Clay Bavor | 20VC Summary

1. Key Themes

Unbounded Demand for Frontier Intelligence Will Persist

Clay argues forcefully that the ceiling on demand for the highest capability AI is essentially unlimited — not just for niche scientific problems, but broadly across coding, legal, medicine, and discovery. The assumption that open-weight models will erode frontier model demand misunderstands how much latent demand exists.

"We have not yet appreciated the unbounded demand for, call it frontier levels of intelligence... I think it's hard to get your mind around when you have intelligence that can work around the clock to invent, to build, to discover how you would use that and how much of it you could use." 00:00:00 — Clay Bavor

Token Spend Will Become a Core Line Item in Corporate Budgets

Clay predicts token spend will eventually be treated like headcount — with per-employee token budgets becoming standard. Top engineers are already spending over $100,000/year on tokens, and Clay expects the steady-state ratio of token cost to salary to land near 20%, not the current ~3.8%.

"I have heard and I have observed that top engineers who are really leaning in to Claude Code, Codex, and so on, are spending more than $100,000 on a run rate basis on tokens per year... I would not bet on 3.8%. I would bet on much closer to 20%." 00:24:20 — Clay Bavor

The Forward-Deployed Engineer (FDE) Model Is the Winning Enterprise AI Motion

Sierra deliberately borrowed the Palantir FDE playbook, embedding engineers directly inside customer organizations at launch. This drove exceptional time-to-market results — six weeks for Next, 58 days for Cigna — and is what Clay credits for their scale.

"We, I would like to think at least in the AI space, I would say rediscovered and borrowed this model from Palantir... What we generally find is in getting started, having Sierra and help from our teams kind of drive while our customer is in the passenger seat but navigating for the first version. It's what has enabled us to take companies like Next live in six weeks." 00:30:43 — Clay Bavor

Chinese Open-Weight Models Are Largely Distillations of U.S. Frontier Models

Clay offers a pointed structural explanation for why Chinese open-weight models appear so competitive: they are likely derived from large-scale distillation of U.S. frontier models. The U.S. labs won't cannibalize their own frontier business by releasing equivalent open-weight models themselves.

"Part of the driver of the difference is probably the willingness of Chinese companies to do scaled distillation of the frontier models from the labs... Are they going to compete with themselves and drive price pressure on the frontier models by developing and releasing open-weights models that are of similar capability? If I was running that business, that's not something I would do." 00:00:00 — Clay Bavor

The MCP Gateway as an Internal Operating System

Sierra built a single MCP server aggregating all company systems — Slack, documents, operating reviews — accessible by any agent instance (Claude, Codex, their internal "Pinecone" agent). This has become the connective tissue of how the company operates and makes decisions.

"We began by building what we call our MCP Gateway. This is a single MCP server that aggregates all of the main systems and services that we use to run the company... You can interrogate in essence the entirety of the company, all information that is published, whether it's Slack messages or presentations or operating reviews and so on." 00:20:55 — Clay Bavor

Board Cadence Must Match AI Time Clocks

Sierra holds board meetings every six weeks — alternating three-hour and ninety-minute sessions — and uses written memos instead of slide decks. The faster cadence allows priors to be updated with material changes (e.g., the sudden capability jump in coding agents after winter break).

"We came back from winter break and suddenly coding agents were amazing. You had Claude 4.5, Codex 5.2. There is a fundamental step change in the capabilities of these models. It changed our approach to software development. It changed our approach to the core product. And so having a cadence where you can take in information even from the last six weeks, update your priors, and then change course, I think is quite important." 00:38:21 — Clay Bavor

AI-Native Hiring Is Already Replacing Traditional Engineering Interviews

Sierra has fundamentally changed its engineering interview to be AI-native: candidates are given $150 in token credits, told to choose their own coding agent and tools, and evaluated on how they build something from scratch. Clay expects every interview function at the company to have a strong AI-native component within two months.

"It now looks much more like: here's a kind of prompt. Think through an application you would like to build... Here's $150 to spend on, choose your coding agent. You can use whatever setup you want. Completely bring your own laptop. Bring your own tools. We're going to pay for your tokens. And then build it." 00:53:46 — Clay Bavor

Enterprise AI Is Moving from Support to Full Revenue Generation

Sierra's expansion with customers like Rocket and Next illustrates the company's trajectory from customer support into inbound/outbound sales, product discovery, and loan origination — full lifecycle revenue roles, not just deflection.

