In a recent 20VC interview with Harry Stebbings, Dr. Shiv Rao, founder and CEO of Abridge, explained how the company spent years in what he called the “five-year desert” before becoming one of the leading healthcare AI companies in the world.
Abridge is best known for using AI to capture doctor-patient conversations and turn them into clinical documentation. But Rao’s interview was not just about AI scribes. It was about market timing, trust, vertical AI, healthcare incentives, and why the doctor-patient conversation may be one of the most important data signals in medicine.
If you want to watch the actual interview it is here
Here are the 12 biggest takeaways from the interview.
1. Abridge never changed its core thesis
Rao said Abridge was willing to pivot almost everything except one central belief: healthcare runs on conversations between people.
The product changed. The go-to-market changed. The business model changed. But the core thesis stayed the same: the conversation between doctor and patient is a uniquely human signal, and a new healthcare infrastructure layer can be built on top of it.
That distinction matters for founders. A company can be flexible without being aimless. For Abridge, the non-negotiable idea was that clinical conversations would eventually become a foundational healthcare data source. The hard part was surviving long enough for the technology and the market to catch up.
2. The company had to survive the “five-year desert”
One of Rao’s most useful founder lessons is that being right early can still be painful.
Abridge believed in the doctor-patient conversation before the market was fully ready. For years, the company had to keep building, learning, selling, and staying alive without the explosive demand that arrived later.
Rao’s framing is simple: sometimes the job is just to not die. You need to still be standing when the market finally opens.
That is what happened when clinician burnout, health system budget pressure, and large language models all converged. Abridge had already spent years pre-selling and preparing the market. When demand arrived, the company could move quickly.
3. Burnout made the problem impossible to ignore
The interview repeatedly comes back to physician burnout and administrative burden.
Healthcare systems did not suddenly decide that documentation was annoying. Doctors had been drowning in notes, clicks, and after-hours charting for years. But coming out of the pandemic, the pain became too large to ignore.
Rao points to a market where a large share of clinicians were burned out, nurses were considering leaving the profession, and hospitals were under financial pressure. In that environment, a product that gave doctors time back was no longer a “nice to have.” It became a strategic necessity.
That is why Abridge’s timing mattered. The problem was old, but the willingness to act became new.
4. Vertical AI companies need to keep moving into the newest technology vintage
Rao described vertical AI companies as belonging to different technology “vintages”: post-transformer, post-LLM, post-agent, and so on.
His point is that a vertical AI company cannot get comfortable with the version of AI that made it successful. If the frontier moves, the company has to move with it.
That has product implications and organizational implications. The product has to become more capable as the models improve. The team also has to be willing to change how it builds, sells, and operates.
For healthcare, this is especially important because the market is slow-moving, regulated, and trust-sensitive. A company has to be technologically current without behaving recklessly.
5. Abridge does not want to fight foundation model companies
Rao’s view on OpenAI, Anthropic, and the frontier model labs is pragmatic: if you are fighting against them, you have probably already lost.
Abridge’s strategy is not to pretend the foundation models do not matter. It is to ride their tailwinds while owning the parts of the healthcare workflow that generic model companies are unlikely to own deeply.
That means integration, compliance, clinical workflow, trust, regulated data, enterprise deployment, and the context around how care actually gets delivered.
In other words, the model is not the whole company. In healthcare, the product is the model plus the workflow, data rights, deployment, safety, and trust layer around it.
6. The clinical note was the first product, not the whole opportunity
Abridge’s wedge is ambient clinical documentation: listen to the visit and help generate the note.
But Rao makes clear that the note is only the beginning. The doctor-patient conversation touches almost everything that happens after the visit: orders, follow-up, coding, billing, quality measurement, patient communication, and the medical record.
This is why the conversation is such a powerful wedge. It is present in every clinic, it captures the richest part of care, and it connects naturally to the rest of the workflow.
The lesson is broader than Abridge: the best healthcare AI wedges start with a real, high-frequency workflow and then earn the right to expand.
7. In healthcare, notes become bills
One of the sharpest business points in the interview is that clinical documentation is not just documentation. In the U.S. healthcare system, documentation often becomes the basis for reimbursement.
Rao says Abridge realized that the notes it generated would eventually become bills. That changed how the company thought about product architecture and enterprise value.
A product that begins as a doctor-facing time-saver can become important to the health system’s financial infrastructure. That is why Abridge has to speak to multiple buyers at once: the CMIO, the CIO, and the CFO.
Each buyer sees a different version of the problem. The doctor wants less burden. The clinical leader wants quality and safety. The CIO wants integration and security. The CFO wants documentation and reimbursement to work.
8. Abridge does not see Epic as the main competitor
Stebbings asked Rao how Abridge competes with Epic, given Epic’s dominant position in health system software.
