Kollectiv AI
Late one night in San Francisco, long after his kids had gone to sleep, Andrey Akselrod sat at his desk staring at a small stack of papers that had come to define his year. They were hospital bills—simple on the surface, catastrophic in practice.
The visit itself had lasted only 24 hours. The aftermath stretched to twelve months.
Insurance wouldn’t pay. The hospital wouldn’t communicate. Phone calls spiraled into new phone calls. Conflicting statements turned into collections letters. It was a bureaucratic labyrinth with no map.
For most people, the nightmare would end with frustration. But Akselrod is not most people. He is a builder—twelve years straight of nonstop startup building, from Smartling to People.ai, scaling engineering teams and automating complex global workflows. He understands systems. He understands failure modes. And this system, he realized, was failing spectacularly.
“There has to be a better way,” he kept thinking.
It would take another year, and a serendipitous reconnection, before that thought became a company.
A Partner From Another Frontier
Across the Bay, Sasha Rohachova was living a very different chapter of her career—but she, too, was growing restless.
She had built one of the first augmented reality “virtual tattoo try-on” apps, Inkhunter, which exploded to 16 million downloads. She had led frontier AI and visual discovery teams at Pinterest, piloting 0-to-1 products in machine learning and computer vision. She spent her days immersed in innovation—yet she felt increasingly pulled toward a different challenge.
“I’ve always been unafraid of unfamiliar territory,” she says. “I started my AR company before I owned a smartphone. I moved to the U.S. before I knew how to speak English well. What matters is curiosity. What’s broken? What’s possible?”
One afternoon, a mutual investor reached out.
“You should talk to Andrey,” he told her. “Your skill sets might fit together.”
They reconnected. Talk turned into exploration. Exploration turned into alignment. They sat down with a “co-founder prenup,” the brutal list of questions every founding team should answer but almost never does. Ambition. Values. Conflict. Vision.
They matched.
But the partnership still needed a problem—one big enough to carry a decade of work.
Into the Back Office
The founders started sampling industries. Finance. Logistics. Consumer software. But nothing clicked until they stepped into the fluorescent-lit back rooms of dental practices.
If the hospital billing system was broken, dentistry’s was positively prehistoric.
Rohachova remembers shadowing billers for days.
“I was shocked,” she says. “Copy-pasting. Sitting on hold with insurance companies. Manually re-submitting claims. Hundreds of tiny workflows stitched together by human effort.”
The software was outdated. The processes were brittle. The labor inefficiencies were staggering. And most importantly—the pain was universal.
The founders interviewed hundreds of people: dentists, billing managers, RCM specialists. They sat silently in offices, mapping real-world workflows from the inside out. At one point, they even tried doing the work themselves.
“We needed to know where the pain lived,” Rohachova says. “And whether people would pay to eliminate it.”
They quickly learned something surprising: dental practices weren’t asking for AI. In many cases, they didn’t even have the vocabulary for it. What they were asking for was relief.
Building the Administrative Engine of the Future
Kollectiv AI was born from a simple but radical premise:
AI shouldn’t just recommend actions. It should take them.
This is the essence of agentic AI—software that performs work, not just produces output. And for dental revenue cycle management, the fit was uncanny.
Akselrod paints the picture:
“Insurance verification, claim filing, denial management, appeals, payment posting—it’s a chain of dozens of steps. If a human misses one, revenue is lost. An AI agent doesn’t miss steps.”
Kollectiv’s system logs into insurance portals, extracts benefits, checks claim statuses, identifies denials, writes appeals, posts payments, and reconciles accounts—autonomously.
It doesn’t ask. It acts.
It doesn’t fatigue. It improves.
And it doesn’t replace dental staff—it frees them.
“People don’t become dental assistants to spend their day on hold with insurance companies,” Rohachova says. “AI elevates their work. It removes the drudgery so humans can be humans again.”
Trust, Not Speed
But Akselrod knew something from the beginning: healthcare is not a place for Silicon Valley’s reckless mantra of “move fast and break things.”
“We handle sensitive patient information,” he says. “You don’t earn trust with slogans. You earn it with architecture.”
Every piece of Kollectiv’s platform is built with HIPAA compliance at the core: encrypted data flows, least-privilege access, audit trails, human-in-the-loop overrides, and explainability at every step.
AI doesn’t go unchecked. It escalates edge cases. It documents why it took each action. And ultimately, it becomes more trustworthy because it embraces supervision rather than dodging it.
The North Star: Fewer Interventions
While most companies measure ARR or customer growth, Kollectiv tracks something different: interventions.
It’s a metric borrowed from autonomous vehicles.
How often does the system need a human to take over?
Rohachova puts it simply:
“Our goal is true autonomy. The fewer interventions required, the closer we are.”
