Christina Cacioppo - Building an Enduring Software Company in the Age of AI - [Invest Like the Best, EP.496]

Category creation and competitive moat

Vanta started eight years ago around SOC 2 compliance and audits, expanding into broader security and "trust" products; it now serves 16,000 customers with five products. Cacioppo attributes its win over early rivals to constant shipping — she argues a software company is only as good as its most recent release — and deep integrations: Vanta connects to 400 tools and aggregates data into what it calls a trust graph, a security data lake broader than competitors', including dedicated security vendors. Go-to-market success also came from brand moments like its widely discussed billboard campaign, whose slogan jokingly suggested compliance work doesn't have to be so unpleasant. She also recounts the "Billboard War": another startup quietly took over Vanta's famous 101/280 freeway spot by paying more on what turned out to be a placement order rather than a locked-in contract, prompting Cacioppo to personally try (and fail) to buy it back by offering double what the startup was paying.

Valuation disconnect and the AI financing environment

Cacioppo agrees Vanta is growing faster with margins still very high, yet the market now assigns dramatically lower multiples to profitable, high-margin SaaS businesses than to low-margin, hyper-growth AI companies. She frames this as cognitive dissonance: she believes a disciplined, high-margin business will outperform over time, but acknowledges the near-term reality that investor appetite is skewed toward labs and infrastructure, plus a talent market pricing people based on speculative AI trajectories, which forces real strategic tradeoffs, including pressure toward token-pricing models that don't fit Vanta's economics.

Agent-first product rebuild

Vanta has rebuilt its core experience around a single user-facing Vanta agent (backed by hundreds of task-specific sub-agents) that proactively performs work with permission — e.g., identifying an offboarded employee's lingering account access and asking whether to remediate it, versus simply alerting. The product now dynamically generates UI rather than relying on static dashboards; a new onboarding flow replaced a traditional step-by-step wizard with a canvas where users can drag in messy files and screenshots for the agent to interpret, after Cacioppo pushed a hesitant team that feared failure to commit to the LLM-based approach. On pricing, Vanta currently eats token costs rather than marking up Claude (customers quickly spot and resist markups), prioritizing usage growth over margin for now, with a long-term goal of outcome-based pricing — which she says remains unsolved industry-wide because defining a universal unit of work is messy, with dozens of candidate definitions and no consensus.

Talent war and company economics

Competing offers dangling trillion-dollar outcomes have pushed total comp for Vanta's top performers up roughly 10x over five years, a trend Cacioppo expects to keep accelerating, checked only by eventual investor concern over dilution. Vanta has not historically done tender offers, but she calls them normalized and expected as part of comp, even if she finds the broader dynamic unhealthy. She says she doesn't personally fixate on valuation or multiples, prioritizing ARR growth and employee sentiment instead, and reiterates no fixed timeline toward an exit — the explicit goal is remaining a large independent company, following the pattern of most major U.S. software firms, with no particular desire to IPO as a milestone in itself.

Overhyped and underhyped trends

Cacioppo calls custom model post-training largely a marketing exercise aimed at investors rather than a genuine quality lever, and is skeptical that enterprise customers actually want a bespoke model handed to them, comparing it to an early, unrefined era of product design. She views the forward deployed engineer framing as disguising professional-services economics that look upside-down outside of Palantir-scale eight-figure contracts, where thinner five- and six-figure deals don't sustain the model. She considers deep company and customer context currently undervalued relative to narrow vertical skills like sales ability, especially as labs build out go-to-market teams with strong sellers who lack domain context. On where capital should go, she flags that the ecosystem of tooling and companies built around open-source models is comparatively undeveloped versus what's grown around the major labs, and names robotics as an underweighted, less-hyped area given improving models.

Security trajectory and market positioning

Looking ahead, Cacioppo describes AI models as more persistent attackers than humans, capable of chaining many low-severity issues into serious exploits (citing an OpenAI talk at RSA on Hugging Face vulnerabilities), which she says will force organizations toward machine-speed, continuous fix cycles rather than traditional severity-based prioritization. She positions Vanta as eventually offering continuous penetration-testing-and-remediation for the broader market — distinguishing it from CrowdStrike (enterprise-grade) and Wiz (multi-thousand-person companies, roughly $150k price point), with Vanta serving customers that would pay closer to $40-50k. She closes by framing the AI era as one of rising excitement but falling institutional trust, with Vanta positioned to supply verifiable trust signals to customers, regulators, and governments undergoing AI-driven modernization pushes, and describes her own personal growth as a public speaker as one of the more surprising skills she's developed running the company.