Will Griffith (00:00):
At ICONIQ, we've known Bret for more than a decade going back to when he was the CTO of Facebook, and we partnered together in 2024. And at that time, what we wrote and believed was that every company would need its own customer-facing agents, and that Bret and Clay were uniquely suited to define that new paradigm.
(00:22):
Today, agents built on Sierra now resolve and have conversations with more than 40% of the Fortune 50 companies. And in July, Sierra launched Horizon, designed for agents to pursue outcome over weeks, not just resolve a single conversation. Bret, we have a year to catch up on. Thank you again for being here. We really appreciate it.
Bret Taylor (00:46):
Thanks for having me.
Will Griffith (00:47):
Bret, you sat on the stage 12 months ago with Matt and you guys talked about what it would take to really build an enduring AI application company as foundation models continue to improve. Can you tell us what you believed then that you no longer believe is true today?
Bret Taylor (01:06):
It's interesting that that was only a year ago. AI moves really quickly. I think at the time, if you go back a year ago, there was a MIT study talking about something like 90% of AI projects fail. In my circles, it got a lot of play. Every board and CEO would send it down to their CTO and be like, "What's going on?"
(01:32):
What's interesting about that is since then you had the explosion of Claude Code and Codex, and we've gone from we don't know what to do with AI to tokenmaxxing and, "What the (censored) is going on with my token bill?" So it's really interesting. It was almost like a complete flip for the problems. And the value proposition that I was thinking about for applied AI conceptually, just beyond customer experience, sales and service, was to help people successfully deploy projects. But now you can sort of see this other value proposition, which is people building AI solutions from component parts from these tokens is leading to all these other problems, which is now people are overspending on intelligence to get the solution to their problem.
(02:16):
I think all of it is, even though it's the literal opposite problem, I think all of it is a demand for more to applied AI solutions. If you look at the few areas where there are successful applications of applied AI, software engineering, customer experience and legal, where there are mature vendors in legal, you have Harvey and Legora. And customer experience, you have us, you don't need to consider anyone else.
Will Griffith (02:44):
We feel the same way about Legora just to clarify.
Bret Taylor (02:46):
Yes, fair enough. I won't say any other. And then coding ... It's the areas where you tend not to have these concerns about what value am I getting because they tend to be off-the-shelf solutions that you can deploy quickly. They tend not to be priced by tokens. They tend to be priced towards more business outcomes that you understand out of value. And it leaves the burden of optimizing token spend to a software vendor who can amortize that R&D cost among all of their partners. And it leaves a company to figure out where to deploy that solution to their problems.
(03:22):
My contrarian take is because I think there's in San Francisco circles, people are like, "Oh, there only two software companies named Anthropic and OpenAI." I just don't see it. And in fact, if anything, my conviction is even deeper now. And I actually think, where are the purpose-built solutions for your CFO? Where are the purpose built? Every single department should have a vendor helping them deploy agents that deliver their business outcomes like auditing their financials. And we don't have any of those companies yet. So I'm incredibly bullish on applied AI. I think the concerns about POCs failing, tokenmaxxings are symptoms of a immature applied AI market.
Will Griffith (04:01):
Let's press on that for a minute. Why do you think those verticals that have been successful have developed and evolved to have structure and really those market leaders that are making it more digestible for enterprises vis-à-vis the ... I mean, off the CFO, great example. These other areas where it's a little bit disorganized and you have the foundation models potentially looking to do forward-deployed engineers and build companies that are going to actually try and address them directly themselves?
Bret Taylor (04:29):
I think the answer is process formality and I'll double click on that, but I think the atomic unit of an AI agent is a process, not a job or a person. And what I mean by that, just take a bank, KYC, AML, know your customer, anti-money laundering. If you're onboarding to a financial services institution, you go through a pretty formal process to ensure that you are who you say you are and that the bank is allowed to do business with you.