"We worked with Redfin to rethink their search experience. We help Rocket reach out to folks who have expressed interest in a refinance... We worked with them to build Rocket Assist to help bring people in and help them shape and size their loan... None of that is service and support." 00:34:38 — Clay Bavor

Willingness to Go Deep Down the Stack — Without Going All the Way

A defining strategic choice at Sierra was to invest deeply in agent architecture and proprietary fine-tuned models on top of open-weight models, while deliberately stopping short of pre-training their own foundation models. Clay frames this as knowing where to slipstream and where to differentiate.

"Wherever we can, we're scanning for opportunities to strengthen our platform... Our first founding head of research was the Princeton professor who literally wrote the paper on language model based agents, the ReAct paper. And so we invented and went further down the stack than I think some companies at that point would have been willing to." 00:36:58 — Clay Bavor

2. Contrarian Perspectives

The U.S. Open-Weight Ecosystem's Apparent Weakness Is Structural, Not a Gap

Most observers frame China's strong open-weight models as evidence of a U.S. capability gap. Clay inverts this: U.S. labs are choosing not to compete with themselves by releasing powerful open-weight equivalents. The Chinese models are downstream of U.S. frontier work, not independent advances.

"If you can't build frontier models yourself, okay, maybe the next best approach is to distill them and offer them up. I think that's probably the main driver of the difference." 00:00:00 — Clay Bavor

Token Spend at 3.8% of Developer Salary Is "Wildly Off" — It Will Reach 20%

The conventional frame (e.g., Salesforce spending $300M on Anthropic sounds huge but is ~3.8% of dev salaries) suggests AI spend is manageable and possibly overvalued. Clay flatly rejects this, saying 3.8% is nowhere near the steady-state level and 20% is the right target.

"I think 3.8% is wildly off from where the steady state will converge... I would not bet on 3.8%. I would bet on much closer to 20%." 00:26:29 — Clay Bavor

Young People With No Work Experience Are Now Among the Most Valuable Employees

Conventional hiring prizes experience. Clay argues that university graduates who have spent unlimited time mastering AI tools have an unprecedented advantage over seasoned professionals — and that Sierra's most effective employees are 22–23-year-olds who are fully AI-native.

"I can't remember a time when a young person with no work experience but with the right mindset and experience using some of these tools has ever been so valued. Some of our most effective employees at the entire company are 22 or 23 years old and have been completely AI pilled." 00:53:18 — Clay Bavor

Deliberate Underpricing in Each Fundraising Round Is a Strategic Advantage

Most founders maximize valuation. Clay and Brett explicitly chose lower prices than they could have achieved in every single round, framing it as milestone-to-milestone discipline rather than valuation optimization.

"Every one of our rounds we actually guided to and took a lower price than we could have." 00:00:00 — Clay Bavor

The "Founders in the Weeds" Debate Is the Wrong Frame — Selective Depth Is What Matters

The trendy "founder mode" framing implies uniform deep involvement. Clay argues what matters is surgical judgment about where founder-applied force actually changes the outcome, not blanket micromanagement.

"You have to have judgment for it... It's pointless to be in quote founder mode 17 layers in the details and something that doesn't matter. It matters a lot if it's our next generation agent architecture and there's something that we can add." 00:45:59 — Clay Bavor

3. Companies Identified

Sierra AI agent platform for the enterprise. Co-founded by Clay Bavor and Brett Taylor. Works with 40% of the Fortune 50, 50% of customers do over $1B in revenue, 30% do over $10B. Raised over $1.5B, valued at nearly $16B.

"We are at multiple larger size than our next nearest similar vintage startup competitors, growing faster, as I said, are working with many of the great companies in the world." 00:29:57 — Clay Bavor

Palantir Enterprise software and forward-deployed engineering pioneer. Directly cited as the model Sierra borrowed its FDE motion from. Clay notes Palantir has "low hundreds" of customers, implying Sierra is targeting much larger scale.

"We, I would like to think at least in the AI space, I would say rediscovered and borrowed this model from Palantir... My understanding is Palantir has low hundreds of customers. We are and intend to be at a lot larger scale than that." 00:30:43 — Clay Bavor

Rocket (Rocket Mortgage) Large Sierra customer. Sierra built Rocket Assist for loan origination, partnered on Redfin search integration, and handles outbound refinance outreach — described as Sierra's clearest example of expansion beyond support into full sales.