The company has already signed partnerships covering 100+ dental locations, including two dental groups generating $30–50 million in annual revenue. Pilot results suggest the platform is already faster—and more precise—than human teams.
But the founders are not rushing.
“This is a marathon,” Akselrod says. “We want the administrative engine of dentistry to run flawlessly. That takes time, rigor, and trust.”
The Future Is Administrative
If Kollectiv AI succeeds, the transformation won’t look futuristic. It will look almost invisible.
Patients won’t see AI agents behind the scenes verifying their insurance instantly.
Dentists won’t see AI resolving claim denials while they’re treating patients.
Office staff won’t see AI updating payment ledgers late at night.
But they will feel it:
Cleaner operations. Faster collections. Fewer headaches. More time with patients.
When the administrative engine runs itself, the care improves itself.
And maybe, one day, no one will spend a year fighting over a 24-hour hospital bill.
Interview Transcript:
Alan Olsen
Welcome to American Dreams. I’m here today with Andrey Akselrod and Sasha Rohachova. Welcome to today’s show.
Andrey Akselrod
Great to be here.
Alan Olsen
Both of you are based in Silicon Valley and are experienced, veteran entrepreneurs. I’m really looking forward to today’s conversation. We’re going to get into your new company, but before that, I want to spend a little bit of time reflecting on your backgrounds and how you both arrived where you are today.
Andrey, many listeners may know you as the founder of Smartling and former CTO of People.ai. What inspired you to start another venture, and why this time in the healthcare space?
Andrey Akselrod
I had been building companies nonstop for about 12 years. My transition from Smartling to People.ai was very quick. I took my family to France for a week, and that was essentially the only break I had between the two companies. It was 12 very intense years.
After I left People.ai, I took a sabbatical. During that time, as I was recharging and reflecting, I realized that we are living through one of the biggest technological shifts of our lifetimes, which is AI. During technological shifts like this, it is often the best time to create companies.
The first major shift I personally experienced was the creation of the internet. That was when companies like Google and Amazon were born. Then we had the mobile transition, which led to companies like Uber and WhatsApp. I believe the AI transition may be the biggest of them all because it is completely transforming how we work. I just couldn’t sit on the sidelines.
That is one part of the story. The other part relates to healthcare. A couple of years ago, I was in the hospital for just 24 hours. But that 24-hour hospital visit took me a full year to resolve the bills. The insurance company didn’t want to pay. The hospital didn’t want to communicate with them. It was a complete mess.
I kept thinking, “There has to be a better way.” That experience partially influenced our decision to move into the healthcare space.
Alan Olsen
Let’s move over to Sasha. Sasha, your career spans Pinterest, augmented reality, machine learning, and a Y Combinator startup. What personal and professional experiences led you toward founding Kollectiv AI?
Sasha Rohachova
That’s a good question. As you mentioned, I started with a Y Combinator company, and later I led the Innovation Lab at Pinterest. I have always stayed very close to frontier technology. In my first startup, it was augmented reality and computer vision. At Pinterest, it was machine learning and generative AI. Now, with Kollectiv AI, it is large language models and agentic AI.
Beyond staying close to frontier technology and being excited about what is happening in research and what may be possible next, I have also always been unafraid of unfamiliar territory.
For example, my first company was an augmented reality iOS app, but I started that company before I even owned a smartphone. I came to the United States to learn business before I knew how to speak English properly.
Similarly, with Kollectiv AI, I am entering a new territory and a new vertical. The process is about spotting what is broken, understanding it from the ground up, and then rebuilding it using frontier technology and a first-principles approach.
Alan Olsen
How did the two of you meet and decide to co-found this company? Was there a specific “aha” moment that brought you and your skill sets together?
Andrey Akselrod
Great companies are built by great founding teams. I believe the team matters the most, or at least close to the most. It is one of the most important factors in building a company.
I knew I needed a co-founder, and when you choose a co-founder, it is important that they complement you skill-wise. At the same time, you also need to be aligned on culture and ambition.
Sasha and I actually met briefly at a conference in 2016, so we were aware of each other, but we were not really in touch after that. Later, we were reintroduced by a mutual friend and investor. They said, “Andrey, you should talk to Sasha. Sasha, you should talk to Andrey.” They thought we might click, and we did.
I have a friend who is an executive coach, and she recommended a book called From Startup to Grown-Up by Alisa Cohn. One of the chapters talks about a “co-founder prenup,” which is a set of questions you are supposed to ask a potential co-founder to understand alignment and build a strong foundation.
When Sasha and I got together, we went through many of those questions and found that we were very compatible. We thought about building a company in a similar way.
