Podcast Summary

NVIDIA AI Podcast – GTC Live Washington, D.C. (Chapter 2): Agentic AI for Every Industry

Date: November 11, 2025
Panelists:

• Aravan Srinivas (CEO, Perplexity)
• Shiv Rao (CEO, Abridge)
• Scott Wu (CEO, Cognition)
• George Kurtz (CEO, CrowdStrike)
• Brad (Panelist/Moderator)
• Moderator (Unspecified)

Episode Overview

In this high-energy GTC Live episode, NVIDIA’s AI Podcast dives into “agentic AI”—intelligent agents that plan, reason, and take actions to transform work across industries. Four leaders at the cutting edge of AI discuss real-world impact in sectors ranging from cybersecurity to healthcare, developer tooling to consumer technology. The lively panel tackles how AI agents are redefining workflows, the shift from simple chatbots to autonomous assistants, impact on security, scalability of infrastructure, and the economic forces at play in the rapidly evolving AI landscape.

Key Discussion Points & Insights

1. Agentic AI – From Research to Real-World Impact

• Main Thesis: The game in AI is now about deploying intelligent agents everywhere, not just faster chips or bigger models.

  “America is currently winning that race. And that dominance in AI is really fueling a new era where millions of AI agents will exist to help us in every part of business and in life.” – Moderator (01:00)

a. The Rise of Personal AI Agents

• Aravan Srinivas (Perplexity): Comet, their browser-as-personal-assistant, exemplifies this. It gives users a ‘second brain’, empowering them to delegate both mundane and complex tasks seamlessly, no matter the application.

  “...the number of questions a user asks on Comet is 6 to 18x more... That’s just because the AI is there with them everywhere and they’re starting to do a lot of awesome things like setting up their own Shopify stores, Facebook ads, listings... We’re just beginning to see this explosion of people getting a lot more agency and autonomy.” (02:25)

b. Autonomy and the Move from Answers to Actions

• Perplexity’s vision: Move past ten blue links (classic search), straight to answers, and ultimately to action-taking agents that can book hotels, manage tasks, and interact with the world on the user’s behalf.

  “...three years ago we started this transition from 10 blue links to answers... The real transition... to just asking your agent to go do [book a hotel] for you, is only going to be possible with something like a browser agent.” – Aravan Srinivas (14:49)

2. AI in the Enterprise: Coding, Developer Tools, and Productivity

• Scott Wu (Cognition):

  • The firm’s AI coding agents dramatically increase engineering productivity—6-10x on repetitive tasks. This means tackling more projects that previously languished on the wish list.

    “...engineering toil, that’s things like migrations and re-platforms... we’re seeing gains... one hour of an engineer’s time using AI tools corresponds to about 6 to 10 hours not using the tools.” (04:15)

  • Despite hype, real adoption is happening at the largest enterprises—from code generation to automating processes.

  • The “endgame” is not the elimination of programmers, but a shift in how we communicate with computers—closer to natural language and higher abstraction.

    “It will always be up to us as humans to decide what the computers should do, right?... Despite [automation], the number of software engineers has just gone up and up because we have so much more demand for software.” (12:50)

3. AI in Healthcare: Restoring Humanity, Managing Workload

• Shiv Rao (Abridge):

  • AI is addressing critical clinician burnout—many doctors and nurses are leaving due to administrative “toil”.

    “Two out of five doctors don’t want to be doctors in the next two to three years... 30% of nurses [are considering leaving]. We have a public health emergency... That’s where AI comes in.” (06:00)

  • Tools like Abridge unburden clinicians by handling documentation and clerical work, making care more present and human again.

  • Notes are essentially bills in healthcare; proper, compliant documentation ("agentically" generated) saves time and helps keep hospitals running.

    “We’re not, as clinicians, compensated for the care that we deliver. We’re compensated for the care that we documented that we deliver... you can keep the lights on for the health system.” (10:49)

4. AI and Security: New Enemies, New Defenses

• George Kurtz (CrowdStrike):

  • Security risks have accelerated; adversaries now exploit vulnerabilities in minutes instead of weeks/months.

  • To keep up, security operations centers (SoCs) must be automated with AI agents, as humans can’t respond fast enough.

  • Yet AI also democratizes the attacker’s tools:

    “AI... has democratized destruction, and it’s given this level of sophistication to a much broader group that are not as sophisticated.” (07:14)

  • There is no “one-stop” security solution—the key is integrating data and context across endpoint and cloud.

    “We never lose context. And this is the key... It’s not about collecting a pile of data and putting it somewhere. It’s never losing the context.” (19:16)

  • Startups vs Scale: Scale is an asset, but startups can still disrupt with focus and speed by leveraging modern APIs and infrastructure.

5. The Economics of AI: Cost of Inference & Scaling

• Cutting inference costs is crucial for all panelists—they all use or build with massive models.