Rao’s answer was that Abridge is not trying to be Epic. Epic is the medical record. Abridge is building an intelligence layer around the conversation and the workflows that flow from it.
That distinction is important. Many healthcare startups fail by trying to rip out entrenched infrastructure. Abridge’s strategy is different: integrate into the existing system, make the workflow better, and become valuable on top of the record rather than replacing it.
For healthcare AI companies, this may be the more realistic path. The winner is not always the company that replaces the system. Sometimes it is the company that makes the system usable.
9. Trust is the real speed limit in healthcare
Rao said healthcare moves at the speed of trust. That may be the most important line in the interview.
Abridge could make money in ways that might look attractive in the short term, including selling data, but Rao says the company refuses to do that. In healthcare, trust is not a slogan. It is a prerequisite.
Patients share deeply personal information. Physicians need to know that the tool will not compromise them. Health systems need compliance, security, and governance. A single trust failure can destroy years of progress.
Rao’s rule is that Abridge has to “earn the right” before expanding what it does with partner data. That is a useful standard for all healthcare AI companies.
10. Some AI tasks should use frontier models; others should move in-house
Abridge does not treat all AI work the same.
Rao described a split between tasks that are binary and relatively bounded, and tasks that are open-ended and never fully perfect. For some tasks, the goal is to build or fine-tune a fast, cheap, reliable in-house model. For others, the company should keep riding the frontier model wave as the technology improves.
This is a practical view of AI economics. Using the most powerful model for everything can be expensive and unnecessary. But using a weaker model for a task that requires deep reasoning or nuance can hurt the product.
The right answer depends on the workflow, the user experience, the cost structure, and the consequence of errors.
11. Rao would choose great people over temporary model access
Asked whether he would rather have early access to the best frontier models or six months with the best researchers and engineers, Rao chose the people.
His reasoning is that model-access advantages are temporary. A great team compounds. Strong researchers and engineers can build primitives, improve the product, and adapt as the underlying technology changes.
This is a useful corrective to the current AI market, where many companies pitch themselves around access to a model or partnership. Rao’s view is that the durable advantage is still talent, judgment, taste, and execution.
Model access is rented. Team quality compounds.
12. The bigger opportunity is moving healthcare from sick care toward prevention
Near the end of the interview, Rao zoomed out from documentation and described a bigger ambition: changing healthcare’s business model.
He described a system where providers, insurers, and life-sciences companies are siloed and often misaligned with the patient. The deeper promise of AI, in his view, is not only automating paperwork. It is helping create new care models that align incentives around prevention.
That connects directly to the distinction between sick care and health care. Much of the current system activates after people are already sick. A more aligned system would make it easier to intervene earlier, manage risk, and keep people healthier longer.
For NextMD, this is the most relevant part of the interview. Concierge medicine and direct primary care already move in that direction by giving physicians more time, smaller panels, and stronger patient relationships. Healthcare AI becomes most valuable when it supports that relationship rather than replacing it.
The Bottom Line
The Abridge story is not just a story about an AI scribe company reaching a $5.3 billion valuation. It is a story about how healthcare AI actually becomes valuable.
It starts with a real workflow. It respects trust. It integrates into the messy reality of healthcare. It gives doctors time back. It understands that documentation, billing, care quality, and patient experience are connected. And it uses AI to strengthen the doctor-patient relationship instead of trying to route around it.
That is the version of healthcare AI worth paying attention to.
Frequently Asked Questions
What was Shiv Rao’s main point in the 20VC interview?
Rao’s main point was that Abridge’s core thesis never changed: healthcare runs on conversations between doctors and patients. The company pivoted product, go-to-market, and business model over time, but kept building around that central belief.
What is Abridge?
Abridge is a healthcare AI company that helps capture doctor-patient conversations and turn them into clinical documentation and related workflows. It was founded by Dr. Shiv Rao, a cardiologist.
Why is Abridge important?
Abridge is important because it shows how vertical AI can work in healthcare: start with a real clinical workflow, earn trust, integrate into health systems, and expand from documentation into broader care and business processes.
What does Abridge’s rise mean for doctors?
The most immediate implication is less administrative burden. If AI can reduce documentation work, physicians can spend more time with patients and less time charting after hours.
What does this mean for concierge medicine and direct primary care?
Concierge medicine and direct primary care depend on time, trust, and relationship-based care. AI tools like Abridge are most relevant when they protect that relationship, reduce clerical work, and help physicians focus more on prevention and patient communication.
Sources
20VC with Harry Stebbings, “The Five Year Desert to Product Market Fit & a $5.3BN Valuation with Shiv Rao, Founder @ Abridge,” YouTube, published May 16, 2026. https://www.youtube.com/watch?v=byZkrYBF-N0
The Daily Summary, X post summarizing the 20VC interview, June 13, 2026. https://x.com/thedailysummary/status/2065875281626456363