"We worked with Redfin to rethink their search experience. We help Rocket reach out to folks who have expressed interest in a refinance... None of that is service and support." 00:34:38 — Clay Bavor

Anthropic Frontier AI lab. Sierra's primary model partner; Claude is integrated into their MCP Gateway and Pinecone agent system. Clay uses Claude Code as a primary coding agent example.

"You can add this single gateway to your Claude instance, to your Codex instance, and indeed to Pinecone." 00:20:55 — Clay Bavor

OpenAI Frontier AI lab; specifically the O1 model cited as a landmark development in reasoning and test-time compute.

"One of the most underrated developments of the past few years was the O1 model from OpenAI in late 2024 where if you recall there was a chart that showed okay test time compute or amount of inference done, amount of thinking out loud and performance and it just keeps going up into the right." 00:13:53 — Clay Bavor

Nebius GPU cloud provider. Founder cited to underscore the depth of compute demand: even at 10x current supply, they'd sell out in a day.

"We had the founder of Nebius on the show the other day and he said that if they 10x supply they could still sell out in a day." 00:15:25 — Harry Stebbings

Cigna Major healthcare enterprise. Sierra customer. Went live in 58 days from kickoff.

"Cigna, right? One of the largest healthcare companies in the world. Live and I think it was 58 days." 00:32:39 — Clay Bavor

Redfin Real estate platform. Sierra partnered to rethink their home search experience for Rocket.

"We worked with Redfin to rethink their search experience." 00:34:38 — Clay Bavor

Opera Technologies Japanese company acquired by Sierra to establish a Japan presence and bring cultural expertise in omotenashi (extreme hospitality).

"We recently acquired a company in Japan, Opera Technologies... to hit the ground running there and to have a team that can be attuned to the cultural nuances of Japan." 00:33:43 — Clay Bavor

Next (clothing/retail) Sierra customer. Went live in six weeks from kickoff. Sierra also builds personalized product recommendation and outfit-building experiences for them.

"We work with them on personalized product recommendations. How do you help someone build an outfit, a bigger basket of things they will love? Again, that's much more sales than support." 00:35:00 — Clay Bavor

SiriusXM Early Sierra design partner and one of the foundational companies Sierra built its first platform with.

"Olokai, SiriusXM, Sonos, Weight Watchers. We built the first version of our platform with our engineers deeply embedded inside those companies." 00:31:13 — Clay Bavor

Sonos Early Sierra design partner.

"Olokai, SiriusXM, Sonos, Weight Watchers. We built the first version of our platform with our engineers deeply embedded inside those companies." 00:31:13 — Clay Bavor

Weight Watchers Early Sierra design partner. Sierra's founding engineer Mihai was embedded so deeply he received Weight Watchers' internal performance review emails.

"Our founding engineer, Mihai, was actually an employee of Weight Watchers, including getting like it's performance review time emails and so on." 00:31:13 — Clay Bavor

Olokai Early Sierra design partner and one of the first companies they built the platform with.

"Olokai, SiriusXM, Sonos, Weight Watchers. We built the first version of our platform with our engineers deeply embedded inside those companies." 00:31:13 — Clay Bavor

Salesforce Named both as a competitive incumbent and as the training ground for Brett Taylor's world-class sales instincts. Mark Benioff cited for spending $300M/year on Anthropic for dev teams.

"Brett, having spent time at Salesforce, really learned from the best. Like Mark is extraordinary... when it comes to instincts on how to sell, it was like, yep, okay, makes sense." 00:57:08 — Clay Bavor

Google Clay's employer for 18 years. Noted for its willingness to build infrastructure from scratch, its alignment of mission and talent, and as the breeding ground for both Clay and Brett (via the APM program).

"What people underestimate about Google is when you have the alignment of an ambitious, enduring mission, incredibly smart people, and a culture that values truth and building and service of that mission, that company can kind of solve anything." 00:58:28 — Clay Bavor

Framer No-code enterprise website builder used by Perplexity, Miro, Mixpanel. Podcast sponsor.

ROX Revenue agents platform for Global 2000 enterprises. Podcast sponsor.

Superhuman Go AI-powered inbox and productivity tool from the makers of Grammarly. Podcast sponsor.

Lovable AI app-building platform. Cited for reaching $500M ARR with only 149 people, used as a contrast to Sierra's enterprise headcount requirements.

"You had Lovable announce yesterday hitting I think 500 million ARR with 149 people." 00:18:24 — Harry Stebbings

Perplexity AI search company, mentioned as a Framer customer.

