From Salesforce co-CEO to Sierra: Bret Taylor’s journey in tech innovation | TechCrunch Disrupt 2024
- Aisha Washington
- 8 hours ago
- 7 min read
Bret Taylor’s career has repeatedly moved between building startups and leading some of Silicon Valley’s most influential companies. He helped create Google Maps, founded FriendFeed before its acquisition by Facebook, served as Facebook’s chief technology officer, built Quip, and eventually became co-CEO of Salesforce after it acquired the productivity software company.
At TechCrunch Disrupt 2024, Taylor explained why he left that executive role to co-found Sierra, a company developing customer-facing AI agents. His conversation also covered the leadership advice that changed his career, his decision to join OpenAI’s board during a crisis, and his view of how generative AI could reorganize the enterprise software market.
Why Taylor left Salesforce for another startup
Taylor described his departure from Salesforce as the result of several overlapping considerations. The decision was partly personal: after years inside large organizations, he felt drawn back to entrepreneurship. It was also intellectual and technological. The rapid development of large language models suggested that a new category of software was becoming possible, and he wanted to participate directly in building it.
In Taylor’s view, periods of major technological change favor teams willing to reconsider established assumptions. Startups can begin with the capabilities of a new platform rather than adapting products, organizational structures, and business models created for an earlier era. He pointed to companies such as Amazon, Google, and Salesforce, all of which emerged as the browser and internet reshaped computing.
That did not make leaving Salesforce an easy rejection of the company. Taylor said that if circumstances had made him its sole CEO, he would have remained fully committed. He continues to feel close to Salesforce and characterized Marc Benioff as both an important mentor and a father figure in his business life.
His broader career, Taylor suggested, has been shaped by learning inside several distinct leadership cultures. Working for Marissa Mayer, Mark Zuckerberg, and Benioff exposed him to different ways of building products, managing organizations, and pursuing ambitious goals. Rather than presenting his startup and corporate chapters as opposites, he treated each as preparation for the next.
The advice that changed his approach to leadership
One of Taylor’s most consequential lessons came from Sheryl Sandberg during his time at Facebook. According to Taylor, she challenged him to expect more from the leaders reporting to him and to stop treating his own ability to complete the work as the measure of effective management.
The observation forced him to confront an identity built around personal output. Like many technically accomplished founders, Taylor was accustomed to solving problems himself. That approach can work in a small team, but it becomes limiting when a leader’s responsibility expands across an organization. At scale, doing more individual work may prevent other leaders from taking ownership.
Taylor responded by asking himself a different question each morning: what could he do that day to have the greatest positive effect on Facebook? The answer was not always to write code or take over a project. It could involve clarifying a decision, enabling another executive, resolving an organizational obstacle, or directing attention toward a neglected opportunity.
He said this change made him both more effective and happier. By the time he joined Salesforce through the Quip acquisition, he had learned to define his role through impact rather than a fixed set of tasks. He also adopted an unusually open posture toward new responsibilities, making a point of accepting opportunities Benioff offered him.
For founders, the lesson extends beyond delegation. Companies change as they grow, and the identity that made someone successful at one stage can become a constraint at the next. Taylor argued that leaders must be willing to revise how they see themselves as their organizations acquire new needs.
Sierra and the opportunity in customer-facing AI
Taylor’s enthusiasm for founding Sierra began with what modern AI systems could suddenly do. Tasks such as extracting meaning from large collections of text, synthesizing information, working across languages, detecting nuance, and responding conversationally had shifted from difficult research problems to accessible product capabilities. Models were also beginning to demonstrate limited forms of reasoning.
Sierra applies those capabilities to interactions between companies and their customers. Its platform enables businesses to deploy AI agents that can answer questions, support purchases, and help resolve technical or account-related problems. Taylor framed the opportunity as more substantial than improving the familiar customer-service chatbot.
Traditional chatbots generally follow narrow scripts and often fail as soon as a request falls outside a predefined path. Taylor contrasted that experience with newer generative AI products, which can interpret varied language and maintain more natural exchanges. Sierra’s ambition is to give companies an agent that can act as a capable digital representative rather than merely redirecting users to help articles.
At the time of the interview, Sierra had announced a $175 million funding round at a $4.5 billion valuation. Taylor argued that the company’s value would ultimately depend on production results, not polished demonstrations. Sierra was working with large enterprises and consumer brands, including ADT, on agents intended for real customer interactions.
