The Moment Everything Changes
Somewhere between the invention of the web browser and the rise of cloud computing, the enterprise software industry settled into a comfortable rhythm. Giant systems of record — CRM platforms, ERP systems, IT service management tools — became the gravitational centers around which entire ecosystems of integrations, workflows, and partner networks orbited. Switching costs were astronomical. Incumbents seemed invincible.
Now, in the age of AI agents, that gravity is shifting. And Brett Taylor — co-founder and CEO of Sierra, board chairman of OpenAI, and former co-CEO of Salesforce — believes we're witnessing a platform shift as profound as the birth of the internet itself.
"Clearly, in three years, we could talk about what are the best practices to set up a software team that's optimized for this technology, and we'll know what those best practices are. And right now, we're just figuring them out in real time. And like, my hypothesis is the companies that figure it out first will move the fastest."
In a wide-ranging conversation with Jack Altman on the Uncapped podcast, Taylor unpacked the forces reshaping enterprise software, why incumbents struggle during platform shifts, how Sierra is building AI agents for the world's largest companies, and what the future of software engineering — and humanity itself — looks like in an era of exponentially increasing machine intelligence.
The SaaS-pocalypse: Anxiety, Not Obituary
Public market investors have been punishing software stocks broadly, and the narrative on social media has turned apocalyptic. The fear: if AI can write software in seconds, moats evaporate, and the entire SaaS model collapses.
Taylor sees it differently. The selloff, he argues, isn't an indictment of individual companies — it's collective anxiety about an uncertain future.
"Every software stock is down, but I don't think that means every software company is equally disadvantaged. It's just basically anxiety about the future."
The real question isn't whether software goes to zero. It's about what happens to the systems of record — those massive databases with workflows built around them — when AI agents start doing the work that humans used to do by clicking buttons in a web browser.
Taylor offered a vivid thought experiment: imagine a CRM system where, instead of salespeople staring at dashboards to manage leads and opportunities, you simply tell an AI agent, "Generate me some leads." The database still exists, but its role fundamentally changes.
"How much do you value the database of leads versus the agent that generates the leads? Ancient history, 30 years ago, those were the same thing. But now you're like, 'Gosh, I actually probably care more about the lead generation,' and how it's stored and tracked is actually maybe a more tactical part of it."
This dynamic applies across every major system of record — CRM (Salesforce), ERP (SAP, Workday, Oracle), ITSM (ServiceNow). The value could shift from the database to the agents running on top of it. And yet, Taylor is careful to note that every incumbent could transform and benefit from AI. The question is whether they will.
Strategy Tax: Why Incumbents Stumble
If large software companies have infinite resources, world-class talent, and deep customer relationships, why do startups with 50 engineers consistently outpace them during platform shifts?
Taylor's answer is elegant and damning: strategy tax.
"In these moments of big platform shifts, what were your strengths can become weaknesses."
He walked through the anatomy of how this plays out. When Siebel Systems — the dominant on-premises CRM before Salesforce — faced the rise of cloud-native software, their leadership couldn't simply start from scratch. They had existing assets, existing customers, existing revenue streams. Every decision became a compromise: How do we transition without abandoning our strengths? What if someone still wants on-premises? These seemingly clever strategic moves accumulated into an anchor that dragged them under.
The tax compounds across multiple dimensions. Product architecture must accommodate legacy systems. Business models must shift from perpetual licenses to subscriptions — a transition so difficult that Adobe's successful pivot under Shantanu Narayen remains one of the few celebrated examples. Sales compensation structures must change. Revenue recognition shifts. Public company obligations demand quarterly results that make bold pivots nearly suicidal.
"All of the advantages that you had all of a sudden become anchors that are holding you back from actually doing the right thing."
Meanwhile, startups carry none of these burdens. They can build natively for the new paradigm. And when the technology wave is bigger than any single category — as Taylor believes AI is — the window of opportunity for best-of-breed upstarts to establish themselves before incumbents catch up becomes the defining race of the era.
Sierra's Playbook: Complexity as Moat
Sierra, the company Taylor co-founded with Clay Bavor, has emerged as one of the fastest-growing AI startups in the world, closing $100 million in annual recurring revenue in seven quarters and $150 million in eight. But what truly distinguishes Sierra isn't just speed — it's the deliberate choice to serve the hardest customers in the most regulated industries.
Over a quarter of Sierra's customers have more than $10 billion in revenue. The company serves most of the Fortune 20, the majority of US healthcare insurers, major banks across the US, UK, and Spain, and large telecommunications companies.
"It's easy to make a demo in AI. It's why you can go on X and just see a thousand demos. Demos are cheap. But making an agent sort of industrial grade is hard."
Taylor described how the market conversation has shifted dramatically over the past year. In Sierra's early days, much of the sales process involved explaining what an AI agent even was. Customers worried about non-deterministic behavior and the risks of AI engaging directly with consumers. Now?
"The conversation is clearly we needed this yesterday."
Sierra's competitive edge comes from two deliberately constructed pillars. First, a product that balances ease of use with deep extensibility — allowing deployments at companies like Cigna to go live in just two months, even within heavily regulated environments. Second, a partnership model that combines technical AI expertise with genuine business understanding.
"There's a circle of people who understand the next generation of AI, and then people who understand business, and we're like the company right in the middle of that, maybe the only one."
The company has expanded well beyond customer support. Taylor highlighted Sierra's relationship with Rocket Companies, where AI agents now power the entire homebuying journey — from searching for a house on Redfin, to originating a mortgage on Rocket Mortgage, to servicing that mortgage afterward. For telecommunications clients, Sierra's agents handle billions of dollars in subscription negotiations.
