By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
CHAPTERS
- 0:49 – 2:14
Vanta’s origin: building and proving trust through security and compliance
Christina gives Patrick a fast history of Vanta, starting from the premise of helping companies build and earn trust. She describes how Vanta began with SOC 2 and expanded into a broader set of security and trust products used by thousands of customers globally.
- •Vanta was started about eight years ago with the premise of helping companies build and earn trust.
- •Early focus centered on audits like SOC 2, and expanded to other frameworks like ISO 27001 and HIPAA, plus security questionnaires and trust centers.
- •The company’s goal is helping customers implement security because that’s what their customers demand.
- •Vanta evolved from being seen as audit-centric to covering more security domains and trust-building methods.
- •Today Vanta serves roughly 16,000 customers globally across about five products.
- 2:14 – 4:17
How Vanta won: relentless shipping, deep integrations, and a “trust graph” data foundation
Christina attributes Vanta’s category leadership to execution—shipping continuously in a world where building is getting easier. She highlights Vanta’s heavy investment in integrations and data aggregation as a compounding advantage that now powers automation and agents.
- •Christina says software companies are “only as good as the last thing you shipped,” making continuous delivery essential.
- •Competition is harder because new products can be built extremely quickly, increasing pressure to keep shipping.
- •Vanta invested heavily in integrations, now connecting to hundreds of tools.
- •Those integrations pull in data that powers automation, AI, and agents.
- •Vanta frames the resulting dataset as a “trust graph,” essentially a security data lake that accelerates new product development.
- 4:17 – 7:37
Living through the “SaaSpocalypse”: multiple compression, investor narratives, and founder cognitive dissonance
Patrick and Christina discuss how market sentiment changed even as businesses like Vanta maintained strong growth and margins. Christina describes holding two realities at once: long-term business fundamentals versus short-term valuation and narrative shifts favoring AI-maximalist bets.
- •Christina calls the last year “SaaSpocalypse,” creating fear that systems-of-record companies are suddenly devalued.
- •Patrick notes the market assigning much lower multiples to high-growth, high-margin businesses than before.
- •Christina describes “cognitive dissonance” between long-term cash outcomes and today’s market pricing.
- •She observes some investors focusing on funding AI labs and extreme endpoints rather than “in-between” software businesses.
- •She flags the talent market as a practical consequence of shifting narratives and multiples.
- 7:37 – 9:39
What Vanta means by ‘agent’: moving from alerts to automated remediation workflows
Christina explains why she considers Vanta an AI business and describes an agent-oriented product experience. Instead of simply monitoring and alerting, the product increasingly proposes and executes actions with user permission to keep work moving forward.
- •Christina says Vanta benchmarks itself against AI applications and believes it is unusually far along in agent-driven product behavior.
- •She describes chat-based interaction plus long-running agents and dynamically generated UI to keep users in flow.
- •Patrick clarifies the shift from “red alerts” to systems that recommend and execute remediation.
- •A concrete example: detecting incomplete offboarding when an employee leaves and offering to automatically remove access.
- •Christina emphasizes permissioned action: the agent asks, the user approves, and then Vanta executes.
- 9:39 – 13:12
The AI pricing trap: why token-based and credit pricing backfires with savvy customers
Christina details customer pushback on token-based pricing, arguing it forces applications into direct price/performance comparisons with underlying models. She explains why even “credits” that obscure token costs often fail because customers recognize the markup.
- •Christina says charging per token prompts customers to compare your token costs to model providers (e.g., “you marked up a Claude token 4X”).
- •She argues token pricing can put an application in direct competition with the models it depends on.
- •Customers quickly test alternatives if they suspect markup without clear differentiation.
- •She notes that not marking up tokens connects directly to margin pressure and business model trade-offs.
- •She says credit systems that hide token exchange rates are usually transparent to customers anyway.
- 13:12 – 16:14
Model selection and evals: task-specific measurement, SME-written frameworks, and prioritizing quality
Christina describes Vanta’s approach to evaluating and swapping models: building task-specific evals and involving subject matter experts to design the evaluation frameworks. She notes that, for Vanta’s use cases, model choice is currently driven more by quality than cost.
- •Vanta invested in evals early in the year, including task-specific or agent-specific eval frameworks.
- •Subject matter experts (with security/compliance backgrounds) help write and validate evaluation frameworks.
- •Although Vanta presents “one agent” to users, Christina says it’s many agents behind the scenes performing specialized tasks.
- •Christina says model swapping is mostly driven by quality at the moment, not cost.
- •She contrasts Vanta’s token usage with coding and document drafting, saying Vanta uses fewer tokens because it’s more “decision/action” oriented.
- 16:14 – 22:42
Rebuilding for AI: modularizing an aging codebase and adopting AI-native onboarding flows
Christina contrasts how easy it is to use coding agents on small codebases versus large, older ones, and explains Vanta’s work to modularize for better agent effectiveness. She also tells a detailed story of replacing traditional onboarding wizards with an LLM-style “canvas” that ingests messy files and screenshots.
- •Christina says throwing coding agents at an eight-year-old codebase can fail due to limited context windows and the need for high-context “babysitting.”
- •Vanta invested in modularizing parts of the codebase so agents can understand smaller, self-contained areas.
- •She frames the problem as wanting to “onboard 9,000 new engineers tomorrow,” which heightens code comprehension constraints.
- •The team initially kept proposing traditional onboarding wizards because they felt they weren’t “allowed to fail.”
- •Vanta shipped a Claude-like onboarding experience: drag in messy files/screenshots, the system asks questions, and it sets customers up quickly.