(04:59):
If you imagine giving a bank ChatGPT work to every employee and you fast-forward two years and say, "Hey, how digitized and agentified is there a KYC process?" You probably wouldn't see a lot of movement because it turns out it's a collaboration between a business team that's onboarding customers, a compliance team of technology. All these teams need to work together and you need a top-down view of the process to actually apply an AI agent to fully digitize it.
(05:28):
If you look at the few areas of software engineering, customer service in particular, and legal, those happen to be areas where the processes were extremely formal. Engineers, we've been, as we do it as a hobby, we've been trying to act like robots for 20 years and now we finally can. You have everything from unit tests to code reviews to these processes that are almost like a glove fit. The context of the code base is available to the agent.
(05:54):
Because of the offshoring phenomenon, customer service, the standard operating procedures or SOPs that the industry have also been formalized for a third party. And now, you're just taking a third party rather than being a BPO as an AI platform. And then with legal, similarly, just because of the natural formalism around legal processes, think of an antitrust process, you do the same process over and over again for a ton of companies. So it's just a glove fit for an AI agent.
(06:24):
If you look at other parts of your business, the processes tend to be less formal or at least certainly not formalized to be used for an outside law firm or a BPO. And so you end up where there's sort of this process of maturity and as a consequence, you end up with a lot more change management and industry's peak required to deploy AI at scale.
(06:45):
I think the companies, the AI native companies of the future will have a very strong opinionated view of the processes that drive their business, customer acquisition, customer acquisition costs, onboarding new ... If you think about a vendor, a marketplace like Uber, you have onboarding merchants, onboarding drivers, you have the consumers. And the more you formalize those processes, the more you can apply AI agents to do it better, faster, cheaper. And I think that is the key barrier for most incumbents, i.e., all the biggest companies in the world is you tend not to have those formalized processes in most parts of your business.
Will Griffith (07:23):
I know you have a day job, but it sounds like you should set up an incubator and really help solve this. But shifting to a little bit the cost equation, we saw in Tuhin's presentation recently, just Baseten, just the growth of what's happening with open source. And you have a unique perspective also as the chairman of OpenAI. You guys use open source for certain tasks and have been employing that and have been talking about that more frequently. Can you talk about what and where you're using open source today and how you see the balance between the use of open source and frontier models evolving for Sierra going forward?
Bret Taylor (07:58):
Yeah. Sierra, we use a ton of open weights models and fine tune a lot of them probably as a measure, just percentage of our inference, it might be 80% or so. Or at least I'll say 80% is at least not frontier scale parameter counts. I would say there's three reasons why one uses open weights models. One is if you want to fine tune the model, which particularly for applied AI firms like ours is important. We do billions of phone calls and chats and you don't need to have a frontier intelligence to understand is that a car horn or someone talking, which turns out to be an important thing in my line of business. And so, fine-tuning a smaller parameter count model is the way you achieve that with price performance quality.
(08:45):
Number two is cost. So for a long time, if you're using the true frontier from Anthropic and OpenAI, it could be quite costly and some of these open-weight models were less expensive. And number three is sovereignty, sort of the Alex Karp rant. If you haven't seen the rant, it's quite entertaining in my opinion, just this idea of, are you seeding some strategic value to the models?
(09:09):
The last one I think you can largely solve with a contract. In my opinion, I think we're moving towards a world where most companies will rent their intelligence but own the context. I don't think most companies can or even should try to differentiate on raw intelligence.
(09:26):
On cost, I think we're seeing some really interesting evolutions. If you haven't used GPT-5.6 Luna, you should. I think it's probably cheaper all in than most open weights models for most companies. Put another way, the frontier labs can distill their own models and rent them out. It's not obvious to me why you're hosting your own model should be cheaper than that. There's not really a scientific reason why that should be.
