Lenny's PodcastProduct lessons from Waymo | Shweta Shrivastava (Waymo, Amazon, Cisco)
CHAPTERS
- 0:17 – 3:52
Shweta’s background and what this episode will cover
Lenny introduces Shweta Shrivastava and her career path across Waymo, Nauto, AWS, and Cisco. He sets up the conversation’s focus: what it’s like to build an autonomous driving product, how Waymo measures progress, and broader product leadership lessons.
- •Shweta’s roles across startup and big-tech environments
- •Waymo as a live, real-world autonomous ride-hailing service
- •Preview of topics: KPIs, trust-building, and product/team lessons
- 3:52 – 4:34
Shweta’s scope at Waymo: onboard autonomy, simulation, and ride-hailing scale
Shweta explains the three major areas her organization owns at Waymo. Her teams span core onboard driving software, simulation/validation tooling, and commercial scaling of the ride-hailing product.
- •Software that determines vehicle behavior and trajectory
- •Simulation tools to validate system performance and safety
- •Commercial scaling of Waymo’s ride-hailing go-to-market
- 4:34 – 5:30
Riding in a driverless Waymo: what the experience feels like
Lenny shares his first fully driverless Waymo ride in rainy San Francisco and how quickly it starts to feel normal. The discussion highlights how the app and in-car experience resemble familiar navigation flows—but with an autonomous vehicle executing them.
- •The “mind-blowing” shift: no driver in front seat
- •Waymo app experience likened to Google Maps + a car arriving
- •Ability to reroute mid-trip and how quickly riders acclimate
- 5:30 – 8:14
Teaching cars “body language”: human-like driving, intent, and social norms
They unpack a specific moment where the car subtly inched into a lane—communicating intent without eye contact. Shweta describes how Waymo uses human driving data and deep learning to interpret road users, while also adapting to local social driving norms.
- •Training models on human driving (while filtering ‘bad’ behaviors)
- •Inferring intent from pedestrian orientation, gestures, and context
- •Driving is social: norms vary by city/intersection and must be learned
- •Balancing naturalness with safety and predictability
- 8:14 – 11:03
How Waymo differs from traditional software companies: complexity, mission, and MVP bar
Shweta contrasts Waymo with more typical software companies: the system’s technical depth, uncertainty, and long time horizon demand different PM muscles. She also reframes MVP: safety raises the minimum viable bar dramatically.
- •PMs must go deeper technically than many software-only roles
- •Higher ambiguity tolerance and long-game tenacity required
- •Mission-driven culture centered on reducing traffic deaths
- •MVP still exists, but ‘minimum’ is constrained by safety requirements
- 11:03 – 13:57
Trust-by-design: L4 autonomy vs driver-assist and why the first 5 minutes matter
Lenny asks about human trust and why people disengage driver-assist systems in tricky scenarios. Shweta clarifies Waymo’s L4 approach (no expectation of takeover) and explains how trust is designed into the experience via transparency and support.
- •Waymo is built for full autonomy; ADAS expects human takeover
- •Design goals: credible, predictable, trustworthy behavior
- •Riders often go from ‘wow’ to ‘uneventful’ quickly—and that’s intentional
- •Trust features: in-car monitor view, rider support, seatbelt interventions
- 13:57 – 15:55
A concrete trust cue: speed limits, hills, and tuning for rider comfort
Shweta shares an example of an initially logical behavior (never exceeding speed limits) that still needed refinement for comfort. Waymo adjusted downhill behavior to match what riders perceive as ‘natural’ even when it’s not strictly necessary for safety.
- •Strict adherence to speed limits is appreciated by riders
- •But ‘safe’ isn’t always ‘comfortable’ or ‘natural’ in context
- •Downhill driving in SF: riders expect subconscious slowing
- •Behavior tuning based on rider expectations and experience
- 15:55 – 17:32
How Waymo measures progress: commercial, operational, and system behavior metrics
Shweta outlines Waymo’s KPI stack, separating business health from driving performance. Because Waymo operates a paid, public service, it tracks standard marketplace/funnel metrics alongside detailed safety and driving-quality measures.