- 22:42 – 30:38
Silicon Valley talent wars: trillion-dollar pitches, comp inflation, tenders, and keeping people motivated
Patrick presses on hiring dynamics where competing companies promise massive valuations and rapid markups. Christina explains how Vanta counters hype with transparency and data, describes comp changes over time, and discusses why tenders have become normalized even if they feel unhealthy.
- •Christina describes pressure to match market “vibes” that every company can become enormous, and the risk of seeming unambitious if you don’t.
- •She says Vanta tries to counter hype with facts: sharing context and metrics broadly with candidates.
- •She recounts recruiting pitches she hears in-market, including rapid valuation step-ups and “get in early” narratives.
- •Christina estimates some top total compensation has risen roughly 10x over five years and believes it may need to be even higher for top performers.
- •She says tenders are now expected by employees as part of compensation norms, even if she finds it unhealthy.
- 30:38 – 34:38
What’s underhyped vs overhyped: company context, token pricing, low margins, and post-training for marketing
Christina distinguishes durable value from hype in today’s AI discourse. She argues that company and customer context is undervalued, while token-based pricing, intentionally low margins, and post-training models for signaling are overemphasized.
- •Christina says it’s underhyped how valuable company/customer context is versus narrow functional skill; she would sometimes take a worse “vertical skill” person who knows the company better.
- •She suggests some AI companies overweight hiring for generic sales talent rather than contextual understanding.
- •She names token-based pricing as overhyped, along with low margins pursued for their own sake.
- •She criticizes custom post-training done primarily for marketing or investor signaling rather than necessity.
- •She says inference providers often frame post-training primarily as a cost move rather than a quality move.
- 34:38 – 36:29
More AI business-model skepticism: FDEs, services economics, and what founders talk about at dinners
Christina questions the durability of the “FDE” framing, describing it as professional services dressed up to attract a different talent profile. She also shares common founder conversation topics: talent poaching, whether approaches are working, and personal agents.
- •Christina calls FDEs “crazy,” framing them as professional services wrapped to attract highly ambitious technical talent.
- •She contrasts big-contract economics (citing Palantir as the common reference point) with the challenges of supporting smaller contract sizes using FDE-heavy approaches.
- •She describes founder dinners where people compare notes on talent competition (including companies perceived as aggressively hiring).
- •Founders trade notes on what’s actually working, including personal agents and practical adoption details.
- •The discussion includes ongoing curiosity about tools, roles, and repeatable playbooks in the current era.
- 36:29 – 38:39
Global model choices and continuity risk: why some customers avoid dependence on US-only access
Christina shares a surprising observation from visiting customers in France: many Paris software companies use Chinese models. She relays their reasoning as business continuity risk management rather than nationalism.
- •Christina reports learning that many French software companies are predominantly using Chinese models.
- •She initially assumed nationalism or anti-American sentiment, and says she was corrected by customers.
- •Customers cited business continuity risk: fear that access to American models could be restricted.
- •She gives an example rationale involving access being closed off to non-Americans, prompting switching behavior.
- •The theme connects AI infrastructure decisions to geopolitical and availability risk.
- 38:39 – 47:18
Security at machine speed: persistent AI attackers, chaining low-severity issues, and continuous pen-test + patch cycles
Christina explains how AI changes the threat landscape: attackers can be more persistent and combine many small issues into serious vulnerabilities. She and Patrick discuss a future of continuous automated testing and remediation, with humans moving toward system design and evals rather than constant manual decision-making.
- •Christina says models are more persistent than people and will try vastly more attempts, increasing the chance of discovering exploit paths.
- •She emphasizes chaining multiple benign or low-severity issues into meaningful attacks, altering prioritization logic.
- •She argues organizations will need to fix issues as fast as they find them, reducing reliance on traditional prioritization queues.
- •Patrick describes a vision of constant agent-driven penetration testing paired with “buddy agents” that patch and remediate.
- •Christina says humans will increasingly architect systems and write rules/evals, rather than making every micro-decision.
- 47:18 – 55:35
Reinventing software—and procurement reality: agent-forward UX, betting on the future, and enterprise ‘AI off switches’
Christina discusses strategic product trade-offs in going “agent-forward” instead of building traditional SaaS UI, arguing this is what software will become. She also describes a procurement-driven reality: some enterprises want AI features off initially to ease buying, creating tension with an AI-native product direction.
- •Christina jokes about buzzword dynamics while confirming “agent” is essentially AI in today’s discourse.
- •She describes choosing agent-generated reporting (charts on demand) instead of embedding traditional BI-style UI inside the product.
- •Patrick frames the risk of being “right but early,” and Christina agrees time horizons matter.
- •Christina recounts an enterprise asking to disable AI features to simplify procurement, intending to turn them on later.
- •She describes reluctance to build an AI off switch, viewing it as a trade-off against building more useful capabilities.
- 55:35 – 1:04:56
The billboard war and rebuilding trust: how Vanta’s brand grew—and why trust is the long-term wedge
Christina tells the detailed story behind Vanta’s famous billboard slogan, how it spread online, and how Vanta later lost a prime billboard placement due to weak contracting. She closes with a broader thesis: excitement around AI is rising while trust in tech is falling, making measurable trust and “industrial grade” proof increasingly central to Vanta’s expansion.
- •Christina explains how “compliance that doesn’t SOC too much” was chosen, went viral, and reflected Vanta’s brand tone (“serious about work, not about ourselves”).
- •She recounts securing a valuable billboard spot during a market downturn—then later discovering the agreement wasn’t truly locked in long-term.
- •Another company outbid them for the placement; Christina tried to buy the spot back and failed.
- •She says losing the billboard felt like a vibes-driven loss, joking it cost “a billion of market cap.”
- •Christina and Patrick discuss a broader environment of high excitement but low trust; Christina says Vanta’s role is rebuilding trust with real metrics and definitions, and she characterizes Vanta’s next phase as “expansive.”