(09:54):
And then fine-tuning is probably a real thing. My intuition is it's probably more of a thing for applied AI companies than most industrial firms. The moment you fine tune a model, it's like exiting the highway and the AI momentum is continuing. So it's like forking the Linux kernel or something. There might be a good reason to do it, but in general, you really should be a sort of a domain expert if you're going down that path.
(10:26):
I think open-weight models will be a thing. I think I'm just first principles economics around it and I'm really happy that OpenAI has made more options for different price performance parts of it's like a matrix. I think that certainly as the chairman of OpenAI, I just think we should be able to provide models at the right price and quality that you want at a lower cost or certainly better token efficiency than most open-weight models. But I do think there's an importance in open-weight model, particularly for fine-tuning.
(11:00):
So I think the classic ebb and flow of the conversation in Silicon Valley tends to be overly simplistic and I think probably compute scale is the most important thing in that supply chain that doesn't get talked about a lot enough in my opinion. Because at the end of the day, inference is inference and whether it's open weight or closed almost matters zero when you're thinking about the cost of inference. And so I think that's really where I think most of the attention should be.
Will Griffith (11:29):
Do you think the foundation models have an advantage in terms of a little bit what Nikesh was talking about, the ability to really manage the harness, have the memory, have the context between frontier and some of the legacy models they should be able to trade between relative to just shifting to open weight? How do you think that plays out?
Bret Taylor (11:47):
I think it really depends on the application. I think a lot of people conflate intelligence with ChatGPT Work or Claude, like a desktop app for an information worker. And in those contexts, there's a very important market by the way, it's gigantic, so I don't mean to be dismissive of it. In that context, you think a lot about model routing and tokenomics because someone collating in a set of emails for a dinner party has very different economic value than a KYC process for a small business to a bank, but you're using the same token count for both of those things as you end up with just a lot of attention on that, which makes sense. It's just like the ChatGPT moment. This is the moment we're in now.
(12:35):
I think if you buy what I'm selling on Applied AI, it would sort of disappoint me if 5 or 10 years from now. If you just think about the AI agent that's auditing your financials after quarter close before you report earnings, I don't think such an agent exists right now. I think it probably should and will exist. If you just think about how much you pay your auditor every year, agents could do that.
(12:58):
My sense is there will be quite a few fine-tuned models that are sort of purpose built for aspects of that task where the vendor there will be ... Their margins are essentially determined by how much they can maintain their quality and bring down that token spend. And there, I don't think there's an advantage to say the frontier lab ... It's very, you could call it esoteric unless you're an accountant in case it's really important, but I would say that vertical integration of the application with the model is where a lot of the value will come.
(13:31):
So I think you're going to see for these broad-based software engineering information work use cases, I do think the frontier labs have structural advantages. For industrial vertical applications, I'm a huge believer in the power of verticals in this and the ability to vertically integrate and get price performance that you couldn't otherwise get.
Will Griffith (13:54):
Makes sense. Makes a lot of sense. Let's shift a little bit to existing non-native AI software businesses. You have a very unique perspective as co-CEO, former co-CEO of Salesforce before you started Sierra. How do you see and what advice would you have for legacy software businesses and how they're going to evolve in this new AI world?
Bret Taylor (14:15):
I'll start with my perspective, which you may or may not agree with just so when I give an answer, you can determine whether to dismiss it or not. I think the difference between AI agents and traditional software is that AI agents are actually completing a task rather than simply being a productivity enhancer. And as a consequence of that, I think there's an opportunity to do outcomes-based pricing, sort of paying an agent on commission, if you will, for a job well done.
(14:47):
I think we've seen parallels to this in the past when we went from impression-based ads to cost per click ads. And now for a lot of the mobile app ecosystem, you do pay per install or pay per conversion. The more you can align your business model towards an accountable, measurable outcome, the more of your margin you're willing to spend on that outcome because it's more measurable and easier to track.
