Podcast Summary: Decoder with Nilay Patel

Episode: Zocdoc CEO: "Dr. Google is going to be replaced by Dr. AI"
Date: October 20, 2025
Host: Nilay Patel, The Verge
Guest: Oliver Karas, CEO & Co-founder of Zocdoc

Episode Overview

This live on-stage interview explores the rising influence of AI in healthcare through the operational and ethical lens of Zocdoc—a central platform for booking medical appointments in the U.S. Nilay Patel and Oliver Karas discuss the evolving role of platforms amid AI-driven disruption, the differences between Zocdoc and other app-era marketplaces, the technical and cultural challenges of automation in medicine, and the deeper societal stakes as healthcare and technology converge.

Key Discussion Points & Insights

1. What Zocdoc Really Is (05:32 – 06:54)

• Zocdoc is more than a doctor-finding app; it's an expansive care-access platform connecting patients to providers through multiple channels (web app, insurance partnerships, phone).
  • Quote: “We are making sure that wherever you are as a patient, you can get access to care... even the phone, which seems weird for us to do given that we started out to eliminate the phone from the healthcare process.” — Oliver Karas (06:10)
• The newest development: an AI agent that autonomously books appointments over the phone.

2. Telehealth and Patient Preferences (08:09 – 10:10)

• Telehealth was boosted by the pandemic but is preferred mainly in mental health; for all other specialties, most patients want in-person appointments.

• Option value matters: patients value having telehealth as a backup even if they rarely use it.

  • Quote: “One of the things about Somatic medicine is that telehealth is a little bit like telepizza. It’s great, except you can only eat the pizza when you’re in the same room with it.” — Oliver Karas (09:29)

3. AI Agents and Their Limits in Healthcare (10:10 – 17:47)

• Zocdoc is using AI for operational and routine tasks (bookings, rescheduling, triage), but draws a line at direct medical advice.
  • On agents replacing people: “Dr. Google is going to be replaced by Dr. AI and the patient will develop their own judgment... Not everything that is possible is actually useful.” — Oliver Karas (11:01)
• AI should be transparent—users must know when they’re interacting with AI instead of humans.
• Zocdoc’s AI (Zoe) autonomously schedules via phone. Simple queries go to AI, complex ones escalate to appropriate humans.
• The company employs deterministic orchestration layers around LLMs to triage tasks safely, with clear boundaries on when handoffs to humans are necessary.

4. Business Model, Company Structure, and Data Handling (18:04 – 21:05)

• After a switch from flat fees to per-referral payments, Zocdoc achieved profitability, even though high-volume doctors initially objected.
• Company functions around singular metrics: revenue and profitability, with over 1,000 employees organized functionally.
• The marketplace’s power curve means most value goes to doctors on one end; the new model drives more equitable participation.

5. AI in Patient-Doctor Matching & Return on AI Investments (24:42 – 29:14)

• AI is core to Zocdoc’s patient-doctor matching process.
  • Match factors: insurance, new patient status, case type, availability, visit duration, doctor specialty, and hidden workflow patterns.
• Investment in improving AI-powered matching is long-term, with gradual but demonstrable returns.
  • Quote: “The smaller you go in and measure, the more you realize, oh God, I will never be done with that... Now, AI is great because we can accelerate the kinds of problems that we can solve.” — Oliver Karas (27:49)

6. Decision-Making at Zocdoc (29:38 – 31:58)

• Karas’ priorities: hiring senior leaders, pushing the company to take smart risks, and triangulating future opportunities based on technical feasibility, customer insights, and regulatory pressures.

7. AI Disintermediation and Competitive Moat (31:58 – 43:37)

• Zocdoc, unlike commodity marketplaces, is not easily “disintermediated” (cut out) by AI voice agents or meta-platforms (e.g., Alexa, ChatGPT), due to the complexity and deep custom “cartography” of its database.
  • Quote: “Our cartographers have been at this for 20 years. There’s no way that anyone will catch up to us anytime soon.” — Oliver Karas (33:23)
• Zocdoc is prepared to partner with these agents rather than compete, seeing leverage in hard-to-build infrastructure.

8. Tech Giants, API Access, and the Future of Agentic Integration (48:00 – 52:45)

• Zocdoc is willing to let platforms like Siri access its services, prioritizing optimal patient experience over locking down APIs.
  • Quote: “I would absolutely allow Siri to do that.” — Oliver Karas (48:56)
• The company is open to experimentation (e.g., MCP, visual versus conversational paradigms), but technical and privacy constraints are non-negotiable, especially with healthcare data.

