Y CombinatorSpenser Skates: How Amplitude Rewired Itself Around AI
Through a bottoms-up experiment phase that triggered hard reorgs; Amplitude rebuilt its analytics roadmap around AI-native teams and dropped SaaS assumptions.
CHAPTERS
- 0:00 – 0:59
The founder’s edge: persistence, clarity, and why analytics will be reinvented
Spencer frames two core beliefs that show up throughout the conversation: startups are filtered by persistence through the “rationally you should quit” window, and founders must be crystal clear about what they’re trying to learn. He also tees up Amplitude’s ambition to lead a coming reinvention of analytics.
- •Most founders hit a 1–2 year point where quitting seems rational; the winners don’t
- •Clarity on what you’re trying to learn is prerequisite to getting useful advice
- •Many people start companies without clear motivations and get derailed
- •Amplitude believes analytics is about to be reinvented and wants to lead it
- 0:59 – 3:53
From skepticism to urgency: why Amplitude initially resisted the AI wave
Harj asks how an established company reorients around AI, especially with internal skepticism. Spencer recounts Amplitude’s early doubts, frustration with “AI strategy” pressure, and the sense that AI capabilities were jagged and often oversold.
- •Amplitude discussed AI in 2023 but didn’t materially shift until late 2024
- •Board/executive pressure for an “AI strategy” can push teams into shallow thinking
- •Early models felt “jagged”: brilliant at some tasks, terrible at others
- •Engineers were frustrated by hype and perceived grifting from non-builders
- 3:53 – 4:25
What changed: AI proved itself in software engineering productivity
The turning point wasn’t a single board memo—it was seeing AI transform coding workflows. Tools like Cursor and others made the productivity gains undeniable, forcing a reassessment of what was now possible for Amplitude’s domain.
- •Coding assistants provided a concrete, measurable productivity jump
- •Seeing engineering transformation made AI feel ‘real’ rather than speculative
- •The shift began in earnest around October 2024
- •AI’s impact on how software is built preceded its impact on many SaaS products
- 4:25 – 8:00
Change agents and early bets: new leadership, Command.ai acquisition, first launches
Amplitude’s AI push accelerated with two catalysts: hiring engineering leader Wade Chambers and acquiring Command.ai. These brought practical experience and demonstrations that helped convert the broader organization, leading to a string of AI product launches.
- •Wade Chambers and the Command.ai team acted as internal catalysts
- •Command.ai had experience with guides, user-triggering, and support-style chatbots
- •Amplitude launched AI Feedback, AI Visibility, and an MCP server
- •A bigger upcoming bet is positioned as a “Cursor for Analytics”
- 8:00 – 10:22
AI Week playbook: training first, then bottom-up innovation
Instead of starting with a top-down AI product spec, Spencer describes training the org to understand model capabilities. An “AI week” combined hands-on demos, leadership participation, and a hackathon-style push—creating a shared baseline and momentum.
- •Step one was organizational fluency, not a fixed AI roadmap
- •Live demo (vibe-coding dark mode) helped leaders ‘show, not tell’
- •Training + hackathon format let teams apply AI to real work immediately
- •AI week outputs became the seed ideas for multiple new initiatives
- 10:22 – 11:36
Why AI product building breaks the SaaS loop: technology-first discovery
Spencer contrasts traditional SaaS product development with AI-driven development. With SaaS, customers can describe what they want; with AI, customers can’t reliably specify what’s possible, so teams must start from model capabilities and map them back to product value.
- •Classic SaaS loop: ask customers, prioritize, build, repeat—Amplitude mastered this
- •AI requires ‘capability-first’ exploration because customer asks are mis-specified
- •“Faster horse” problem: customers describe the old paradigm, not the new one
- •Understanding what models can/can’t do becomes a core product competency
- 11:36 – 15:32
Top-down AI pressure and the engineer skepticism gap
The group explores why AI adoption is often pushed from executives downward rather than bubbling up from engineers. Spencer credits massive vision-setting (and selling) at the societal level, while engineering teams remain constrained by today’s uneven reliability.
- •AI adoption is often “tops down” vs typical developer-led tech adoption
- •Spencer calls Sam Altman an exceptional vision-and-adoption driver
- •Society’s expectations can outrun real product reliability
- •Engineer frustration grows when hype doesn’t match day-to-day capabilities
- 15:32 – 20:30
Reorgs, talent mixing, and ‘burning the boats’: what it cost to transform
Diana asks what Amplitude had to give up to execute the shift. Spencer describes repeated reorganizations, moving out leaders anchored in old SaaS modalities, and integrating acquired founder-talent with longtime Amplitude employees to raise the AI bar.
