Capturing 5% of Google Searches

Yeah, happy to. Thanks for having me, Mustafa. We started about ten years ago, actually as a decision support system for doctors — the original idea was really to reduce misdiagnosis. My two co-founders are both neuroscientists. One is a grandson of the German Nobel Prize-winning physicist Werner Heisenberg — not the Breaking Bad Heisenberg, the real one — and he'd done decades of research into how the human brain works and how we think. He looked at the way really good doctors make their diagnosis, and he calls it "constellatory thinking": a really good doctor, even as the patient walks in, already sees five, six, seven different things at the same time. He felt there was a way to use software to support doctors in doing that better, by applying some of the learnings of what really good doctors do. That was the starting point.

My other co-founder is a paediatrician and geneticist from the UK. She also founded a network of young doctors and medical students called Doctorpreneurs before Ada. As a geneticist she'd worked on finding really rare cases, using a dysmorphology database called the Winter-Baraitser database — which was still on CDs and that kind of thing. She was drawn to the idea of helping doctors who only have eight to ten minutes per patient and would struggle to hold the symptom constellations of 7,000-plus rare diseases under that time pressure.

So that was the original idea — we started in 2011. I was brought on as the business guy in this constellation. I think we do really well on the medical side; we can always do better on the business side. But we worked with doctors — professors in the UK from the early-cancer-diagnosis champions in the Royal College of General Practitioners, and the equivalent in Germany, first with neurology specialists and then GPs — and we tested the concept. We were encouraged early on by the chair of a specialist clinic in Germany, probably the leading clinic for vertigo and dizziness, which is often misdiagnosed. He told us a young doctor joining his department would need about half a year to reach the accuracy of our software. So we hired a lot of doctors, retrained them as medical programmers, and started building a massive medical knowledge base covering all of general medicine and hundreds of rare diseases. Today we cover about 30,000 ICD-10 codes.

But as you can imagine, convincing doctors to use new technology, with all the pressures and constraints they're under, wasn't easy ten years ago. My background is in consumer internet, and I'd always been convinced that patients really want this. My argument was simple: about 5% of all Google searches are health-related, which makes it the second-largest category of Google searches. I won't tell you what the largest category is — it also has to do with the human body, but not so much with health. And eight out of ten people in the Western world Google themselves before they go to the doctor. So there was a strong need on the patient side, and I'd always lobbied my co-founders to give the technology to patients too.

In early 2017 we launched our patient-facing app. Since then it's been downloaded over 11 million times, we have over 20 million cases in the system, and right now, every three seconds, someone somewhere in the world enters a new case in Ada. So we've evolved from working with specialists, to GPs, to patients directly — and now that we're offering the technology to patients, we're seeing more and more interest from health systems again. We're on the way to going full circle.

The concept of clinical decision support software had been around since probably the mid-eighties or even longer — there were systems, though of course not as usable; things weren't in the cloud. There was a lot more skepticism ten years ago than you'd see today, though it was always mixed. There were real enthusiasts — some doctors said, this is amazing, this is the future, I want this. And others said, why would I need this? I already have this in my head. It was a learning experience for me. Personally I come from a family of doctors — I'm the first black sheep who became a lawyer and a business person instead.

Some doctors who were initially quite skeptical tested it. We had one case with a doctor in Germany, probably in his mid-sixties, who clearly didn't believe in it. He had a patient he'd known for years who'd previously had episodes of abdominal pain — in two or three prior cases it had been diverticulitis. She came in complaining of abdominal pain, and he was very close to falling victim to premature closure: oh, I know this patient, it'll be the same thing. He was about to say diverticulitis and send her on her way. Then he remembered — oh, I have this thing from that weird software company, let me test it. He entered her symptoms, and our system suggested a kidney stone as the top priority. He had further tests done, and it was indeed a kidney stone. So he went from a doubter to a believer.

We needed those small successes. As you can imagine, the first few years of building a startup are extremely rocky, so when something like that happens it gives you the courage to carry on. And as we kept working over the years, we saw more and more such encouraging stories — that's still what drives us today.

The first year, all we really had was an idea and a prototype. There was a lot of testing, iteration, talking to patients and doctors, collaborating with clinical institutions, and continuous improvement. It was also an interesting experience to marry software engineering and medicine — and, eventually, to try to make a business out of it. My two co-founders, Claire and Martin, are extremely purpose-driven, and business certainly wasn't their top priority. We were lucky: we had very supportive, patient early investors. They weren't VCs — they were private, high-net-worth individuals who believed in what we were doing and have stuck with us through the years. Getting doctors and programmers to work together was also interesting. Now you see a lot of digital health companies, but in the early years there weren't many, and what you learn in medical school is different from what you learn when you learn to code.