  • Perplexity builds its own inference stack, tunes open-source models, and introduces high-value subscription tiers (like background agent services ~ $2,000/year).

    “We work with Nvidia and build our own inference libraries... $2,000 a year is pretty cheap for something that can do all of this in parallel at the same time.” – Aravan Srinivas (24:08)

  • Cognition optimizes by using the right model for the job—reserving expensive, powerful models for the most complex tasks and using lighter models elsewhere (ensemble approaches).

    “The biggest and most expensive models you only use in the times that you absolutely need them.” – Scott Wu (27:49)

  • Abridge faces similar challenges; it’s costly to process every doctor-patient encounter with commercial models, so they rely heavily on distilled, open, and fine-tuned models, with de-identified data to ensure privacy.

    “So much of our playbook is around distillation,... fine-tuning and... post-training. Being able to create that feedback loop in healthcare is easier said than done because of... privacy and security.” – Shiv Rao (29:00)

6. The Big Picture: Agents as OS, Hardware, Policy & Future

• Will Agentic AI replace operating systems?

  • Skepticism on OS replacement—hard limitations due to phone OS monopolies, but browsers serve as a “backdoor” for personalized agents, and new hardware (like AR glasses) could change interaction paradigms.

    “The browser lets you access third party apps without actually having to do that at the OS level on the phone. We think cracking that problem is way more important than going out and building a new hardware.” – Aravan Srinivas (30:42)

• Government’s Role (Washington, D.C. Perspective):

  • Security and technology procurement in government must speed up—current processes lag behind AI innovation cycles.
  • Call for government to pioneer agentic stacks and public-private partnerships for “security AGI” and autonomous SoC.

    “Instead of buying technologies that are five years old... in AI dog years, that’s like 50 years.” – Brad (23:06)

Notable Quotes & Memorable Moments

• On AI Shifting Paradigms:

  "You're just faster as a software engineer if you're working with the best AI tools..." – Scott Wu (04:15)

• On AI as a Force Multiplier for Attackers and Defenders:

  "AI, in many senses, is great because we're able to deal with these threats, but it's minting new adversaries because it's now democratized destruction, and it's given this level of sophistication to a much broader group..." – George Kurtz (08:11)

• On the Willingness and Limits of Automation in Healthcare:

  "...we're not going to be able to fully automate a doctor or a nurse. If we're not fully automating them, then the conversations that they're having with their patients are really upstream of so many of the workflows..." – Shiv Rao (10:41)

• On the Role of Scale in Security:

  "Scale is incredibly important now more than ever, when you look at competitive advantages and moats, one of them is scale. It’s not only the amount of data you have, but the customers you can touch." – George Kurtz (19:16)

• On the Future of Personal Agentic AI:

  "Our goal is that even if it takes a few minutes... you should be able to take your mobile app and just speak out the task, forget about it, delegate it... It’s running on the background, on the server, asynchronously comes back to you..." – Aravan Srinivas (15:29–16:36)

Timestamps for Key Segments

• 01:00 – Framing: America’s AI leadership & the agentic era
• 02:25 – Aravan Srinivas on Comet and the "second brain" browser
• 04:15 – Scott Wu on real productivity gains for developers from AI
• 06:00 – Shiv Rao on AI combating clinician burnout
• 07:14 – George Kurtz on security risks and automated SoC
• 10:49 – Shiv Rao on agentic AI and healthcare documentation as billing
• 12:50 – Scott Wu on the endgame for developer tools and employment
• 14:49 – Aravan Srinivas: When will AI actually book your hotel?
• 16:48 – Discussing the tradeoffs of speed vs. background execution for agents
• 19:16 – George Kurtz on scale & security data
• 24:08 – Aravan Srinivas on inference cost and monetization
• 27:49 – Scott Wu on ensemble models for cost and speed
• 29:00 – Shiv Rao on healthcare AI, privacy, and feedback loops
• 30:31 – Will agentic AI replace operating systems?
• 31:40 – Browsers, new hardware (AR glasses), and the future of interface
• 21:40–23:26 – George Kurtz and Brad: How Washington can accelerate security innovation

Tone & Style

• Conversational, engaging, and occasionally irreverent (“not my interns…”)
• Candid about challenges and hype cycles; enthusiastic about possibilities.
• Practical optimism; all speakers blend realism about industry limitations with excitement for rapid AI-driven change.

Summary Takeaway

Agentic AI is leaving the lab and permeating every industry. From security and healthcare to the very way consumers interact with the web, agents are driving explosive gains in productivity, efficiency, and autonomy. While challenges remain—costs, security, adoption curves, and societal adaptation—the panel remains optimistic that with the right blend of infrastructure, policy change, and product innovation, agentic AI will underpin the next transformation in how we live and work.

End of Summary