(15:14):
And I think agents that work in domains where there is a quantifiable, measurable outcome, I believe the market will just move towards outcomes based because vendors who do it will find buyers more receptive to spend money on it. Because it's just easier to pay for a loan originated or a client onboarded or a customer service call resolves than it is to think about the tokenomics of it all. That's my perspective. That's something we've obviously pioneered at Sierra and all of the outcomes I just said are outcomes we get paid for today as an example.
(15:51):
So going back to your question around incumbent or as you call them legacy software companies, I think the challenge will be kind of threefold. One is you have to transform your technology. So you're going to make agents to perform tasks where you previously made a web app where people did those tasks. That's a different type of product than you were making before. So you need to sort of reimagine your product around agent-based workflows rather than forms and fields, which is sort of the traditional software as a service application.
(16:26):
Number two is you need to think about different business models. So you're going from a seat-based business model to something else. Again, I'm a strong advocate for outcomes in that.
(16:38):
And number three is you need to make that transition while most of these companies are public companies, which is a nightmare. It's just really hard because public market investors are just incredibly impatient and they want to see everything.
(16:56):
They want your organic business growing and your new thing growing and all these other things. There's just not a lot of companies who have made that transition. Obviously, Microsoft probably being the most famous with Windows to Azure, Shantanu going to ratable revenue, being maybe a smaller version of that at Adobe. But it's incredibly hard and I think doing all three at the same time, there's going to be the Satya of this new era, but I think there's going to be a lot of others just because it's both.
(17:27):
And I think that's why I've said this a lot publicly and I believe it is ... I've sort of now apparently the old guard in enterprise software, but you tend to see a pendulum swing from best of breed solutions towards platforms. So when there is a new technology like the internet, you see people go from Oracle to Salesforce because Salesforce was native to the web and what is now called software as a service. And then same with mobile.
(17:57):
And then as the technology becomes commonplace, it's not hard to make an agent anymore and you can get agents from everybody, then you tend to see a lot of consolidation where the native companies that grew to scale will become new incumbents, the rest will get consolidated. And then you'll see a lot of CTOs, CIOs, CEOs say, "Let's simplify our vendor landscape," and you tend to get that sort of best of platform consolidation. We're just solidly in best of breed right now just because you can't really buy great solutions from most of the incumbent companies, but it is essentially like we've been through this many times, it's sort of a race to see which incumbents can upgrade their platforms and business model. It's a race to see which of the AI native companies can earn the right to sort of become an incumbent, if you will, in the future.
(18:50):
I certainly hope Sierra's on that latter list, but it's interesting to see in these different domains that will happen, but the consolidation will happen. It's just a matter of when. Is it a year from now? Is it two years from now? But it's just a natural ebb and flow that I think will just repeat itself.
Will Griffith (19:09):
We love where you are on that frontier. You're really defining it very confident that you're going to be one of these winners. Let's jump ahead a little bit to outcome-based pricing. Talked about the analogy in the advertising world from impressions to clicks to CPA. Makes sense in terms of really viewing agents as performing a process and task and getting paid for that. Has that been hard for enterprises that traditionally played for software and workflows, but not for outcomes? How hard has it been for them to adopt that mentality? Have you been having to champion that, and how well has that been received?
Bret Taylor (19:46):
Sometimes I see a few clients in here can keep me honest on this, but broadly, I think people are fairly receptive. I'll just tell you our model, which I think is intertwined with that, we almost always start with a live proof of concept, which mean by live, I mean it's actually talking to customers. Maybe it's onboarding a merchant on your marketplace, maybe it's doing a first notice of loss for a insurance company. So you can actually measure the performance of the agent in production so that by the time you're deciding, "What does this mean? How many outcomes am I going to buy next year?" You can actually extrapolate it in a spreadsheet rather than measure it.