Outcomes Over Tokens: The Business Model of the Future
Perhaps Taylor's most provocative thesis concerns pricing. While much of the AI industry debates token-based or usage-based pricing, Sierra has committed to outcomes-based pricing — charging customers based on whether the agent actually solved the problem or made the sale.
"If you look at the history of software, we went from impression-based ads to cost-per-click ads... from on-premises licenses to subscription-based software. Could outcome-based software be the next?"
The logic is compelling. If you deploy an AI agent to generate sales leads, you care about the number and quality of leads — not how many tokens the model consumed. Token cost is an input uncorrelated with the output customers actually value.
Taylor went further, suggesting that the ability to price on outcomes is almost a litmus test for whether a company is truly building applied AI:
"I actually define applied AI as, can you describe your value proposition without mentioning models? If you have to mention token utilization, it's probably a tool. It's not an application of AI."
The Codex Moment: When Code Gets Written by Machines
As chairman of OpenAI's board, Taylor has had a front-row seat to the rapid advancement of coding agents. But even with that privileged vantage point, the emotional experience of actually using tools like Codex caught him off guard.
"You can talk about it all the time, and then the first time you one-shot something and it turns out really good — and not like slop, but like, really good — it's an emotional experience. It was just sort of like, holy shit. This is real."
Taylor drew a parallel to the introduction of CI/CD (continuous integration/continuous delivery) in software engineering — a paradigm shift that required entirely new processes, testing frameworks, and organizational structures. Teams that started with CI/CD thrived; those that tried to retrofit it onto manual release processes struggled enormously.
The same dynamic is playing out now with AI-assisted coding. The best practices for building software teams optimized for AI don't yet exist. But the companies that discover them first will move dramatically faster than those that don't.
Yet Taylor cautioned against over-extrapolating from software engineering to the broader economy. Code and finance may be limited primarily by intelligence — they're digital, abstract, and highly amenable to AI automation. But most of the economy isn't purely digital.
"If you need to ship a T-shirt from Vietnam to here, yeah, you could automate some of that stuff, but at the end of the day, that cargo ship still needs to be in the water."
Pharmaceutical companies still need wet labs. Clinical trials still require human patients. Construction still involves physical materials. The 10-person billion-dollar company may emerge, but it won't be the norm in a competitive world where rivals use efficiency gains to reinvest and compete harder.
Building on Sandcastles — And Being Okay With It
One of the most striking aspects of building an AI company today is the relationship with impermanence. Sierra invests heavily in proprietary technology — multilingual voice support, background noise handling, proprietary voice activity detection — knowing full well that much of it will become commoditized within a year or two.
"I'm building this and I'm 100% certain we'll throw it away in the next 48 months. But I have to build it because if I don't, I can't serve the bank that has a big business in Hong Kong."
This echoes a broader insight from Tobi Lütke of Shopify: when generating code becomes easy, the durable asset becomes the system of prompts and product decisions that led to it — the ability to "terraform" your software from scratch. Taylor sees this as a fundamental shift in what constitutes intellectual property in software.
The progression mirrors the early web. In 1996, selling a web server was a technology sale. Today, nobody cares how web pages are served — it's a commodity. The value moved up the stack to platforms like Shopify. AI agents are on the same trajectory, moving from a technology-centric sales cycle to a product-centric one.
Humanity in the Age of Superintelligence
Despite spending his days at the frontier of AI, Taylor maintains a remarkably grounded optimism about humanity's future. He doesn't believe machines being smarter than humans diminishes what it means to be human — any more than machines being stronger than humans has.
"We've been weaker than machines for my entire life. And I don't think it makes me feel weak as a person. This is for the first time we have computers that are going to be more intelligent than us."
He described his own emotional journey with Codex — the initial jolt of seeing an AI write high-quality code that he might have tied part of his identity to, followed by the next morning's realization that it was simply a better tool. The accountant before and after Excel is his favorite analogy: the value you provide doesn't change, but the act of doing it becomes completely different.
"We will all be status-seeking animals. We will all compete for the real estate here in San Francisco. Even though our standard of living will go way up, we will all be jealous of people still. We will all just compete. And as a consequence, I think humanity will be just fine."
Looking Ahead: When Regulators Ask for Agents
As for what's next, Taylor offered a striking prediction. The most exciting development this year, he believes, will be the adoption of AI agents in heavily regulated industries — moving beyond early adopters to universal deployment.
And then, the hot take:
"My intuition is regulators will start asking for agents. The idea that you have a human set of controls over a regulated process will start to feel like a risk, rather than the risk being AI."
It's a perspective that flips the current regulatory conversation on its head. Instead of asking whether AI is safe enough to be trusted, the question becomes whether human-only processes are reliable enough to meet regulatory standards.
The Race Is On
Brett Taylor's view of the current moment is simultaneously clear-eyed and exhilarating. The categories of AI that will matter are already obvious — just as search, e-commerce, and digital payments were obvious in 1995. The winners are not yet determined. Incumbents have the right to compete but carry the burden of strategy tax. Startups have the freedom to build natively but must achieve scale before the giants catch up.
The companies that will define this era, Taylor believes, are the ones that figure out how to harness AI not as a technology demo but as a business outcome — measurable, accountable, and deeply integrated into the messy reality of how the world's largest enterprises actually operate. Whether they have 50 employees or 50,000, the ones that move fastest in this window of radical change will earn the privilege of winning.
And somewhere in the middle of it all, the craft of building software is being reinvented in real time — by the very technology it's creating. If that isn't the most fascinating time to be alive in technology, it's hard to imagine what would be.