Sep 19, 2024 · 48m · no-priors
No Priors Ep. 82 | With CEO of Sierra Bret Taylor
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of No Priors, Sierra Co-Founder and OpenAI Board Chair Bret Taylor joins Sarah Guo and Elad Gil to explore the architecture, business economics, and future trajectory of enterprise AI agents. Taylor outlines how goal-oriented guardrails, outcome-based pricing models, and multimodal interfaces are transforming enterprise software and human-computer interaction.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 13.2% of the talking time here. How this is scored →
The hosts as informed peer 5.0 Guest teaching 4.9 Guest disagreement 1.0 The hosts pushing back 1.7
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 16:44 Bret disagrees on latency and security consolidation
Bret explicitly pushes back against Elad's assertion that security reviews and round-trip latency will force consolidation exclusively within hyperscalers.
Hardest push from the hosts ▶ 16:16 Elad challenges cloud consolidation premiseElad pushes on Bret's market thesis, arguing that running applications across separate substrates introduces unacceptable round-trip latency and procurement hurdles.
Biggest teaching moment ▶ 10:04 Bret deconstructs the inadequacy of simple RAGBret breaks down why off-the-shelf retrieval-augmented generation is woefully insufficient for complex enterprise action execution across multiple systems of record.
The host holds their own ▶ 22:25 Elad reframes LLM wrappers as database SaaS equivalentsElad demonstrates strong industry pattern recognition by reframing the dismissive 'AI wrapper' critique as identical to early SaaS platforms acting as wrappers on SQL databases.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Categorizing AI Agents: Personal, Persona, and Company | 4 | 5 | 1 | 1 | Bret opens by establishing a comprehensive taxonomy of agents into personal, persona-based, and company agents. Sarah demonstrates quick technical comprehension by pinpointing constraints around task scope and system integration scaffolding. | |
| Sierra's Platform and the Economics of Customer Conversations | 4 | 5 | 0 | 1 | Bret details Sierra's product proposition and the economics of customer service, explaining how AI shifts contact costs from $13 to sub-$1. The hosts ask clarifying framing questions to anchor the company's enterprise focus. | |
| Beyond RAG: Orchestrating Actions with Goals and Guardrails | 3 | 6 | 1 | 1 | Bret educates on why standard RAG pipelines fall short in real enterprise customer journeys, framing the engineering problem around orchestrating stateful business actions using goals and guardrails rather than static rules engines. | |
| Market Structure: Comparing AI to the Cloud Era | 7 | 5 | 3 | 5 | Elad actively challenges Bret's cloud market analogy, pressing on whether latency and security concerns will force foundation models to consolidate directly inside hyperscalers. Bret politely dissents, and Elad adds sharp SaaS historical context about wrappers over databases. | |
| Outcome-Based Software Pricing and Emerging Enterprise Applications | 6 | 4 | 1 | 1 | Both hosts and guest engage in an aligned peer discussion on outcome-based pricing models versus token billing, with Elad bringing in the concrete Klarna case study as evidence of operational displacement. | |
| Agent OS, Brand Personalities, and Instant Resolution | 5 | 5 | 0 | 1 | Bret breaks down how Agent OS handles non-deterministic customer behavior and multi-model supervision. Elad and Sarah probe on implementation details like prompt tuning versus post-training and instant customer gratification. | |
| Future Form Factors and Human-Centric Computing | 6 | 4 | 1 | 2 | The conversation shifts to future human-computer interfaces. Elad probes Bret on social implications, specifically asking whether conversational agents will cannibalize human-to-human interactions. |
Statements from this episode (13)
Taylor: Truly great personal AI agents require more technology than currently exists
“And so as a consequence, I think it's probably a prerequisite for a great personal agent Probably demands more technology than is currently available, though there's lots of interesting startups in this space, and you could imagine some interesting companies c…”
Taylor: Large consumer giants like Apple, Google, and OpenAI will dominate personal agents
“My take is the domain of personal agents is probably of the very large consumer companies like Apple and Google and OpenAI and others that have big consumer brands.”
Taylor: Branded AI agents will define company digital presence by 2025
“In 2025, existing digitally will probably mean having a branded AI agent that your customers can interact with.”
Taylor: AI cuts call center conversation costs from $13 to well under $1
“You know, for most phone calls, it's called 13 dollars, you know, to service that phone call. Now with AI, you can bring down that cost to well below a dollar, you know, and so all of a sudden you've literally decreased the cost of a conversation by an order o…”
Taylor: RAG is woefully insufficient for meaningful enterprise customer service
“What we found in practice is that broad category of technology investment is woefully insufficient for almost any meaningful customer experience. If you think about, you know, all of the interactions you've had with brands that you care about, what percentage …”
Taylor: AI pre-training will consolidate into a small number of frontier model builders
“I think that will probably play out with the frontier models. We'll end up with a relatively small number of companies doing pre-training you know, which is the really capital intensive part of model building not because, you know, there's not, they're the onl…”
Taylor: Startups degraded by new model releases are wrappers, not real solutions
“If every time there's a new release of an AI model, somehow it decreases your value, it probably indicates you're not actually a solution. You're, you might be a slight value add on top of the models. I think there's a number of startups that unfortunately sor…”
Taylor: Sierra charges based on completed outcomes rather than tokens
“At Sierra, we're really focused on what we call outcome-based pricing, you know, charging for the job done.”
Taylor: Outcome-based pricing could be as big as the SaaS subscription model
“That's another really powerful part of software as a service in the era of AI, is I think you can, you know, I think the best AI companies are aligning their business model with their customers' business models charging for the outcome, you know, and I think t…”
Taylor: AI will augment back-office analyst roles like an 'Iron Man suit'
“I'm not sure this is one job, but I'm really excited for automating the role of an analyst, especially back-office analysis, and not necessarily replacing, but sort of the Iron Man suit, you know, for analysts. If you think about the very superficially high-le…”
Taylor: Sierra deploys supervisor models to monitor other AI models
“We do a lot of what we call supervisor models, so we have models supervising other models.”
Taylor: AI agents will personalize personality per customer within years
“Agents start off with one personality, and then, you know, maybe a few years from now, people have the confidence of saying, let's actually reflect back the personality or demographic of the person talking.”
Bret Taylor: Smartphones will remain primary compute hub via audio and alternative interfaces
“Because of the perseverance of the smartphone, probably if I had to pick, I think the smartphone will remain, but coupled with things like AirPods and CarPlay and others, you'll interact with it more through different modalities. But the Anker supercomputer in…”