Then we decided to test the partnership. We worked together for a few weeks, and it went well. At that point, we said, “We can build something great together.” That is how Kollectiv AI was formed.
Alan Olsen
Andrey, you mentioned your experience in the healthcare industry and how it took 12 months to reconcile expenses from a single hospital visit. With Kollectiv AI, you moved into the dental industry. Sasha, what pain point did you see that others were overlooking?
Sasha Rohachova
When we started the company, we knew we wanted to build our next big venture. We also had the conviction that we could build it together.
We explored multiple markets and multiple ideas. We personally reached out to and interviewed hundreds of people. At some point, we started exploring healthcare, including different verticals and different elements within healthcare.
Dental pulled us in. We spent months talking with dentists, RCM specialists, and back-office teams. We shadowed billers, and honestly, I was shocked by how outdated the software was. So much time was being wasted on copying and pasting, checking insurance information, waiting on hold, and losing money because the system was inefficient and broken.
We saw real pain. We also saw that new technology could solve that pain. On top of that, it was a sizable market, yet concentrated enough to build within. That combination made us very excited about the industry and influenced our decision to move into dental.
Alan Olsen
The whole AI industry is moving so quickly, and the technology is developing rapidly. What we say today may not be true tomorrow based on the problems being solved in real time. Give us a high-level pitch. What exactly does Kollectiv AI do, and what makes it different from other automation tools in healthcare revenue cycle management?
Andrey Akselrod
We actually think even bigger than revenue cycle management. Revenue cycle management is the first problem we are addressing, but we think of the company as automating the administrative engine of dental practices and large dental groups.
This is the work that keeps revenue flowing, but currently, it drains a lot of time and resources from practices.
Revenue cycle management is how dentists get paid. Imagine you want to set up an appointment with a dental office. You call, and one of the first things they ask is, “What is your insurance information?” Before you come in, the office needs to verify your insurance, make sure it is active, and determine what insurance covers and what you, as the patient, are responsible for.
Then you arrive, sit in the chair, and the work is done. At that point, the dental office needs to file a claim with the insurance company. The claim is either paid or denied. If it is denied, someone has to figure out why and potentially appeal the claim to get the money from the insurance company.
Once the claim is paid, the office has to take that information, put it back into the practice management system, and reconcile the accounts.
That is a simplified version of the workflow a dental office must go through to get paid by insurance. In reality, it is much more complex. That complexity is what we are automating.
If you imagine the dental office of the future, the doctor and staff are laser-focused on the patient and the patient’s health. Everything else — the administrative workflows — is handled automatically by AI agents, flawlessly, without errors, and using best practices.
That is what we do.
What makes us different is our AI agentic approach. Instead of simply suggesting things or simplifying workflows, the AI actually does the work. It logs into insurance portals, verifies insurance, checks claim statuses, identifies denials, writes appeals, and follows up until the last dollar that should be collected is actually collected.
We are replacing a lot of repetitive manual labor with always-on AI agents that unlock revenue for dental practices and improve efficiency.
Alan Olsen
In today’s world, when people think about AI agents, automation, and autonomous systems, they often worry that people within an organization may be displaced from the workflow.
How did you validate demand before writing a line of code? Were dental groups actively asking for AI-based solutions, or did you have to educate the market first?
Sasha Rohachova
I’ll take that question.
Overall, educating the market is not the best strategy. It is hard to force people to change. What we learned from previous experiences is that you have to deeply understand the customer. You need to understand who you are working with, what frustrates them, what they are already spending money on, and what gap exists between what they need and what is currently available in the market.
We spent months inside dental offices, shadowing back-office teams, mapping workflows, and talking to dentists and back-office staff. We interviewed many people. We also sat silently in offices and observed the work being done. Then we started trying to do parts of the work ourselves.
Once we were confident we had identified real gaps, we tested willingness to pay for specific projects. We would say, “Here is what we are going to build,” sometimes even showing visuals, and ask, “Would you be willing to pay for this?”
That gave us confidence that the solution was truly needed.
Andrey Akselrod
Sasha is really good at this — identifying the right questions, asking them in the right way, and figuring out the real answer.
It is not enough for people to say, “Yes, we are interested.” That often means nothing. The real question is whether they are willing to pay money for it. One of Sasha’s biggest strengths is getting that information out of people.
Alan Olsen
I think it can be difficult when you are trying to do something new that nobody has done before — introducing a solution that people immediately gravitate toward.
I remember a story from the accounting industry. I knew a very talented programmer who had connected the motherboard to a CD-ROM drive and helped revolutionize technology in that area. As he was looking at accounting, he said, “Accounting is so simple. I can write my own program and automate everything.” He said, “For $29, I can save you a lot of time.”
But sometimes educating the client or helping them move away from their current systems is not easy, because it requires change.