We had doctors who turned into programmers, who said: I enjoy working directly with a patient, but if I can amplify that impact at scale — not just help 30 or 40 people a day, but contribute to helping 30 or 40,000, eventually a few hundred thousand a day — that gives me satisfaction too. Another challenge for me was understanding at least the basics of how different health systems work. In the UK you have a single-payer system, which people are rightly very proud of and is dear to the heart of Britain. In Germany you have a mixed private-public system, and then the US is a whole different story. If you think about how to develop a business out of all of this, you need to understand how the reimbursement models work and what motivates the different players. That was another challenge, and I'm still learning.

I don't think there's a clear-cut answer where I'd say we only want to focus on Germany, for instance — there's a bit more bureaucracy in Germany than in other countries. From the outside, having a national health service seems attractive: ideally you convince one top decision-maker and then apply a technology like Ada for the benefit of an entire population. But then I learned about the ins and outs of CCGs, AHSNs and trusts, and realised procurement very much happens at a local level, so it's certainly not that easy. And when you look at the US, they're approaching about 20% of GDP on healthcare spend, which probably isn't sustainable — so in theory they should be very interested in solutions that pursue the triple aim: optimising outcomes while lowering costs and providing a great patient experience. But again there are a lot of challenges. It's not one system either — you have a mix of nonprofit and for-profit systems, and all these PBMs; it's super complex. Still, based on where we come from, the US, the UK and Germany are our core markets — my co-founder Claire is from the UK and started the business here in Germany, and the US is just a huge market.

In addition, we've always had our eye on low- and middle-income countries — not so much for commercial reasons. The technology we've built lets you have a conversation with a bot, like a WhatsApp chat with your trusted family doctor, and at the end it gives you two things: an idea of what might be causing your problem — we never call it a diagnosis, because only doctors are allowed to make a diagnosis — and some advice on what next steps to take. More than a billion people in the world have very little access to doctors at all, but they'll increasingly have access to the internet and a cheap smartphone. So we think this is a fantastic way to bring more tailored, personalised, helpful health information to those who need it most.

In low- and middle-income countries the delta of improvement we can give an individual is actually greater than in the Western world, where we all have complaints about the health system but still have really good access to very qualified doctors by global standards. So we have this strange tension: on the one hand we need to focus on developing the business, and for that the wealthier countries are the focus; on the other hand, the whole reason we got into this is to help people, and we can make a bigger impact in low- and middle-income countries. We have partnerships — for instance with Fondation Botnar from Switzerland, which is focused on improving access to healthcare and education for children and adolescents. We have a project with them in Tanzania where we make our technology available to the population, which is why we have a version of Ada in Swahili. We also work with them in Romania, and we strive to empower community health workers in such countries. And we have a project with the Rockefeller Foundation in South Africa, where we aim to help young mothers and teenage pregnant women. That's always a big part of what we do.

That's a great question. I've heard that quote so many times over the first few years. On the one hand, I believe in the lean startup model — you build an MVP, you iterate. My co-founder Martin believed more in: I have this genius idea, and eventually people will realise how amazing it is; you can't just ask people, because they'll only tell you what they already know. The truth is probably somewhere in the middle, and we had a mix of both approaches.

What didn't surprise me at all was that in our conversations with doctors, the predominant theme was that the things they were looking for weren't really what we were building. Our goal was to help them make a better diagnosis, but their problems were much more mundane — and understandably so. They'd say: how do I deal with all this bureaucracy? Why do I have to enter the same data in three different systems? I just want to save time.

It was actually quite obvious to me early on, because I grew up in a doctor's household, that doctors have to deal with so much admin. Their feedback was clear: your product's amazing and it would help us make a better diagnosis, but we can't use it with every patient because we don't have time to enter the data. In Germany you already have to enter data into your practice-management system so you get paid — and those systems weren't built to benefit patients or improve diagnosis, they were built to let doctors get paid. Very old legacy software, so there was little hope of connecting our cloud-based, modern software to these legacy private-practice EHR systems. We learned a lot from talking to doctors, but to me it wasn't surprising. I was quite sure that if you want to change how healthcare is delivered, the change would come from the patient side, not originally from the doctor side.

My hypothesis was that we need to give patients something that empowers them and changes the information asymmetry you traditionally had. Eric Topol, the US cardiologist, described this really well in his book The Patient Will See You Now — a play on waiting for hours in the doctor's office until someone finally says, "the doctor will see you now," and you say, "oh, thank you so much." He describes how modern technology will eventually help patients sometimes know a lot more about their own condition than the doctor possibly could. The patient is always the person who should be most motivated to find out what their problem is and how best to address it.

