Podcast Summary: "From Copilots to Autopilots: Building the Connected Operating System for Healthcare"
Podcast: Becker’s Healthcare Podcast
Host: Lucas Voss (A)
Guest: Tanay Tandon (B), CEO at Kimmure
Date: October 16, 2025
Overview
This episode explores the rapid evolution and adoption of AI in healthcare, particularly focusing on the shift from ambient AI tools (which assist with documentation and scribing) to agentic AI models (which handle complex, autonomous tasks). Tanay Tandon, CEO of Kimmure, shares behind-the-scenes perspectives on explosive adoption rates, the critical importance of integrated platforms, lessons learned from health system deployments, and key principles for evaluating technology partners in healthcare.
Key Discussion Points & Insights
1. The Rise of Ambient AI in Healthcare
[01:17–03:20]
- 2024–2025 marked exponential growth in ambient AI adoption, likened to the viral spread of consumer apps (Snapchat, TikTok) rather than traditional B2B software.
- Ambient scribing/documentation evolved in three phases:
- 2023: Early adopters started experimenting
- 2024: Mass proliferation
- 2025: Adoption by late adopters, often prompted by better EMR integrations (e.g., Epic deploying its own scribe tools).
- Quote:
“Ambient AI...had adoption that looked more like a consumer app, like a Snapchat or a TikTok than B2B software than we've ever seen in healthcare.”
— Tanay Tandon [01:41]
2. The Power of Integration: Beyond Point Solutions
[03:20–04:50]
- Early deployments saw low provider adoption (10–20%) when ambient tools were isolated.
- Integration with multiple workflows (patient engagement, revenue cycle, intake, summarization) drove adoption up to 60–80%.
- Takeaway: Ambient scribing must be one part of a connected operating system, not a standalone tool.
- Quote:
“The reality is most health systems that sign ambient scribing contracts a year or two years ago have abysmally low adoption because it's not part of this operating system...that connect[s] every point solution into a single one.”
— Tanay Tandon [03:55]
3. Emergence and Impact of Agentic AI
[05:25–06:56]
- Agentic models are not new, but current language models can now process and execute complex chains of action autonomously.
- Major use cases: Handling denials, appeals, prior authorizations.
- Results: Kimmure saw 5x growth in revenue cycle throughput with the same staff, thanks to agentic LLMs.
- Driven by the broader labor shortage in healthcare; automation is now essential rather than optional.
- Quote:
“You have to deploy agentic LLMs that can handle actions themselves and go from being copilots to physicians to being back office autopilots that can run lights out.”
— Tanay Tandon [06:39]
4. How Health Systems Should Evaluate and Select Technology Partners
[07:34–09:13]
- Encouragement to pilot many solutions with multiple vendors rather than making a single “forever” choice based on presentations or relationships.
- Real-world ROI and platform integration must be tested under actual conditions.
- Example: Kimmure pilots 100+ vendors within its own operations for maximum insight.
- Quote:
"Too often I found that health system leaders select their vendors a little bit like selecting the name of their baby, which is, you get one shot at it and then you're done...I think it's kind of ridiculous."
— Tanay Tandon [07:57]
5. Principles for Selecting Vendors: Engineering Talent & On-Site Collaboration
[09:13–10:18]
- The most important differentiator is engineering talent and the concept of forward deployed engineering (engineers embedded on-site at client health systems).
- Customization is essential in healthcare because no two systems are alike.
- Without hands-on, on-the-ground engineering, vendors will not be able to deliver true integration or value.
- Quote:
“If your partner isn't sending engineers on site to do hand to hand combat and work on the front lines with your staff, that's a bad partner in today's day and age...”
— Tanay Tandon [09:56]
6. Co-development with Clinicians: The Path to Better Products
[10:40–11:09]
- Best software arises from partnerships with clinicians; vendor engineers and healthcare professionals should collaborate closely.
- Invites more health systems and clinical advisors to join in co-development work with Kimmure.
Notable Quotes & Memorable Moments
- [01:41] On ambient AI adoption:
“2024 and 2025 really exciting because, you know, speaking for Kamir, going from maybe in 23, doing 100,000 appointments a year to this year annualizing to 20–30 million appointments is shocking growth.” - [03:55] On the shortcomings of point solutions:
“Most health systems that sign ambient scribing contracts a year or two years ago have abysmally low adoption because it's not part of this operating system...” - [06:39] On labor shortages and agentic AI necessity:
“The only solution...is deploy agentic LLMs that can handle actions themselves and go from being copilots to back office autopilots that can run lights out.” - [07:57] On vendor selection process:
“Selecting vendors a little bit like selecting the name of their baby...you get one shot at it...I think it's kind of ridiculous.” - [09:56] On the value of forward-deployed engineering:
“If your partner isn't sending engineers on site ... that's a bad partner in today's day and age.”
Timestamps for Important Segments
- [00:33–01:17]: Tanay Tandon introduces Kimmure’s mission and reach.
- [01:17–03:20]: Growth and phases of ambient AI adoption.
- [03:20–04:50]: Lessons learned—why integrated operating systems outperform point solutions.
- [05:25–06:56]: The maturation and applicability of agentic language models in healthcare operations.
- [07:34–09:13]: Best practices for vendor selection—building a culture of piloting and experimentation.
- [09:13–10:18]: The critical importance of engineering talent and on-site deployment.
- [10:40–11:09]: Co-development with clinicians; call for clinical partnerships.
Conclusion
This episode delivers valuable insight for healthcare leaders navigating the crowded AI and automation landscape. Tanay Tandon advocates for an integrated “operating system” approach, extensive solution piloting, and deep technical collaboration. The move from co-pilot tools to genuine autonomous autopilots is underway, and health systems are urged to use rigorous, pilot-driven processes to separate hype from lasting value. The message is clear: success in AI-driven healthcare depends less on any individual technology and more on the quality of integration, partnership, and constant adaptation.
