No PriorsNo Priors Ep. 108 | With Abridge Founder and CEO Shiv Rao, MD
CHAPTERS
- 0:00 – 0:42
Meet Shiv Rao and Abridge’s mission: letting clinicians focus on patients
Sarah introduces Shiv Rao—practicing cardiologist and founder/CEO of Abridge—and frames the company’s goal: reducing clerical and financial documentation burden through AI. The episode sets up the broader theme of how AI is reshaping healthcare delivery.
- •Abridge converts medical conversations into clinical documentation
- •Focus on reducing clinician clerical load and burnout
- •Context: major Series D fundraising and rapid growth
- •Episode theme: AI’s real-world impact on healthcare delivery
- 0:42 – 1:32
Abridge’s founding thesis: conversations are the upstream signal for healthcare workflows
Shiv explains Abridge’s core thesis since 2018: clinicians won’t be fully automated soon, but clinical conversations are the first signal that drives many downstream workflows. Documentation is the initial wedge, with expansion into broader use cases over time.
- •Healthcare workflows originate in patient–clinician dialogue
- •Start with documentation as a wedge product
- •Long-term expansion into multiple value propositions from the same signal
- •Assumption: augment clinicians rather than replace them
- 1:32 – 3:03
What the product does in practice: “hit record,” talk naturally, get a note and billing-ready output
Shiv describes Abridge’s day-to-day usage: clinicians conduct a normal visit and receive an immediate draft note. The system aims to meet clinical, specialty, system, and billing requirements—far beyond a generic summarization output.
- •Real-time (or near-real-time) draft note generation from conversation
- •Output is tailored by specialty, setting, patient context, and system needs
- •Documentation supports both clinical communication and revenue-cycle requirements
- •Goal: eliminate after-hours “pajama time” documentation
- 3:03 – 3:38
Clinical note vs. billable note: why documentation is a high-stakes artifact
Sarah asks about clinical versus billable notes, and Shiv explains that reimbursement depends on what’s documented, not only what’s delivered. This makes notes central to both patient care and health-system revenue integrity.
- •Healthcare reimbursement is driven by documented care
- •Notes function as bills and revenue-cycle inputs
- •Accuracy and completeness are financially consequential
- •Raises the bar for reliability and auditability
- 3:38 – 5:38
Go-to-market strategy: choosing the hardest customers (large health systems) first
Shiv explains why Abridge deliberately targeted large health systems and academic medical centers instead of starting with small practices. The quality bar is higher—across specialties, care settings, and languages—but defensibility and differentiation are stronger.
- •Healthcare is heterogeneous: DPC/small practices vs. large health systems
- •Abridge ran toward the “hard end” for higher barriers to entry
- •Need to work across many specialties and contexts (ED, inpatient, outpatient)
- •Deep ML research bench as a differentiator (e.g., CMU ties)
- 5:38 – 10:23
Clinician-founder advantage and why adoption accelerated dramatically
Elad probes how Shiv’s clinical background shaped strategy and execution, and Shiv contrasts today’s adoption environment with historically slow healthcare IT sales cycles. He attributes acceleration to an urgent post-pandemic burnout crisis plus the market’s awakening after ChatGPT.
- •Shiv’s ongoing clinical practice informs workflow and credibility
- •Post-pandemic staffing shortages made productivity tools urgent
- •ChatGPT shifted buyer understanding of generative AI value
- •Early demos in 2021–2022 acted as “pre-selling” that paid off in 2023
- 10:23 – 11:23
Enterprise virality and the high-risk/high-reward bet on academic medical centers
Shiv details the 2023 decision to push further into major academic systems, where executive networks spread information quickly. Success with early large deployments created momentum, but failure could have damaged reputation across the CIO/CMIO ecosystem.
- •Health system executives share vendor experiences rapidly
- •Early “home runs” created enterprise-level word-of-mouth
- •Starting at the top increases scrutiny but accelerates scaling if successful
- •Examples of early enterprise traction and rapid multi-year agreements
- 11:23 – 14:36
Ecosystems, trust, and integration: partnering with incumbents while competing with Microsoft
The conversation shifts to navigating incumbents and partnerships—where trust is the core currency. Shiv describes infrastructure expectations (downtime risk) and why integrations and partnerships (e.g., with Epic) are critical in enterprise healthcare.