9. Hybrid AI-Deterministic Frameworks vs. Pure LLM Approaches (52:45 – 55:15)

• Zocdoc believes combining traditional algorithms with LLMs ensures both safety and performance for critical tasks; pure LLMs are not yet ready to shoulder these responsibilities alone.
  • Quote: “For some tasks, obviously the LLMs as they come out of the box today are just wonderful. For some tasks, you can’t trust them enough and you need to put them into an orchestration layer.” — Oliver Karas (54:33)

Societal Stakes: Healthcare Policy, Access & Zocdoc’s Role

1. Marketplace as a Pressure Valve (55:15 – 57:12)

• Zocdoc fills critical last-minute doctor capacity (estimated unused 30%)—especially vital as system-wide financial pressures and budget disputes increase.
• Their ambition: improve access, quality, and cost efficiency for patients via better matching and transparency.

2. Price Transparency and Patient Leverage (59:01 – 61:49)

• Zocdoc sees itself as a “union of all the patients,” using aggregated demand to press for provider accountability on things like pricing and hours.
  • Regulatory requirements for transparency often create compliance games; Zocdoc can incentivize better provider behavior directly in the marketplace.
  • Quote: “Whenever you [regulate], you have all the smartest people... trying to figure out how to obey the letter of the law but circumvent the spirit. Whereas Zocdoc can actually reward you for the right behavior.” — Oliver Karas (60:37)

3. Working with Government and Institutions (61:49 – 63:25)

• Zocdoc has partnered effectively with VA and state-level health programs (e.g., Blue Shield of California), shortening wait times from weeks to days for patients.
• The company maintains bipartisan relationships, working with five administrations and adapting to the evolving role of government in tech adoption.

AI and Medical Advice: The Hard Boundary (65:35 – 68:42)

• Zocdoc’s hard line: It will not give direct medical advice—unlike ChatGPT, which willingly does so (sometimes to dangerous effect).
  • Quote: “We don’t give medical advice and that’s going to stay firm until something else changes.” — Oliver Karas (67:19)
• If AI is to play any advice-giving role, it would need robust “know-what-it-doesn’t-know” safeguards and be limited to very low-stakes recommendations.
• Karas is skeptical that full medical automation will replace doctors anytime soon: the human factor and complexity of real-world care remains crucial.

The AI Bubble Question (68:54 – 69:30)

• Karas is agnostic about whether current AI investment levels are sustainable.
  • Quote: “Whether we’re paying the right prices for certain assets right now, who am I to judge?” — Oliver Karas (69:28)
• He’s confident that, bubble or not, AI will remain a durable, valuable technology in healthcare operations.

Notable Quotes & Memorable Moments

• On Telehealth vs In-Person Care:
  “Telehealth is a little bit like telepizza. It’s great, except you can only eat the pizza when you’re in the same room with it.” — Oliver Karas (09:29)
• On AI Agents’ Role in Medicine:
  “Dr. Google is going to be replaced by Dr. AI and the patient will develop their own judgment...” — Oliver Karas (11:01)
• On Zocdoc’s Competitive Moat:
  “Our cartographers have been at this for 20 years. There’s no way that anyone will catch up to us anytime soon.” — Oliver Karas (33:23)
• On Zocdoc as a Patient Union:
  “We’re using [patients’] collective purchasing power to start affecting change in the system.” — Oliver Karas (59:45)
• On AI Giving Medical Advice:
  “We don’t give medical advice, and that’s going to stay firm until something else changes.” — Oliver Karas (67:19)

Timestamps for Key Segments

• 05:32 – Zocdoc’s evolution beyond the classic app model
• 08:09 – Telehealth vs. in-person behavioral patterns
• 11:00 – AI replacing Dr. Google and the limits on agents
• 14:35 – Role of AI in triaging and the importance of human oversight
• 18:04 – Company structure, metrics, and business model shift
• 24:42 – AI powering patient-doctor matching
• 31:58 – The “DoorDash problem” and Zocdoc’s unique position
• 33:23 – Why Zocdoc can’t be easily replaced by AI agents
• 48:00 – Willingness to integrate with Siri and other agents
• 55:15 – Systemic issues in U.S. healthcare and Zocdoc’s impact
• 59:45 – Pushing toward real price transparency
• 65:35 – Line between booking and medical advice
• 67:19 – Policy on AI advice: hard boundaries

Conclusion

This episode offers a revealing look at how entrenched marketplaces like Zocdoc are integrating AI to improve efficiency—while fiercely maintaining human oversight in areas of medical judgment. Karas argues that AI will supercharge access and operational convenience (replacing “Dr. Google” with “Dr. AI” for information), but draws a strict line at automated clinical recommendations, prioritizing patient safety, regulatory compliance, and marketplace integrity. Zocdoc’s “coast of England” metaphor for the complexity of healthcare infrastructure underscores why it remains a tough platform to disrupt, even as tech giants and LLM-driven agents eye the medical front door.