That focus reflects a practical test for enterprise AI: can it reliably solve a problem in a complex operating environment? For Taylor, technical novelty matters only when it produces measurable improvements for customers and the businesses serving them.
From websites and apps to branded AI agents
Taylor expects conversational interfaces to become a major channel for digital business. Companies currently invest heavily in websites and mobile applications because those products mediate much of the customer relationship. He believes AI agents could eventually command a comparable level of attention and investment.
The economic argument is important. Human-assisted service can be expensive, which has encouraged companies to steer people toward static self-service tools. AI could lower the cost of a conversation enough to make responsive assistance available for far more interactions. That might include text, phone calls, or push-to-talk experiences rather than a single chat window.
In Taylor’s longer-term vision, a company’s agent would travel across interfaces while preserving its brand, knowledge, policies, and ability to act. A customer might reach the same underlying service through WhatsApp, a smart speaker, earbuds, a website, or another device. The agent would represent more than a support department; it could become a unified layer across the customer experience.
The challenge is to make that experience trustworthy. An agent representing a business must do more than generate plausible language. It needs to follow company rules, connect to operational systems, complete authorized actions, and behave consistently across different situations.
Joining OpenAI’s board during a crisis
Taylor also discussed his decision to join OpenAI’s board after Sam Altman’s removal in November 2023 triggered an institutional crisis. He said he considered the commitment carefully, including the time it could require, before concluding that he was in a position to help stabilize an organization that might otherwise come apart.
The choice was personal as well as professional. Taylor acknowledged that Sierra’s work, and his own return to entrepreneurship, had been enabled in part by advances associated with OpenAI. That connection gave him a sense of responsibility when the organization faced an uncertain future.
Taylor did not claim that his participation alone saved OpenAI. Instead, he said he acted on the information available and did not regret the decision. Since joining, he had helped assemble a substantially different board and had watched the organization continue releasing significant new technology.
He portrayed OpenAI as several entities at once: a mission-led nonprofit, a research laboratory, and an exceptionally fast-growing technology company. Those identities create difficult governance questions. Its stated goal of ensuring artificial general intelligence benefits humanity exists alongside the enormous compute requirements, capital needs, and operational pressures involved in developing frontier models.
Taylor did not offer a simple answer for how OpenAI’s nonprofit and commercial structures should evolve. His comments instead highlighted the central tension: the organization must fund and manage development at extraordinary scale while remaining accountable to a broader mission.
How the AI market could resemble cloud computing
Taylor compared the developing AI economy with the cloud-computing market. In that analogy, foundational model providers occupy a position similar to infrastructure platforms such as AWS, Microsoft Azure, and Google Cloud. Both layers demand immense capital investment and provide technical foundations on which other companies build.
A second layer supplies tools, data services, and development infrastructure. Above that, application companies package the technology into solutions for particular users and industries. Taylor positioned Sierra in customer experience, while citing products focused on areas such as software development and legal work as examples of other specialized categories.
This structure matters because most enterprises do not ultimately want raw technology. They want a dependable solution to a business problem. Although building internally can initially appear attractive, the total cost includes maintenance, integration, security, ongoing model changes, and scarce engineering talent. The same logic that drove adoption of software as a service may therefore apply to AI applications.
Taylor expects competition to remain intense. He compared the current excitement with the dot-com period, when many companies pursued similar opportunities but only a smaller number built durable businesses. Access to a model will not by itself create a defensible advantage. Product quality, packaging, distribution, technical execution, and customer outcomes will all influence which companies last.
For Sierra, that means its moat cannot rest on the broad popularity of AI. It must come from serving complex customers well and improving the consumer interactions those customers care about.
Building Sierra as an independent company
Taylor has sold two previous startups: FriendFeed to Facebook and Quip to Salesforce. That history naturally raises questions about whether Sierra is being built for another acquisition.
His answer at Disrupt was direct: Sierra is intended to remain independent. The statement fits the scale of the opportunity he described. If conversational agents become a primary interface between businesses and customers, the category could support a major standalone software company.
Taylor’s journey also reveals a consistent pattern beneath his changing titles. He is drawn to moments when a new computing platform makes established assumptions unstable. His move from Salesforce to Sierra was therefore less a departure from his earlier career than a return to its recurring theme: entering a technological transition early, learning what the new medium makes possible, and building for the market that may follow.