- •Trust and reliability are prerequisites in provider-facing healthcare tech
- •Abridge becomes “core infrastructure,” so uptime and redundancy matter
- •Partnerships help “borrow” trust within established ecosystems
- •Frequent head-to-head evaluations versus Microsoft; importance of integration
- 14:36 – 17:40
Why “speech-to-text API” isn’t enough: medical ASR accuracy, jargon, and multilingual reality
Elad challenges whether speech recognition is commoditized; Shiv explains why small error rates matter in medicine. Abridge must handle specialty terms, drugs, varied pronunciations, and multilingual conversations—then produce English documentation and structured EHR fields quickly.
- •3–5% error differences can be clinically and financially significant
- •Medical vocabulary and drug pronunciations vary widely by clinician
- •Healthcare visits are dialogues (not dictation monologues)
- •Multilingual and even mixed-language encounters require dynamic language biasing
- 17:40 – 19:54
Inside the stack: distillation, information extraction, and stakeholder-specific outputs
Shiv outlines how Abridge structures the conversation into multiple artifacts: clinician note, patient-friendly after-visit summary, and revenue-cycle documentation. Style transfer and extraction map conversation content to standardized dictionaries and formats needed by each stakeholder.
- •Conversation is transformed into different outputs for different audiences
- •Information extraction pulls meds, symptoms, diagnoses, procedures
- •Mapping to dictionaries and discrete fields enables structured workflows
- •Style transfer reduces patient confusion and supports clearer communication
- 19:54 – 22:12
Beyond notes: orders, claims, trials matching, and real-time decision support
Building on the conversation-as-signal thesis, Shiv describes expansion paths: auto-suggesting or drafting orders, enabling revenue-cycle steps, surfacing clinical trials eligibility, and eventually supporting clinical decisions at point of care using population-level patterns and evidence.
- •Orders can be extracted and placed into the EHR
- •Downstream revenue-cycle actions include coding/claims support
- •Clinical trial matching can be surfaced during the encounter
- •Clinical decision support framed as the longer-term “holy grail”
- 22:12 – 28:43
AI’s broader impact and the adoption curve: wedges, clinician-in-the-loop, and changing training culture
Elad asks about AI’s big-picture trajectory and why adoption lags technology; Shiv argues the key is choosing the right initial workflow—high-frequency but lower-stakes due to clinician review. He shares how clinicians already use tools like GPT/Claude dialectically, and why younger clinicians may accelerate adoption.
- •Adoption depends on workflow stakes and frequency; start where review mitigates risk
- •Clinician productivity and CFO-visible ROI drive deployment decisions
- •LLMs are useful but often require iterative back-and-forth to reach best answers
- •Cultural shift: trainees adopt AI tools faster than prior generations
- 28:43 – 32:49
Shipping at enterprise quality: buyer personas, specialty variability, and “mutant” teams
Shiv explains how Abridge balances imperfect-but-useful product delivery with the high enterprise bar. He breaks down stakeholder needs (CMIO, CIO, CFO), the challenge of serving many specialties and settings (like the ED), and how interdisciplinary clinician-engineers accelerate iteration, backed by heavy R&D investment.
- •Enterprise sales requires satisfying CMIO, CIO, and CFO (often “2 of 3” to start)
- •Specialty-specific documentation styles and ED workflows complicate product design
- •Stitching discontinuous conversations is necessary in settings like emergency care
- •Clinician-engineers (“mutants”) and strong R&D spend enable faster iteration
- 32:49 – 38:42
Origin story and the “agency” insight: helping both patients and clinicians reclaim control
Sarah prompts Shiv’s personal journey into founding Abridge; Shiv traces the idea to deep learning’s rise and a formative patient encounter where a husband’s note-taking gave the patient presence and control. He connects that to clinician agency—reducing “pajama time”—and closes with real-world testimonials that reinforce the mission.
- •Shiv’s path: medicine, corporate VC, CMU ties, then founding Abridge
- •Key patient story: note-taking as a tool for presence and agency
- •Clinician agency: impossible workloads and after-hours documentation debt
- •Customer feedback loops include both critical edits and powerful outcomes stories
