UNPAK3D Podcast

Ep. 26: From Anniversary Side Project to 2 Billion Agent Executions, Joe Moura, CEO of CrewAI

CrewAI just crossed 2 billion agent executions — but founder Joe Moura thinks building agents isn't even the hard part. The real value gap? Getting them to production. Joe joins Vignesh and Arvind to break down why most enterprise AI projects fail at planning, not technology, how Oracle cold-approaching him at the SF Ferry Building validated the business, and what it takes to get Fortune 500 companies actually running agents at scale. Time Stamps:(0:00) Introduction and Joe's path from Clearbit to CrewAI(4:47) What makes CrewAI different in the agent builder market(10:24) Why code-first, multi-agent, and the no-code platform customers demanded(16:16) Real use cases: KYC automation and fraud detection(18:23) How Joe became the "face" of CrewAI to drive adoption(25:04) Enterprise AI: why planning fails companies more than technology(33:17) The data readiness trap — when security pulls the plug(42:44) Partnership strategy: selling with IBM to the US DoD(46:16) Low-ego CEO: learning sales from his own AEs(52:42) The future: fine-tuning comeback and long-horizon agents This episode is brought to you by Stifel Bank — our trusted partner for fund banking and wealth management. Learn more at stifel.com [www.stifel.com]. Where to find Joe Moura: * LinkedIn: https://www.linkedin.com/in/joaomdmoura/ [https://www.linkedin.com/in/joaomdmoura/] * CrewAI: https://www.crewai.com [https://www.crewai.com]

Ep 25: 2 Co-Founders, 5 Dead Ideas, Then Pylon w/ Marty Kausas, CEO @ Pylon

For 40 years, the safe bet in enterprise software was Zendesk or Salesforce. Marty Kausas, founder of Pylon, thinks picking either for customer support today might now get you fired. Before Pylon worked, Marty spent 2 years cycling through 2 co-founder breakups and 5 dead ideas. Today, Pylon has 5x'd revenue two years running, raised a $31M Series B from a16z and Bain Capital, and migrated 150+ customers off legacy support platforms. We cover the "happy grinder" culture, why B2B support is a completely different problem than consumer, the per-seat vs outcome pricing debate, and why Marty thinks buying Zendesk is now the risky move. This episode is brought to you by Stifel Bank — trusted banking and venture debt partner for Sierra Ventures' portfolio. Learn more at stifel.com. Chapters:(0:00) Cold open and intro(2:02) Meet Marty and Pylon(3:06) Co-founder breakups and 5 dead ideas(10:13) Salesforce → Support Force(13:29) LinkedIn as 80% of Pylon's pipeline(22:05) Why B2B support isn't ticket deflection(28:33) Not Sierra, not Decagon(32:34) The context layer(39:34) Shipping 40% faster as a compound startup(55:12) Why you might get fired for buying Zendesk Where to find Marty Kausas:- LinkedIn: linkedin.com/in/mkausas- Pylon: usepylon.com

Ep. 24: From $3.7B Exit to $5.5B Company: What Jyoti Bansal Actually Learned the Second Time

Jyoti Bansal was days away from taking AppDynamics public when Cisco made a $3.7 billion offer that changed everything. Now his second company, Harness, is valued at $5.5B — and he's built it on a contrarian thesis: AI is producing more code than ever, but code isn't the finished product. We cover the real story behind the Cisco acquisition, why his best sales reps from AppDynamics failed in the early days of Harness, how he runs 16 product modules as "startups within a startup," and why 70% of engineering time still happens after the code is written. This episode is brought to you by Stifel Bank. Timestamps(0:00) Trailer(0:52) Introduction and welcome(2:36) Jyoti's journey from small-town India to Silicon Valley(6:01) Three lessons from AppDynamics that shaped Harness(10:29) Proving value — not just finding the problem(17:23) The sales hiring mistake: matching reps to your stage(27:42) 72 hours from IPO to Cisco's $3.7B acquisition(33:57) Startups within a startup — Harness's M&A playbook(39:06) AI's productivity paradox: more code, less throughput(48:44) The "two AIs" framework and where moats really live(58:50) How Jyoti decompresses

Ep 23: The Anti-Vibe AI Revolution, Scott Dietzen of Augment Code

Is "vibe coding" enough for the enterprise? Scott Dietzen explains why professional engineers need an expert AI, not a toy. In this episode, we sit down with Scott Dietzen, Former CEO of Augment Code and Former CEO of Pure Storage. Scott breaks down why the current "vibe coding" trend—where non-engineers generate simple apps from scratch—fails to address the needs of large-scale enterprise software. Key Topics Discussed: * The Context Engine Moat: Why passing 20 million lines of code into a context window doesn't work, and how Augment solves this with a real-time semantic map. * Agentic AI: Moving beyond simple chat to AI agents that can handle parallel tasks like testing and code reviews. * Security First: How Augment secured SOC 2 and ISO certifications to win enterprise trust where competitors failed. * Leadership & Hiring: Scott's framework for building high-trust teams and why he never had to ask for a raise in his career. Timestamps * * (00:00) Intro * (02:52) Augment: The Anti-Vibe Coding * * (05:39) From Chat to AI Agents * * (11:04) The Problem with Fine-Tuning * * (13:05) Building a Context Engine * * (23:23) Security as a Moat * * (29:40) Hiring for Trust * * (43:30) PLG vs. Enterprise Sales * * (54:37) The Myth of the Solo Founder * * (57:19) Leadership & Burnout

Ep 22: Building Category-Defining AI Companies with Gokul Rajaram

In the final episode of the year, we sit down with legendary operator and "company helper" Gokul Rajaram. Gokul shares his unfiltered thesis on the current AI landscape, explaining why he is bearish on infrastructure and middleware companies while betting big on the application layer. We dive deep into his "Compound or Die" philosophy, why startups need to launch their second product immediately after finding product-market fit, and the return of the "996" work culture in Silicon Valley. Key Topics Discussed: * The Big Three in AI: Why Document Processing, Browser Automation, and Voice Agents are the most exciting unlock right now. * The Infrastructure Trap: Why the model layer and "middleware" companies are being squeezed, and why application layer companies have the real leverage. * Compound or Die: Why the old rule of waiting for $10M ARR to launch a second product is dead, and why you must build a compound startup now. * Metrics that Matter: Why you should ignore top-line growth in favor of Gross Revenue Retention (GRR) and Net Revenue Retention (NRR). * The New Work Ethic: The reality of "996" (9 am to 9 pm, 6 days a week) and "007" (12 am to 12 am, 7 days a week) culture in AI-native companies. * Future Roles: The rise of the "AI Ops" engineer and the return of the "Forward Deployed Engineer" Timestamps * * (01:10) - Introduction and Gokul’s journey from Google to Angel Investing. * (04:05) - AI themes to watch and avoid * (11:03) - How to build defensibility at the Application Layer. * (19:50) - "Compound or Die": Launching your second product early. * (22:27) - Why GRR and NRR matter more than top-line revenue. * (30:03) - Advice for non-AI native SaaS companies: Pivot or Perish. * (35:50) - The future of Voice AI and why middleware might be dead. * (43:58) - Founder Archetypes and the return of intense work culture (996/007). * (53:07) - Emerging Roles: The AI Ops Engineer and Forward Deployed Engineers. * (1:00:00) - Gokul’s mental model for decompressing: The "Second Brain".