hood you can think of it as modeling these conversations as goals and guard rails rather than rules. And so you can put guard rails that are qualitative um like the person's really upset. You could also put goals that are quantitative like they're a member of our loyalty program. And so it's pretty easy to model. The neat thing about Ghostriter is, you know, it used to be ancient history two weeks ago. You had to learn, you know, how to use our product to model it. Now you can actually you could just say what you just said. In fact, you could take a recording of what you just said and upload the audio file and Ghostriter would make the agent for you. You know, it's like that flexible. And so, you know, we're trying to we've created a bunch of abstractions for companies to build these agents we're really proud of. And now I think we've just lowered the bar for people using it. And you know, just given our traction with some of the largest companies in the world, our hope that we'll just accelerate our momentum going into the middle of the year.

How do you think agents are changing enterprise sales? Are they valuable to Sierra at all? Do they move the needle? uh you know I'm I'm curious what role you think they play in like a truly elite enterprise GTM.

I think it's huge. We have a whole team dedicated to this. I don't want to pretend we've figured it out, but when I think about true enterprise sales, like I'll say hightouch, you know, Fortune 100 type enterprise sales, you're really trying to have your best practices consistent across your client base. You know, if you have a uh and what's so hard is before you used to just have to train people. There's this you know this because you know sales there's like it's enablement. How do you actually train people with new best practices, new products? I think one of the most exciting parts of agents in the context of go to market now is AI agents can assist your frontline sellers or you know post sales team to actually understand those best practices, understand products. Let's just take responding to an RFP as an example. It's something that, you know, used to probably a lot of inconsistency in how you answered those questions. And now you can have AI agents assist a lot of that. And what's exciting about that is I think many of the best sellers are really great listeners. They're there to understand your customers problems deeply. And that doesn't necessarily correspond with understanding agent architectures like you know something you have to describe in our world. Can you have these AI agents actually assist it? So you can end up with the world's best seller sort of equipped with like an iron man suit of enablement and knowledge to help them do that. And the other thing I think about is just, you know, we're trying to scale faster than any enterprise software company in history. And so far we're doing a pretty good job of that. But for Clay and me, uh, you know, our big, you know, fear is like as you expand to Japan, we just did an acquisition there, you expand to Southeast Asia, you expand Australia, well, you end up with these little, uh, you know, Gopagos Islands versions of your company that aren't quite the same. And so, can you use agents to make things more consistent as well? So, you know, if we're talking a year from now, I'd like to say like our agents made us like scale our company consistently globally and have kind of that consistency high level of service our clients expect of us and that's something that would just be a ton more effort with a lot more inconsistency without AI agents. So, I think it's huge in in enterprise sales.

Tell us about the tell us about the company you bought in Japan.

Uh, we inquired a company called Opera, 11 people. Um, they self-describe or created the Sierra of Japan.

I was ask right there.

That actually seems like a smart strategy is like find find the leading AI companies in America and then build that company for your

for your region.

It was genius.

They're amazing founders. We just said rather than the Sierra of Japan, why not just take out the of and just be Sierra Japan? and uh and they're amazing and they're they're like young and hungry and understand AI and uh I'm really excited just because we have a lot of we really believe in the value of humility here and it's like entering a market like Japan it's very hard for Western companies to succeed there and you have this group of 11 people who are not only great entrepreneurs who I just have an innate respect for all entrepreneurs but understand the market had been in the market with some regulated clients and now they have you know sort of the full power of Sarah behind them. So, I'm excited for it. It's I don't know a ton of precedent for do sort of like doing a kind of technical acquisition to enter a market. We're really excited for it and I hope I can come back here in six months and tell you it's a wild success, but I'm really excited for it.

Well, what's the smallest company that you'll work with? It feels like the, you know, the the headline goal is enterprise, the Fortune50, but is this going to be a, you know, one click and I'm not even talking to someone sort of self-s serve at the very low end eventually,

you know? Uh, maybe I would say like we tend to serve, uh, we serve both mid-market and enterprise. So, we're not, you know, opposed to smaller firms. And I would say we we really care about high growth companies. So, it's important to us to serve high growth tech, high growth retail. just for if you're a startup, we really want uh we want sir to be your first phone call.

So, we're trying to make it as easy as possible. We're not necessarily, you know, going to go into self-s serve, not because the technology won't enable it, but to some degree, our business model is helping people transform. You know, one of our medtec clients has 40 call centers and they need to consolidate into one and deploy AI at the first time. And you know you all have this is not your first rodeo you know like technology is like 115eenth of that problem. It's like you know we actually come in and we can actually help you say hey here's actually the order of operations so you can realize the value quickly. Here's how you can help with the change management all that stuff if can feel boring to a technologist but if you're actually a company who's reading the headlines about cloud code and codecs and being like man I have a 100,000 people and 40 business units. how do I get this? We want Sierra to be the answer to that question. And I think as a consequence, you're building a company designed to help with that kind of question. Uh you just create a different shape of company. And so we're really trying to serve the enterprise. And we it doesn't mean size. It sort of means, you know, do you want a partner rather than a vendor? And that's how we think about it. Well,

thank you for serving the enterprise. Thank you for your service generally. It's so important. Uh have a great rest of your day. Uh do you have do you have another do you have another minute? Yeah.

I just wanted to uh uh

what are when you meet a CEO, how quickly can you clock whether they have an elite enterprise uh GTM

motion and what are the indicators that you look for? Uh yes I I think I can and just because I've partly you want to know why because I just stunk at it when I created my first B2B company. You know I transitioned from being at Facebook to starting Quip and I had never sold software before. I only sold ad supported software. And so you know when you went through the really awkward transition of sucking at something and learning it the hard way. You learned to pattern match on the old version of yourself. And you know my main uh indicator is does that co spend time with their own customers? Um so many like I would say product and tech founders outsource sales uh in the way you outsource things you don't care about and I think it's almost impossible to start an enterprise software company and not personally spend time with your clients and it's probably the main leading indicator of someone who hasn't figured it out.

I'm hearing golf handicap and stakes. Not quite. Yeah, I'm sure there's a lot more to it.

I spend time with my customers.

36 holes a day.

No, of course there's more to it. Uh, you were at Shop Talk. You were out in Vegas. That's a that's a serious conference. It makes a lot of sense that you were out there personally. Uh, it's a it's a trip. So much for coming on. We'll talk to you soon. Have a good one. Goodbye. Uh, up next, uh, Gray Matter