(20:27):
So broadly, this is extremely reductive, but if you look at the history of the software industry, for every $1 most companies spend on software licensing costs, they spend $4 on implementation. And that's been true for a really long time. And I heard that from someone said trust, but verify. I was trying to ... I have no idea where the stack came from, but it was interesting when I said it, I saw a lot of people nod like, "It's basically right." I was like, people are like, "I don't know what's right, but it's basically right."
(20:53):
And so you end up with this interesting thing where you buy a piece of software and then you implement it. And in that period, you bought it already and you're not using it, you're not getting any value from it. And then you spend millions of dollars making it work for your business.
(21:07):
And then by the time you got to the part where it's live, you probably have a different account team, you probably have all these things are changed and people forgot why they did it in the first place often. And you go back to some PowerPoint deck with an ROI calculator and you're like, "Wait, what? What happened here?"
(21:25):
And so, it's sort of like the dirty little secret of the enterprise software industry that there's a stark lack of accountability, I think, for delivering on the promises of these enterprise software sales cycles. And so, outcomes-based pricing while novel, it just has so many second order effects.
(21:49):
First, most of our company is focused on post-sales, like the delivery, not sales, because we don't get any value until it's actually delivered in production. I think that's really great. I think it's something that is a positive change for the enterprise software company.
(22:08):
Second is, if you think about what happens a year after you've partnered with us, that quarterly business review, we're all focused on the same metrics. We're trying to drive the same outcomes that you are. And so, it actually creates this enduring partnership that feels more always on, relatively speaking, that I think is really different. So I always like to say we are a partner, not a vendor. And we don't always meet that bar, but that's what we want to be.
(22:35):
And that's why I love business model shifts because they actually force you to change everything about your company to meet the requirements of that new business model. So I'm an outcomes evangelist on this front and we'll sort of make anything work if it's not appropriate for a use case, but I think it is a positive step forward for the industry.
Will Griffith (22:57):
It's incredible. I think the alignment there is really profound and hopefully simplifies a lot of things. You've launched Horizon, which enables outcomes to really transpire over weeks as opposed to single conversation. Was that targeted any particular verticals and that continue to just open up the market? Are there markets that were necessary for that to happen in terms of just as opposed to a quick interaction?
Bret Taylor (23:20):
Yeah, I'll start with where it came from and then it definitely was a few industries. So we at Sierra, we've always tried to imagine our company as going beyond traditional service to helping with just broadly being the digital front door for your business. And the digital front door for your business involves a lot more than service.
(23:39):
We ended up working with a few companies like Rocket Mortgage as an example, doing things like loan origination, which were distinctly not customer service. It wasn't loan servicing, it was on the origination side and really learning about what that takes. And if you go to rocket.com, you can use Rocket Assist and do a big chunk of the mortgage origination process because it's sort of an expensive thing to try. Buy a house, no problem, but it's really neat and we're just learning about it.
(24:06):
But then, when we zoomed out from that and a few use cases in healthcare, like getting a referral from a primary care physician to a specialist and finding out that most of those referrals get dropped and if you're a healthcare system that's lost revenue and bad outcomes for the patients as well, and you think about, "Okay, AI should be able to help with this, but how?"
(24:29):
You realize that the one conversation in there is just one part of a broader process. And in fact, it's not a well-defined process. If you think about just the primary care physician to a specialist, so let's say you run a big healthcare system and most of your physicians are in your network except for the orthopedic surgeons. This happens to be a true story of one of our clients. So if you're referring someone to an orthopedic specialist, it's not in your Epic instance. You can't actually make the appointment, you actually have to call them on the phone. So you actually rely on the patient to bear that burden and you end up with all this lost revenue in the way.
(25:05):
So the AI agent that we've developed with them actually will call the primary care physician, call and navigate the IVR of the specialist, and call the patient and get that thing scheduled, and it will take as much time as it takes. It might have to call the patient multiple times, it might have to call the provider multiple times, but that long horizon agent involves multiple people. It involves an undefined amount of time, and all it's doing is trying to pursue a goal within guardrails that you set.