This is not a knock on doctors. The people I admire most in terms of day-to-day work are doctors and nurses — you have to have a fantastic ethos and a caring mindset to even go into that profession. You see it now in the pandemic, where health workers are finally getting some of the recognition they deserve as heroes — but they were doing it before the pandemic too. It's just that there's a limit to how much a doctor can hold in their head at the same time when assessing a patient. They have time pressure, incomplete information, and the patient also feels under pressure — I've experienced this as a patient, where some doctors are always under time pressure and want to get you out the door as soon as possible, so there's even pressure to describe what you actually have.

So with the solution we've built now, we've incorporated some of the learnings from what doctors told us. The doctor doesn't need to enter the information — the patient enters it in their own time. At the end of what we call a pre-assessment — an automated way of taking the patient history — the outcome can be shared with the health professional, combined with hints on what the system thinks the most likely diagnosis might be and what tests, examinations and investigations to consider. So we're addressing the doctors' concern that they don't have time to enter the information, by letting the patient enter it, and then giving the doctor that information before they've exchanged the first word with the patient. The doctor doesn't start from a blank sheet — they're pre-briefed, which has the potential to save them time.

They can double-check with the patient, but it can also be fed into the EHR system the doctor uses, so a big part of the documentation is already done too. Then it starts becoming good not only for the patient but also for the doctor. We're already working on this concept in practice — for instance with large health systems in the US such as Sutter Health in the Bay Area. They have about 10 million patient consultations a year, and they've integrated us as the digital front door in their own website and app. The chief patient experience officer of Sutter Health, who's still a practising doctor, Dr Albert Chan, told me he's personally had patients who told him it saved their life — they used the Ada technology embedded in Sutter's website and app, and said if they hadn't, they wouldn't have gone to the hospital; they'd have thought it probably wasn't that serious. But because they used it, and it pointed them to see a doctor right away, they did. That's really encouraging for us.

We did actually build an Alexa version of Ada a few years ago. It's interesting — I don't know if this is publicly known, but a Cambridge-based computer scientist who's one of the world's top experts on artificial intelligence created a company called Evi, which he sold to Amazon and which later became the AI behind Amazon Alexa. His name is William Tunstall-Pedoe. He's an investor in Ada and an advisor to us — so we have a direct connection to basically the creator of Alexa.

He'd actually looked at some other companies in the UK trying to do similar things to us, but concluded we were years ahead, invested in Ada and has been advising us. So we did create an Alexa version — but as exciting as it sounds, it's actually quite tricky. If you've ever been stuck in an automated telephone hotline with British Airways or some travel provider, you want to pull your hair out, because it reads all the options to you and it takes a lot longer. Sometimes just tapping on something is easier than having options read to you and then responding. We're still looking at the concept, and eventually the vision you're describing could become true.

But because it's healthcare and we have to be really careful to avoid any potential source of error — even within the UK, all these different accents; and imagine I go to the UK with my German accent, tell the system something, it comes out wrong, and I get completely different treatment based on that. That would be scary. So we're still looking at voice-based interaction models as well.

The number 150 is quite accurate. When we started we were about 12 people; we quite quickly grew to about 50.

Then for several years we were around 100, and that was still quite manageable. We had almost no attrition — people were really loyal; it was more like a family setting. We were mostly working out of a grungy Berlin backyard office, but we had a lot of fun and were a tightly knit group. After some funding rounds we did probably go too fast, in hindsight — we wouldn't be the first to make that mistake. It's exactly around that number, 150, when I didn't know everyone anymore. Before that I really knew a lot about each individual I worked with — even some of their private concerns — and you can be much more understanding if someone can't perform at work for some reason. As long as you're a smaller team it's a much more personal relationship.

When you try to grow fast and blitzscale, things happen where you come into the office as the guy running the company and you see people and don't know whether they're visitors or work here. That's when you have to be careful it doesn't slip away from you. We did have one situation, early last year, where for the only time in our history we had to make a correction — we did some restructuring, and that was incredibly painful. It felt really bad, but it had to be done. We reduced the number of people, and since then we've been a lot more focused, and more focused on advancing the business side, which is the stage we're at now. I feel we're in a better place, but it's a big challenge. Scaling a company is hard.

What I've learned is that you can't do everything — you need a management team that supports each other and where you can trust each other; that's the only way to build a successful company. And preserving the friendly culture doesn't come for free; it's not automatic. Once you get to that level of scale, you can't automatically expect everyone to be friends and understanding of each other's mistakes and failures, including mine. You have to put some processes in place, and at the same time try not to become too corporate — still preserve a friendly culture. It takes a lot of work. Over the last year we've implemented tools to more systematically collect the team's feedback, where it used to be done around the water cooler or over a beer.

