Grit, Glean, and Kleiner Perkins' next big Enterprise AI incubation — Joubin Mirzadegan, Roadrunner

Joubin explains the core problem Roadrunner solves: the horrific state of CPQ (Configure, Price, Quote) software like Salesforce CPQ, which creates 30-second loading screens and breaks salespeople's backs during critical deal closures. He describes how LLMs can abstract away complexity in structured and unstructured text, similar to how they revolutionized coding and legal work.

• •CPQ software is universally terrible with 30-second page loads during critical deal moments
• •LLMs can reason with both structured and unstructured text to solve complex pricing problems
• •Similar pattern to successful AI use cases: Harvey (legal), Cursor (coding), now CPQ
• •The problem compounds as pricing models shift from simple seat-based to consumption-based

Joubin reveals how he started the Grit podcast as a strategic hiring tool for Glean, needing an excuse to meet top CROs he didn't know. KP was initially skeptical until he recorded a pilot episode. He committed to 100 episodes before judging success, tuning out noise and vulnerability of being exposed publicly.

• •Started podcast specifically to help Arvind at Glean hire a world-class CRO
• •KP was skeptical initially - most VC podcasts are 'pretty cringe'
• •100-episode commitment rule: don't judge quality until you hit that milestone
• •Transitioned from audio-only to video for multimodal consumption (cooking, showering, running)
• •Eventually shifted from CRO-only to founders for more authentic, unfiltered conversations

Joubin shares his detailed approach to making guests comfortable and authentic: cold room temperature, lights bouncing up not down, no pre-shared questions, starting recording immediately without 'ready, go' cues. He refuses to give guests questions beforehand, even turning down interviews if required, because rehearsed conversations aren't real.

• •Temperature kept cool, lights positioned to avoid 'spotlight' feeling
• •Never gives guests questions in advance - kills authenticity
• •Starts recording immediately without 'are you ready?' to prevent performance mode
• •Deep background research to build mental model, then poke at it during conversation
• •Would turn down even Zuck if questions had to be pre-approved (though acknowledges flexibility for once-in-lifetime opportunities)

The story of Glean's incubation reveals how Arvind tackled the 'eye roll' problem of enterprise search - a category declared dead after decades of failed promises from Google and others. Success came from having Rubrik as a design partner, unfettered access to build the right solution, and proving it worked in production before broader launch.

• •Enterprise search was the 'eye roll of the industry' - everyone had promised and failed for decades
• •Mamoon bet on Arvind's technical genius despite category skepticism (pre-LLM era, 2019)
• •Critical advantage: Arvind built Glean specifically for Rubrik, his previous company
• •Design partner approach: unfettered access, co-development, proof in production
• •Getting past category death requires showing real production success, not just demos

Joubin explains why enterprise AI sales is harder than ever: customers must simultaneously learn how to use LLMs in general AND your specific product. This requires extensive handholding through forward-deployed engineers who co-develop solutions, essentially passing a bill through congress with multiple stakeholder alignment.

• •Customers face double learning curve: understanding LLMs + understanding your product on top of 'quicksand'
• •Forward deployed engineer motion is necessary for early AI companies
• •Enterprise sales is like 'passing a bill through congress' - managing multiple stakeholders
• •Engineers underestimate the difficulty of this process
• •Requires mapping the organization militarily: champions, resistance points, campaign strategy

Windsurf achieved 0 to $100M ARR in 7-8 months through an unwavering commitment to both Google-class product AND Salesforce-class distribution. Joubin credits Liam on his team for building the 'most crack sales team' in dev tools, which later moved to Cognition. The key was moving fast, hiring great talent quickly, and being in a market with strong pull.

• •Fastest growth Joubin has seen in a KP company: 0-$100M ARR in ~7 months
• •Founders committed equally to world-class product AND world-class distribution from day one
• •Hired fast and hired great - built complete go-to-market machine rapidly
• •Coding market has natural pull: engineers are expensive, copilots make economic sense
• •Dedicated entire floor to video production - unprecedented for dev tools company
• •Sales team later moved to Cognition, bringing proven playbook

Joubin reveals critical hiring mistakes: don't just look at fancy LinkedIn logos (Snowflake, Databricks). Better to hire someone who fought tooth-and-nail selling the #3 product. In KP's top 8 companies, 38 out of 40 executive roles are people reporting to CEO for the first time - experience isn't the thing, it's context, trust, and learning velocity.