(25:35):
And if you think about onboarding a customer, onboarding a supplier in a marketplace, if you think about subscription churn management for subscription businesses, if you think about just any care coordination use cases in healthcare, if you think about loan origination, digital claims adjuster, digital loan officer, these are all the use cases that we're working on here.
(26:02):
And I'm just really excited about it because fundamentally, this brings your digital front door from simply lowering the cost of customer service to doing something that can lower your customer acquisition costs and actually drive top-line revenue growth and hopefully really drive true outcomes. I mean, that's why I love the healthcare use cases because we spend 18% of GDP in this country on healthcare.
(26:24):
If you think about prescription drug adherence, care coordination, all these things, there's a possibility to have a digital one-to-one concierge for everyone in this country. I think that's just going to have a huge impact on some of these fundamental drivers of, I would say, the bad kind of costs in our economy.
Will Griffith (26:43):
Love it. It's definitely a massive TAM expander and use case expander, so that's going to be exciting on a lot of different levels. There's a lot of founders in the room, and I think consistently a big question is, will the foundation models continue to advance to such a degree that their focus will become a feature and non-durable? You've talked about really renting the intelligence, but owning the process, owning the relationship. What perspective and advice would you have for the founders in terms of what enables durability in the ever evolving world of AI?
Bret Taylor (27:21):
The answer is I honestly don't know. Even as chairman of OpenAI and running in Sierra, it's very hard to predict the future, but it is telling to me just a few things that smart people in San Francisco have said over the past few years have just turned out to be completely false. So one, if we automate some of the AI, all the jobs will go away. There are more software engineering job postings today than there were three years ago. I don't know why, but it was very smart to say otherwise.
(27:51):
In fact, I think Dario did on TV. And only reason I say that is not to poke at him, but just it turns out the economy is very complicated and the idea that you make a digital agent over here and all of a sudden it's just like there's no second order effect is laughable. I'm a card carrying capitalist too, which is maybe not as popular as it once used to be, but it turns out that everyone in the world has the same access to the same intelligence. So the idea that people can just automate something and pass on profits to shareholders is also laughable because everyone has access to the same technology. So the second order effects are as I think where all the interesting things are.
(28:34):
Second thing is I would say is the idea that the frontier labs will do everything. First of all, I just don't believe that's the way the economy will work. There's just so many antibodies that will fight that. But I think the way I would think about it is I think if you look at ChatGPT as a consumer footprint and you look at what Claude pioneered with the sort of desktop footprint, and I would argue ChatGPT work is much better if you're considering there. Look, I'm a single issue voter.
Will Griffith (29:09):
I get it. I get it.
Bret Taylor (29:12):
I would say if your product is directly adjacent to either of those, you're flying very close to the sun. So if you think about the application for information workers established on their desktop, you have to ask yourself, how are you going to compete with the mind share of things like Claude and ChatGPT Work? If you're making a consumer utility, how will you compete with the mind share of ChatGPT?
(29:38):
I don't mean to dismiss Gemini or all the others. I just mean you're flying very, very close to the sun. The more that you're selling business value to non-developers and not trying to just make an information worker more productive, I think the more credibility there is to the durability of your business where you'll have expertise around the process. You'll have a go-to-market motion that is particular to that buyer.
(30:06):
And then as you're thinking about helping your clients differentiate by building context, as I said, rent the intelligence, own the context. I think there's a credible argument that you can really create a virtuous cycle in your vertical and in your domain. So that's the way I think about it, but there's a lot of uncertainty.
(30:24):
The only thing I'd say is if you think of all the reasons why your business might die right now, you'll never start a business just because you can make a credible case. The tail risk is insane for everything right now. And so part of me is like, there's a little bit like the Paul Graham, like, "Make something people want." I do think there's an important thing, just spending time with your customers and really understanding their business problems and solving them.
