May 7, 2025, 8:00 PM ETFireside Chat, Conference Presentation

How AI is Reinventing Software Business Models ft. Bret Taylor of Sierra

Brett Taylor identifies the primary challenge for entrepreneurs as avoiding "single-issue voter" syndrome, where founders regress to their comfortable expertise (e.g., product) instead of addressing emerging business needs (e.g., go-to-market, competition).
Taylor recounts a pivotal moment at age 29 as CTO of Facebook, where COO Sheryl Sandberg advised him to stop doing the work himself and hold the team to his own high standards to enable scaling.
He adopted a strategy of acting as his own board of directors, prioritizing tasks based on business impact rather than personal interest, which created a "virtuous cycle" of improved performance and joy.
Taylor's core advice to founders is to avoid letting "something you're good at become who you are," emphasizing the need for self-awareness to shift focus as the company scales.

Sierra is building customer-facing AI agents based on the hypothesis that a company's singular, branded AI agent will eventually replace websites and mobile apps as the primary digital interface.
The company currently partners with established brands like ADT Home Security and SiriusXM to build agents that handle specific customer experience scenarios, such as alarm troubleshooting or pricing disputes.
Taylor forecasts three distinct AI markets:
- Foundation Models: Expected to undergo consolidation due to high capital intensity, resulting in a few large players similar to cloud infrastructure.
- Developer Tools: "Pickaxes and gold rush" tools (e.g., Databricks, Snowflake) that may face pressure from foundation model providers.
- Applied AI/Agents: Projected to become the new SaaS form factor, with purchasing decisions shifting toward buying specific agents that perform jobs.
Taylor argues that the highest value in AI lies in selling "outcomes" rather than productivity enhancements, potentially creating the first trillion-dollar applied enterprise software company.
He notes that while large enterprises have shown buyer's remorse over initial model licensing (similar to a "lawn" requiring constant tending), the real value emerges when AI solves high-cost business problems (e.g., antitrust review) previously performed by expensive labor.

Sierra utilizes "outcomes-based pricing," charging a pre-negotiated rate when an AI agent resolves an issue autonomously and waiving fees if human escalation is required to align with customer incentives.
Taylor views this pricing model as the natural evolution of software business models, analogous to the shift from perpetual licenses to subscription SaaS.
He warns that incumbents struggle to change business models (e.g., Microsoft's transition from Windows to Azure, Adobe's move to subscription) due to investor impatience, giving startups a significant advantage in adopting new delivery models.
Pricing strategies must account for the specific budgeting constraints of the department being sold to; for example, HR departments often prefer subscriptions over variable usage to facilitate procurement, unlike marketing departments with flexible budgets.
Taylor advises founders to prioritize "top-line growth" outcomes over cost savings, noting that sophisticated companies often reinvest saved costs into growth rather than retaining them as profit.

Established companies can leverage AI to restructure unit economics, specifically by replacing expensive, large-scale operational costs (e.g., 20,000-person contact centers) with automated agents to lower prices or fund growth.
Successful enterprise sales require deep research into the customer before meetings, using tools like AI to understand the client's specific acute business problems rather than pitching generic features.
Taylor emphasizes that entrepreneurs must act as partners to solve specific business problems, distinguishing themselves in a saturated market where many AI vendors present identical value propositions.
The trajectory for legacy companies mirrors the internet era: some (like Walmart) adapted successfully, while others (like Blockbuster) failed to evolve, suggesting a similar bifurcation may occur with AI adoption.

Taylor advocates for vertical-specific AI agents, arguing that horizontal platforms face a "vitamin vs. painkiller" problem where "good enough" homegrown solutions often prevent market penetration.
He believes success requires deep specialization in core workflows specific to an industry (e.g., claims processing for insurance vs. benefits explanation for health insurance) to provide immediate, measurable value.
Taylor cites the "Marc Benioff rule" of strategic excellence: secure one successful customer, replicate it to ten, then to a hundred, rather than relying on "ivory tower cleverness" in strategy.
He notes that the "job" of an AI agent varies significantly by sector, making generic applications difficult to justify compared to specialized solutions that directly address vertical pain points.