• •Anti-pattern: assuming fancy logos (Snowflake, Databricks) = good hire for your startup
• •Better signal: someone who succeeded selling the #3 product through sheer grit
• •38/40 executives in top KP companies are first-time reports to CEO - experience overrated
• •Must interview deeply for intangibles: chip on shoulder, motivation sources, resilience
• •Sales requires internal flame that burns after repeated rejection
• •Technical bar is rising: salespeople must understand product deeply enough to explain without SEs
• •Stage matching critical: people from 50M+ ARR companies struggle at seed/Series A (no brand, no inbound, pure art vs. science)

The perfect storm: Joubin discovered CPQ was the #1 pain point for CIOs across multiple dinners over 4 years. Pricing models are exploding in complexity (consumption, multi-product, volume discounts). Salesforce has end-of-lifed their CPQ solution but the replacement doesn't exist yet - creating a 2-year window. LLMs can finally handle the complexity, and Joubin has unfair distribution access through his CRO podcast and CIO networks.

• •Validated pain: multiple CIO dinners over 4 years all cited CPQ as #1 problem
• •Pricing complexity exploding: from simple seat-based to consumption, multi-SKU, volume discounts
• •AI will make it worse: everything moving to token-based consumption pricing
• •Salesforce (95% market share) end-of-lifed CPQ, replacement doesn't exist - 2-year window
• •Incumbents can't fix it: would require complete data model rebuild (innovator's dilemma)
• •Joubin's unfair advantage: knows all key stakeholders (CROs and CIOs) through 6 years of network building
• •Technology finally ready: LLMs can reason with structured/unstructured text to abstract complexity

Roadrunner's founding team: AJ (Caltech at 15, finished 2nd in class) and Eugene (Meta) who built NASA Mars rover software together. They independently discovered the CPQ problem. The team spent enormous effort on data model to handle infinite pricing permutations. Following Glean's playbook with 4 design partners - choosing the hairiest, most complex customers to stress-test the system.

• •9 people today, co-founders AJ (Caltech prodigy) and Eugene (Meta) built NASA Mars rover software
• •Co-founders independently found same problem, had complementary skills to Joubin's distribution expertise
• •Massive upfront investment in data model: handling all SKU rules, discounts, geo restrictions, channel partner logic
• •Design partner strategy: 4 companies with hairiest complexity (hardware, software, SaaS, consumption)
• •Shared Slack channels, weekly standups - co-developing solution like Glean did with Rubrik
• •Not demand constrained: 10+ customers wanting to join, bottleneck is engineering bandwidth
• •Clear roadmap, no surprises - just need world-class engineering talent

The AI dream: Roadrunner will proactively suggest deal structures based on historical data. Instead of AEs calling around to deal desk, finance, and top reps asking how to structure deals, the system will say 'you're doing Costco, we just did Nordstrom - adjust these things.' Automating the administrative nightmare that consumes expensive AE time while augmenting their effectiveness.

• •Vision: system recommends deal structures based on all historical quote data
• •Eliminates human-in-loop for deal desk consultations and top rep knowledge transfer
• •AEs spend ridiculous time on admin work in hacky systems - should be solving this
• •Deal desk shouldn't do this either - they should focus on strategic work
• •Current state: active Slack channels with constant ping-pong, complete mayhem
• •Not taking away AE jobs - extending their capabilities and reducing ramp time

Joubin's daily routine: works out every day without exception (Monday bike up Hawk Hill, 2x running, 2x weights, basketball). Salad for lunch every day. Philosophy: easier to do something 100% of time than 90% - eliminates decision fatigue. On grit: Angela Duckworth's definition is passion + perseverance, but the key word is passion. Put yourself in positions where you really care, then it feels like play when it's work for others.

• •Daily workout non-negotiable: bike Mondays, run 2x, lift 2x, play basketball
• •Same lunch every day (salad) - eliminate decision fatigue
• •100% easier than 90%: no cognitive load deciding which days to work out
• •Grit definition: passion + perseverance over sustained time (Angela Duckworth)
• •Passion is the operative word - do things you really care about
• •When you care, you transfer enthusiasm, outlast others, feels light not heavy
• •Sales definition: ability to transfer enthusiasm from one person to another
• •Dream podcast guest: Andrej Karpathy for authenticity, teaching ability, autodidactic approach