When you are dealing with measurable impact, how do you help bring your customers to understand the speed, error reduction, and cost savings of your solution?
Andrey Akselrod
At the end of the day, the proof is in the pudding. We have to put our software out there and measure the actual impact it generates.
We are making workflows faster, less expensive, and more precise compared with manual processes where people may make mistakes along the way. One of the most important things we are doing is ensuring that every dollar that should be collected is actually collected.
We are still very early in our journey as a company, so I do not want to quote specific numbers yet. But as we run our initial pilots, we are seeing meaningful improvements in the cost to collect for our partners.
Alan Olsen
Let’s move into the bigger picture of AI and healthcare. The idea of agentic AI is getting a lot of attention lately. How would you define agentic AI, and what makes it especially relevant to revenue cycle management?
Andrey Akselrod
Agentic AI is different because it is not just generating information or recommendations. It is actually taking action on your behalf.
Revenue cycle work is a long chain of small but critical actions taken one after another. A missed step or an incorrectly completed step can mean a lot of lost or delayed revenue. This type of workflow is a great fit for agentic AI.
Think of it as having an expert team member working alongside you. The AI does the work, and the human supervises the edge cases where the AI may struggle. The AI performs the work that needs to be done, while people oversee the process.
Alan Olsen
Sasha, many people worry about AI replacing people. How do you see Kollectiv AI’s technology changing, rather than eliminating, human roles in the dental office?
Sasha Rohachova
We hear that concern, too, and I think it is very natural for people to worry about change and the unknown. This new technology appears to be doing many smart things, and change is coming. The world is changing.
At the same time, I am a strong believer in human adaptability and the human ability to change and use new tools. I believe AI will help elevate our work, not simply replace it.
When people go into healthcare or become doctors, they are not doing it because they want to chase insurance companies. Even people working in the back office do not want to sit on hold for hours or copy and paste data from one system to another.
AI can help automate that manual, repetitive work. It is work that brings value, but it does not necessarily bring fulfillment. By automating it, staff can focus more on taking care of patients, connecting with them, and providing a better patient experience.
So yes, work will transform. Change is coming. But I believe it will change in a better direction.
Alan Olsen
Andrey, what are the biggest ethical or regulatory challenges you have faced, or expect to face, in applying AI to healthcare administration? How do you view that?
Andrey Akselrod
Alan, have you heard the expression “move fast and break things”? That is not what you do in healthcare.
In healthcare, you have to be very careful with how you handle data and patient information. One of the things we are doing from the ground up is carefully following HIPAA regulations and making sure private patient information stays private.
That involves many technical elements: encryption in transit, encryption at rest, least-privilege access control, audit trails, and more. We have to design the software with all of that built in from the beginning.
The other major component is oversight, accountability, and trust. Humans do not tend to trust AI right from the start. AI needs to earn its keep.
Part of earning that trust is allowing humans to supervise what is happening, set guardrails, and verify what the AI is doing. Another part is explainability. You want to understand what the agent is doing and why. And another part is escalation. If the AI determines that it needs a human in the loop, it escalates the situation and lets a human make the judgment call.
Those are major components of what we are building and how we are positioning ourselves with dental practices and larger dental organizations.
Alan Olsen
How would you define success for Kollectiv AI over the next 12 months? Where do you hope to be?
Sasha Rohachova
I’ll jump in.
Success over the next 12 months means delivering true autonomy for core RCM workflows, where the software runs without a lot of human intervention.
Think about autonomous cars, like Waymo or Tesla. Those companies track how many interventions are needed — how often a human driver has to take control when something goes wrong.
We want to track and understand how many interventions are needed as our software runs autonomously. Success also means not only doing this in one location, but scaling the impact across hundreds of locations, including dental practices and dental groups, with clear and measurable outcomes.
It is going to be an exciting next 12 months for us, with a lot of work ahead.
Andrey Akselrod
We already know that we are faster and more precise compared with humans. Now the question is how many interventions and exceptions occur. The goal is to reduce those to an absolute minimum over time.
Alan Olsen
It is always fun watching new companies launch as they continue to innovate and solve the problems in front of them.
Andrey and Sasha, thank you for being with us here on American Dreams. One last question: for someone who wants to reach out to Kollectiv AI, what is the best way to do that?
Andrey Akselrod
You can write to us directly. My email is andrey@kollectiv.ai. Kollectiv is spelled in a unique way: K-O-L-L-E-C-T-I-V, without the “e” at the end. So again, it is andrey@kollectiv.ai.
Sasha can also be reached at sasha@kollectiv.ai.
Alan Olsen
Thank you for being with us today.
Andrey Akselrod
Thank you.
Sasha Rohachova
Thank you, Alan.