We'd very often stay in the office till late and have drinks together — that used to be the culture. Now, certainly in times of COVID, that's difficult, so we use all the tools we can to at least make sure we have a virtual feedback culture. I take this very seriously, and I recognise things can always be done better, so I see it as a continuous learning process for myself as well.

For me personally, the hardest thing is when you have to make the difficult decisions — like letting someone go when they haven't done anything wrong. If someone's been an ass, then it's different — pardon my French — but if someone really hasn't done anything wrong and it's just not the right fit for the job, and then you have to communicate that decision, I'll never be in a situation where I can easily do it. It's always really difficult for me. If you run a startup there are a lot of other hard things — the first thing you worry about is being able to pay the bills and salaries; and if you want to make and amplify an impact, eventually reaching a billion people, you also have to build a sustainable, successful business, and for that you often need to be on the fundraising trail, pitching to potential partners. All that is part of the job, and I don't find it so hard. The hardest thing is communicating painful decisions to people who haven't done anything wrong.

It's always a work in progress. When it comes to toning down, it goes back to what I described: when you work with people who know you really well, there's a lot of banter — you make a silly joke and people know you don't mean it in a bad way. But suddenly you're 250 people, and someone who's never talked to you hears you make a silly joke and doesn't know how to take it — they think you've done something incredibly politically incorrect. So I've had to tone down the being-silly part, which is sometimes still hard.

I guess you learn a bit from your experiences, especially from your mistakes. Overall I've become a little more serious — not totally serious — but you realise over time that it's not just a game, not just a joke, when you're responsible for more than 200 people's salaries, and responsible to investors who've put their faith in you, and, in our case, responsible for making sure you've done everything you can to provide the safest and most accurate solution for patients. There's a lot more responsibility than when you're a student and can do all kinds of irresponsible things. That's just life.

For me it's a serious goal, not just a North Star. As I mentioned, about 5% of all Google searches are health-related, so we're addressing a need that pretty much everyone will have at some point — people have a health issue and want to find out what's wrong and what's causing it. The addressable population is basically everyone in the world, and we happen to have built the most accurate solution for finding out what's going on with your health. When you're in that situation, it's almost a moral imperative to bring the solution to as many people as you can. There can be various ways to increase your reach, but I'm so deeply convinced the solution we've built will be helpful to many people.

My personal goal, and the impact I measure, is: how can we scale the impact? We know we're already making an impact. Ada is the medical app globally with the most five-star reviews ever — more than 300,000 in the App Store and Google Play Store. Almost every day someone says, they saved my child's life, they saved my father's life; without Ada we would not have found out and would not have taken action. Or people say: I have a rare disease, it took more than 10 years and more than 40 doctors to get my diagnosis; finally I found the specialist who diagnosed me, and I just wanted to test Ada — and now I'm sitting here crying because Ada told me in less than five minutes what more than 40 doctors needed over 10 years for. When you get that kind of feedback, you think, okay, we need to keep working on this and bring it to as many people as possible — eventually through partnerships, or whichever way we make it work. That's a worthwhile goal to strive for.

If you ask the people around me, there are probably a lot more habits that haven't helped me. But I've been told I work better with my back to the wall — I don't think that's a compliment, but unfortunately it seems to be the case. Even as a student I did my studying the night before the exam, so that's not a habit that helps. If I think of something positive, it would be realising you're not the smartest person in the room for everything. One thing I realised very early at Ada is that I work with incredibly brilliant people, and just trusting their judgment and letting them do their work is probably the best thing that's helped me in my career. So the best thing I can do is stay out of the way.

A recent one I've read — twice, actually — is The Hard Thing About Hard Things by Ben Horowitz. It resonated so much with all the challenges of running a startup. When you run a startup you go from conquering the world to complete and utter desperation, sometimes three times within a day, and it's a really interesting book that describes that struggle. Over the years, building a startup is very rarely a completely straight line towards success — it's always small successes that keep you going, and you can't be too dejected by setbacks.

It's a great book, and if you want to run a startup I recommend reading it first — maybe then you won't want to run a startup anymore. But I have to say it's by far the most rewarding thing I've done in my career. As I mentioned, I trained as a lawyer and worked as a strategy consultant, but working on something with a real purpose, that has the potential to make a positive impact on people, is something I can recommend to anyone. At least at the end of the day you feel good about what you're doing, and that's a really strong motivator.