- •Two major reorganizations in the product/engineering/design org in one year
- •Some leaders weren’t a fit for an AI-native future and were moved out
- •Multiple acqui-hires brought in strong YC founders/teams (e.g., Kraftful, others)
- •Cultural commitment (“burning the boats”) encouraged self-selection into AI work
- 20:30 – 23:30
Roadmap in the AI era: AI-native Amplitude without abandoning core products
Jared presses on how AI reshaped the roadmap and resource allocation. Spencer emphasizes that many AI efforts are meant to make the existing product easier to use, while Amplitude still invests in competitive parity and foundational features like session replay improvements.
- •Four priorities: AI-native rebuild, ease of use, parity for non-analytics products, serve marketers
- •AI efforts (e.g., Ask AI) are positioned as interface improvements to existing capabilities
- •Dedicated AI team formed after early side-project experimentation
- •Core roadmap continues (e.g., session replay ‘zoning’ and other fundamentals)
- 23:30 – 26:10
‘Features, not companies’: AI Visibility, commoditization, and incumbent advantages
Harj brings up Spencer’s controversial take that some AI startups are “features, not companies.” Spencer argues AI Visibility is valuable but easy to replicate and quickly commoditizes—while incumbents can subsidize it as free lead-gen and compete downstream where durable business value lives.
- •AI Visibility built quickly and offered free; it materially boosted sign-ups
- •Many ‘visibility’ tools risk commoditization without a downstream business
- •Incumbents can give features away using their existing revenue base
- •Example of stronger model: pair visibility with a broader workflow/business outcome
- 26:10 – 29:14
Where startups can still win: targeting specific buyers, compliance, and services marketplaces
Asked where he’d start a company today, Spencer highlights vulnerable incumbents and the importance of picking a specific buyer/problem rather than generic agent platforms. He points to enterprise security/compliance blockers and even a tech-support marketplace opportunity.
- •Spencer sees large opportunities to disrupt slow B2B incumbents (e.g., Google Workspace)
- •Belief: analytics will have a ‘Cursor moment’ within ~2 years
- •AI businesses win by focusing on a specific buyer + requirements, not generic agents
- •Enterprise AI adoption is gated by security/compliance—an opening for focused startups
- •Idea: an ‘Uber for tech support’ matching skilled youth with older, high-intent customers
- 29:14 – 32:14
Amplitude’s origin story: pivoting from voice recognition to analytics in a crowded market
Spencer recounts the pre-Amplitude company Sonalyte (voice recognition) and why it failed despite a strong demo and press. The pivot came from building internal analytics and realizing other companies wanted it—despite the market being crowded, it matched their strengths and offered deterministic ‘right answers.’
- •Sonalyte was an early Siri-like voice product; tech/product wasn’t good enough
- •They shut it down after Demo Day and pivoted shortly after (mid-2012)
- •Amplitude started from internal analytics built because engineers avoid paying for tools
- •Even in a crowded market, their edge was engineering fit + a scalable, deterministic domain
- 32:14 – 38:08
Learning hard skills (like sales): find coaches, practice, and hyperfocus on the ‘why’
Garry probes how Spencer learned unfamiliar domains like B2B sales. Spencer describes treating it like learning a sport: do the reps, get coached (e.g., painful but effective feedback), and stay anchored to intrinsic motivation and mission clarity to avoid quitting during inevitable low points.
- •Sales isn’t learned from books; it’s learned through practice + coaching
- •A good coach forces clarity on true customer pain, not superficial feature requests
- •Hyperfocus is guided by a mission/goal-tree rather than random rabbit holes
- •Intrinsic motivation sustains founders through long uncertainty and emotional pain
- 38:08 – 42:29
Founder vs public-company CEO: hierarchy, time discipline, and leading in the right places
Spencer explains the hardest transition: from founder who runs toward every fire to an executive who must say no, delegate, and rely on hierarchy. You can’t lead by example everywhere at 800 people—so you choose where to go deep, while learning to deploy resources effectively at scale.
- •Founders lead by running at the hardest problem; execs must ration attention
- •At scale, you can’t be in the weeds everywhere—only in the highest-leverage areas
- •Hierarchy exists for a reason: ownership and accountability become mandatory
- •Public-company constraints require discipline, even when staying authentic matters
- 42:29 – 44:21
Sharing the late-stage playbook: being vocal, learning in public, and closing vision
The conversation ends with Spencer discussing how he manages time, public communication, and authenticity as a CEO, while acknowledging constraints. He thanks YC for telling later-stage stories and reiterates Amplitude’s intent to lead analytics’ AI-driven reinvention.
- •CEO time becomes a scarce resource; discipline increases with company scale
- •Spencer chooses to be more public and candid while balancing public-company limits
- •More late-stage founder stories could form a broader operating playbook
- •Closing vision: analytics will be reinvented, and Amplitude aims to lead it