Miro Collaboration platform, mentioned as a Framer customer.

Mixpanel Analytics platform, mentioned as a Framer customer.

Character AI Early AI startup. Named as an example of companies that tried pre-training their own foundation models — a path Sierra deliberately rejected.

"Character, Inflection, Adapt, great people at these companies. But got the capital expense, the ongoing capital expense to create what is effectively a highly perishable bag of floating point numbers." 00:09:27 — Clay Bavor

Inflection Early AI startup. Named alongside Character and Adapt as cautionary examples of unsustainable pre-training capital expense.

"Character, Inflection, Adapt, great people at these companies. But got the capital expense..." 00:09:27 — Clay Bavor

Adapt Early AI startup. Named alongside Character and Inflection.

"Character, Inflection, Adapt, great people at these companies. But got the capital expense..." 00:09:27 — Clay Bavor

4. People Identified

Brett Taylor Co-founder and co-CEO of Sierra. Former Salesforce President. Google APM program Class 1. Described as having world-class sales instincts and second-to-none system design and architecture judgment.

"Brett's majors are definitely sales and then engineering. He is really good at selling software. He is really good as a software engineer still... Brett's instinct on system design and architecture are second to none. And so I trust his judgment more than I trust my own." 00:56:41 — Clay Bavor

Sundar Pichai CEO of Google. Cited for extraordinary dynamic range — ability to move fluidly between highest-level strategy and pixel-level product details.

"Sundar has a remarkable ability to look at a problem from wildly different zoom levels. His dynamic range in thinking is second to none. Zoomed all the way out, highest level strategy... all the way into the details, the pixels, the drop shadows, the sound, the texture of something." 00:57:37 — Clay Bavor

Pat Grady Partner at Sequoia Capital. Named by Harry as one of the references who called Clay "special."

"When Pat Grady at Sequoia and Neil Major at Green Oats tell you someone is special, well, it kind of means something." 00:00:37 — Harry Stebbings

Neil Major (Neil Mater) Investor. Named alongside Pat Grady as someone whose exceptional reference for Clay carries significant weight.

"When Pat Grady at Sequoia and Neil Major at Green Oats tell you someone is special, well, it kind of means something." 00:00:37 — Harry Stebbings

Ravi Gupta Mentioned by Harry as someone whose references for Clay were outstanding.

"When I speak to Neil Mater, Ravi Gupta, Sangin at GV, and I hear what I hear, honestly they were some of the most astounding references I have had." 00:04:40 — Harry Stebbings

Dan Mies (Dan Miss) Cited by both Harry and Clay as an impressive and deeply humble leader in the AI space. Described with genuine affection by Clay.

"We had Dan Miss on the show. Incredible. I love Dan Miss. I love Dan Miss too. Also one of the most humble leaders I've ever met." 00:17:05 — Harry Stebbings

Mihai Sierra's founding engineer. So deeply embedded at Weight Watchers as a design partner that he received internal HR emails. Described as an extraordinary software engineer whose code is in production.

"Our founding engineer, Mihai, was actually an employee of Weight Watchers, including getting like it's performance review time emails and so on." 00:31:13 — Clay Bavor

Mark Benioff CEO of Salesforce. Called "the GOAT" of enterprise sales by Clay. Cited for spending $300M/year on Anthropic for developer teams.

"Brett, having spent time at Salesforce, really learned from the best. Like Mark is extraordinary... The GOAT. Unbelievable." 00:57:08 — Clay Bavor

Richard Hamming Renowned computer scientist. His talk "You and Your Research" is cited by Clay as possibly the single best thing a new graduate can read — centered on finding great people, working alongside them, and the compound interest of early knowledge accumulation.

"There's a talk that I love by the renowned computer scientist Richard Hamming, 'You and Your Research.' For any new graduate, probably the single best thing on a per word basis I think you can read." 00:50:44 — Clay Bavor

Brandon McCaw (Brandon at McCaw) Unnamed fully but cited by Harry as a founder who spends more on tokens than on headcount — the extreme end of the token-spend spectrum.

"I had Brandon at McCaw on the show who says he spends more on tokens than he does headcount." 00:26:20 — Harry Stebbings

David McCullough Author. Clay's recommended must-read is McCullough's The Wright Brothers, praised as the most accurate portrait of entrepreneurship and invention he's encountered.