(30:51):
And I know that sounds really simplistic and silly, but you can get really wrapped up in the strategery of, "What about this and what about that?" And it's like, well, you'll never actually make forward progress. And I do believe if you follow customer value, durability is nuanced. And I think the companies that compound relentless focus on their customers are more likely to succeed than the people who think about memos written on Twitter.
Will Griffith (31:21):
Bret, thank you. Really appreciate everything. Thank you for the partnership. You're continuing to define the frontier. We couldn't be more impressed and we really appreciate you sharing your perspectives today. Everyone, please join me in thanking Bret for his time and his insights.
Bret Taylor (31:35):
Thank you.
Disclaimer:
This presentation does not constitute investment advice or an offer to sell or a solicitation of an offer to buy any securities (including, without limitation, any investment fund or investment product referenced herein) which will only be made pursuant to definitive offering documents and subscription agreements.
Certain statements provided in the presentation are made by Bret Taylor, Co-Founder of Sierra and Chairman of the Board of OpenAI, both of which are ICONIQ portfolio companies[KK1] . Mr. Taylor’s views are necessarily indicative of the views of other ICONIQ portfolio companies or their executives, whose experiences may differ from those expressed in this presentation.
Mr. Taylor is a client of ICONIQ. ICONIQ did not compensate Sierra, OpenAI or Mr. Taylor directly or indirectly for participation in this presentation, but given his role as a Co-Founder and a Board member of ICONIQ portfolio companies, as applicable, and as a client of ICONIQ, Mr. Taylor may be subject to certain conflicts of interest. ICONIQ holds an ownership interests in Sierra and OpenAI and maintains an ongoing relationship with the companies. As portfolio companies, Sierra and OpenAI may benefit directly and indirectly from their relationship with ICONIQ, including through access to capital, strategic guidance, operational support, networking opportunities, and other resources provided by ICONIQ. Accordingly, Mr. Taylor may have an incentive to make statements that are favorable to ICONIQ in order to maintain or enhance those relationships. In addition, as a client of ICONIQ, Mr. Taylor’s interests may be aligned with the performance, reputation, and fundraising prospects of such funds and the adviser. Positive statements about ICONIQ could indirectly benefit Mr. Taylor in his capacity as a client.
This presentation may contain forward-looking statements based on current plans, estimates and projections. You are cautioned that a number of important factors could cause actual results and outcomes to differ materially from those expressed in, or implied by, the forward-looking statements. The numbers, figures, and case studies contained in this presentation have been included for purposes of illustration only, and no assurance can be given that the actual results of any ICONIQ portfolio company will correspond with the information contained in this presentation. No information is contained herein with respect to conflicts of interest, which may be significant. The portfolio companies and executives highlighted in this presentation are not representative of the entire portfolio or investing history of ICONIQ Venture and Growth. The types, level, and terms of services and resources actually provided by ICONIQ Venture and Growth to portfolio companies may vary. For a complete list of ICONIQ Venture and Growth’s investments please contact us or visit our website.
Certain of the economic and market information contained herein may have been obtained from published sources and/or prepared by other parties. While such sources are believed to be reliable, none of ICONIQ or any of its affiliates, partners, employees, and representatives assume any responsibility for the accuracy of such information.
All of the information in this presentation is presented as of September 1, 2026 (the “Production Date”) (except as otherwise specified), is subject to change without notice, and may not be current or may have changed (possibly materially) between the Production Date and the date actually viewed by you. ICONIQ assumes no obligation to update or otherwise revise any information, projections, forecasts or estimates contained in this presentation, including any revisions to reflect changes in economic or market conditions or other circumstances arising after the date the items were made available to you or to reflect the occurrence of unanticipated events.
ICONIQ is not acting as an adviser or fiduciary in any respect in connection with providing this presentation and no relationship shall arise between you and ICONIQ as a result of this video being viewed by you.
These materials are provided for general information and discussion purposes only and may not be relied upon.