- •Two KPI buckets: commercial/operational + system behavior
- •Commercial: trips/week, active users, funnel conversion
- •Operational: cost to operate the service
- •System behavior: safety, road-rule compliance, progress/throughput
- 17:32 – 20:28
Safety and ‘making progress’: benchmarking against humans and avoiding stranding
They go deeper on performance measurement: ‘safer than humans’ requires benchmarks, such as collisions per distance. Shweta also emphasizes the counterbalance to caution—cars must avoid unnecessary slowdowns, rescues, and impacts on surrounding traffic.
- •Safety benchmark example: collisions per 100,000 miles vs humans
- •Safety measurement is nuanced; benchmarks require careful methodology
- •Progress metrics: undue slowdowns, getting stuck, needing remote/rescue help
- •Externalities: not impeding other road users and traffic flow
- 20:28 – 22:43
L4 vs L5 autonomy: why L4 is ‘good enough’ and what’s still hard
Shweta defines autonomy levels and argues that broad usefulness doesn’t require L5 (unstructured, off-road, no priors). She notes remaining technical challenges (e.g., snow) while emphasizing that L4 deployments already deliver real value.
- •L2/L3 = driver assist; L4 = fully driverless in defined domains; L5 = anywhere
- •L5 may be niche; L4 can enable large-scale real-world services
- •Key capability gaps remain (e.g., robust snow driving)
- •Framing progress around practical deployment rather than theoretical perfection
- 22:43 – 25:24
Sustaining investor and leadership buy-in over long timelines: show real deployment progress
Lenny asks how Waymo maintains enthusiasm as a long-term Alphabet investment. Shweta’s answer: meaningful, visible progress—especially commercial deployments—plus disciplined focus on building a real business that creates customer value.
- •Demonstrate momentum with measurable milestones
- •Commercial deployment is the ‘rubber meets the road’ proof
- •Avoid short-term optics; focus on business fundamentals and user value
- •Let results speak, especially amid industry consolidation
- 25:24 – 29:33
Product leadership lessons: work backwards, choose what not to build, and disrupt yourself
Shweta zooms out to cross-company product lessons. She stresses working backward from customer problems (citing Amazon’s PR/FAQ), being explicit about what you won’t build, and proactively self-disrupting to avoid the innovator’s dilemma.
- •Core tenet: solve a real customer/user problem (not tech for tech’s sake)
- •Amazon PR/FAQ as a forcing function for value proposition clarity
- •Prioritization: define ‘not building’ to avoid being all things to all people
- •Large companies must self-disrupt to stay ahead of upstarts
- 29:33 – 33:05
Underrated PM skills: listening, empathy, and challenging your own assumptions
Shweta argues that listening and empathy are foundational—especially for influencing without authority—but hard to master. She shares a practical heuristic: actively challenge your assumptions; a lack of productive conflict can signal you’re not truly listening.
- •Listening and empathy underpin influence and alignment
- •Skills develop through repeated exposure across diverse environments
- •Understand stakeholders’ constraints and definitions of impact
- •Tactic: proactively challenge assumptions; expect some contention
- 33:05 – 35:24
Promotion advice and lightning round: impact over optics, plus practical team habits
Shweta advises that promotions follow impact—not gaming the system—while still making ambitions known to managers. In the lightning round, she shares favorite books, an interview question, and a high-leverage communication rule to prevent endless email loops.
- •Promotion: focus on business impact; don’t over-optimize for the title
- •Make ambitions explicit so leaders can staff you on stretching work
- •Lightning round: key books and ‘tell me about a failure’ interview question
- •Process tip: ‘rule of seven/ten’ emails—then hop on a call to resolve
- 35:24 – 42:15
Where to try Waymo, rollout notes, and rider pro tips
Shweta shares where Waymo is available now and how riders can access it, plus expansion notes for LA. They close with a simple pro tip for the ride experience: bring a playlist and enjoy the calm of being driven.
- •Availability: Phoenix metro and San Francisco; LA expansion underway
- •How to try: download the app; Phoenix is broadly open
- •Future city list exists but isn’t shared yet
- •Rider tip: use in-car audio (or Google Assistant) for your own playlist