"David McCullough, The Wright Brothers. It's so good... To me it is as accurate a portrait of entrepreneurship and invention as has been written anywhere." 00:59:33 — Clay Bavor

5. Operating Insights

"Think Apart, Think Together" — A Framework for Avoiding Co-Founder Groupthink

Clay and Brett use a disciplined two-phase process for important decisions: each founder independently writes up their full perspective first, then they compare. This was how they defined Sierra's core values and is a repeatable technique for any founding team that risks echo-chambering each other.

"Brett and I have a technique we call think apart, think together, where we'll initialize on a prompt. And the idea is not to group think one another. So we want to get kind of the best of our independent thinking... There was first of all a shocking amount of overlap, which I guess shouldn't have in retrospect been surprising." 00:47:35 — Clay Bavor

"Sierra Brain" — A Strategy Thought Partner Grounded in Deep Company Context

Clay built a living strategic document — 20–30 pages grounding any agent in what Sierra is, how it's organized, its competitive landscape, strengths and weaknesses — layered with every recent board letter and operating review. The result is an AI thought partner that can reason about what the company should do next. Any operator could replicate this architecture.

"Sierra Brain starts with a 20 or 30 page document that grounds any agent in what we are as a company... On top of that, I've given it access to every one of our recent board letters, every one of our recent operating reviews... And then use it to reason about what we should be doing as a company. And so it's a bit like a strategy thought partner, if you will, that knows the company." 00:23:02 — Clay Bavor

Board Letters Over Board Decks — Writing Forces Honest Thinking

Sierra sends a 6–10 page written memo to board members in advance of every meeting, explicitly listing what the company is doing poorly alongside performance. The format forces clarity, removes the ability to hide behind slides, and transforms board members from passive audience to active pre-briefed participants.

"We don't have board decks. We have board memos. So Brett and I write a usually six to ten page memo. There is a saying writing is just thinking on paper. And I think it's very hard to hide from writing." 00:38:50 — Clay Bavor

The Constraining Factor in Software Development Is Shifting — Plan Accordingly

Clay invokes Andy Grove's breakfast factory framework to describe how the bottleneck in software engineering is moving. Writing code was the constraint; reviewing AI-generated code is now the constraint; soon it will be deciding what is worth building at all. Operators should be actively repositioning their teams and evaluation processes ahead of each shift.

"It will be interesting to see how the rate limiter in software development moves around... It used to be writing code, now it's probably reviewing code. Pretty soon it will be deciding what is worth building and editing, kind of what could exist to what should exist." 00:25:17 — Clay Bavor

Build Applications First to Inform the Platform — Not the Other Way Around

Sierra's platform architecture evolved by doing real customer deployments first and then extracting reusable platform components from them, rather than building a platform speculatively and hoping customers would build on it. The first, second, and third customer deployments were unique — but they taught Sierra what to generalize.

"You can build a platform and hope that people come, right? The applications get developed on it. Or you can build applications to inform a platform that makes building the third, fourth, and fifth that much easier. And so wherever we can, we're scanning for opportunities to strengthen our platform." 00:36:58 — Clay Bavor

6. Overlooked Insights

The "Ghostwriter" Coding Agent Writing Production Code Is a Signal About Where Developer Leverage Actually Lives

Clay mentions almost in passing that Sierra built an internal agent called Ghostwriter — an agent for building agents — and that it has written code now running in production. This is a second-order effect almost nobody in the discussion focused on: the real productivity unlock isn't just AI-assisted coding, it's agents writing the agents that run the business. For investors, this suggests the companies that will win in enterprise AI aren't just those deploying agents externally but those most aggressively using agents to build their own products faster than competitors can respond.

"We released Ghostwriter. This is an agent for building agents. It's kind of agents all the way down... Some of the code that is in production, he has written." 00:27:30 — Clay Bavor

Per-Employee Token Budgets Are Coming — And Will Become the New Headcount Line in Finance

Clay briefly mentions a structural shift in how CFOs will think about capital allocation: salary plus a token budget will become the new total compensation and cost-per-employee calculation. This was mentioned once and dropped, but the implications are enormous. Any company building financial planning tools, HR software, or compensation benchmarking products for enterprises should be designing for a world where token budgets are a standard line item — and any operator should be building that framework now before it becomes a reactive scramble.

"Here's your salary, here's your token budget, have at it... For CFOs in the future like capital allocation will look more like how do we allocate OpEx, and then headcount, and headcount will be both headcount for salaries and SBC, and also tokens associated with headcount." 00:24:49 — Clay Bavor