Building Successful Startups: Focus On The Problem, Then The Technology

spotify

Doug Williams leads product and generative AI work on Orbit, the Martin Trust Center for MIT Entrepreneurship’s platform for founders, and he has ridden several technology waves as an engineering leader: an education startup sold to Pearson, mobile at Zipcar and RunKeeper, and more recently healthcare. In this conversation with host Ayush Jain, he makes one argument from several angles: start with the problem and the customer, and let the technology follow.

He explains how he chooses where to work (mission, people, timing), why a startup has to find the place it can be world class, and how RunKeeper drifted into a general health app before refocusing on beginner and intermediate runners. On AI, he treats large language models as a platform to build on, the way founders built on iOS and Android, and argues for tools that assist a human rather than replace one. In healthcare, that means a person checking the output. He closes with a call for smaller, more efficient models, because a technology that doubles our energy use hasn’t moved the bar forward.

Highlights

Problem first, technology second

Doug’s advice to anyone starting something new is to work out what problem you’re solving and how technology can help, instead of getting enamored with the technology and figuring out the use later. In the week-long bootcamp he had just run for Ocean MBA founders, the questions were the same ones he’d ask a team inside a large company: who is your target market, who is your beachhead, and how does this improve on the status quo? “That’s really the whole game,” he says, “and then the technology comes afterwards.” It’s the same order of work we follow in healthcare product development that starts from a validated problem: who has the problem, how they handle it today, and what has to change before anything gets built.

Focus, and a path to world class

When he decides where to work, Doug checks three things: does he care about the mission, does he like the people, and is the timing right. People often tell him they had an idea two years earlier; his answer is that it’s all about timing. He also pushes startups to ask where they will be world class. They might not start there, but they need a path to it, because a team spread across everything is only okay at all of it.

When building becomes a substitute for market knowledge

RunKeeper is his worked example. The app began as a running app, then added biking and other sports until it was “a good general health app as opposed to a great running app.” The team, mostly fast former runners, had assumed the user was like them. When the company refocused, it went out and talked to people on the street and found its real market: beginner and intermediate runners. RunKeeper became “a gateway to running,” the segment ASICS was most interested in when it bought the company. His lesson: “You don’t get smarter talking to your own team. You get smarter talking to customers.”

Build on large language models, and keep a human in the loop

Doug sees large language models becoming a dependable layer to build on, much as iOS, Android and cloud infrastructure were for the mobile wave. His team found the models worked far better when guided step by step through MIT’s 24-step disciplined entrepreneurship framework than when steps were blended. He is wary of AI that removes the human: “the only software without bugs is not used,” but bugs are manageable when a person is there to catch them. In healthcare, he argues, published code systems such as SNOMED CT and LOINC let AI pull information together, as long as a clinician examines the result. That principle runs through how teams approach building LLM applications for regulated work, and it echoes Munjal Shah’s case on this show for a safety-focused large language model in healthcare.

Efficient models matter as much as capable ones

Looking three to five years out, Doug hopes for progress on efficiency: companies working to make large language models smaller and cheaper to run, and smaller models for specific tasks that run on phones and laptops. “Shame on us if we invent this new technology that doubles our energy consumption,” he says. “We haven’t really moved the bar forward.”

  • Full Transcript
  • Ayush Jain [00:19]

    Hello everyone, welcome to another episode of HealthTech with Purpose. Today we are joined by a true leader in the world of innovation and entrepreneurship, Doug Williams. Doug isn’t just passionate about building successful companies; he’s driven by a mission to use technology to tackle some of the world’s biggest challenges, from healthcare equity to affordable housing. Join us as we explore Doug’s journey and delve into some fascinating topics like the intersection of startups and generative AI (GenAI), the power of Fast Healthcare Interoperability Resources (FHIR) interoperability, and the future of AI in healthcare. Hello Doug, how are you today?

    Doug Williams [00:51]

    Good, Ayush. Nice to see you.

    Ayush Jain [00:53]

    Same here. So Doug, why don’t we start with a quick introduction about yourself. I did mention some of the accolades, but I would like to know from you the different things you have been involved in, maybe in chronological order. You could start with where you started, where you grew up, and what led to the different things you were involved in and how you grew them.

    Doug Williams [01:23]

    Yeah, sounds good. In the questions you had “where did you start your journey,” so we can start there. I have the good fortune of being from a family of engineers: both my grandfathers, my dad, as well as folks involved in healthcare, my wife’s dad, my sister, my mom. So I’ve got a technology and healthcare background, which I’m very lucky to have.

    You and I are similar generations; I might be a little ahead of you. But we’ve had the good fortune of the age of electronics, growing up and seeing all the developments that have been basically continuous. As I was growing up I had the Radio Shack electronics kits, the chemistry kits, all very accessible, and I thought it would be a fun way to have a career if I could take these new developments and figure out ways to use them in useful ways. That’s basically what I’ve been able to do, and it’s been a great ride. Along those lines I’ve been able to ride successive waves.

    I was also interested in startups. Large companies are critically important to the world; we need innovation at large companies. But it’s a bit of a cheat to do startups, because you can start small, and if you’re focusing on your customers and thinking about how to use the tech, you can go from small to bigger, which I’ve been able to do. I’ve been able to ride successive waves: the internet, with an education startup we sold to Pearson, the largest education publisher; then the mobile waves, as head of engineering for Zipcar and RunKeeper; and most recently I’ve been focused on healthcare, which is where we intersect. I’m trying to use what we learned in the consumer spaces and apply it to healthcare. We’re all used to consumer apps; that’s the bar. In healthcare we want to raise the bar to have similar capabilities, whether you’re in a hospital or you’re a patient. That’s ultimately where I’m focused.

    Then at MIT, they asked me to come in and build tools for startups. I’ve done multiple startups, and now I’m a kid in a candy store, getting to build tools for entrepreneurs using the latest technology. The latest wave is the generative AI wave, and we’re using the latest AI tools to build what is turning out to be quite useful tools for entrepreneurs, built on MIT’s disciplined entrepreneurship and startup tactics frameworks. So that’s me; that’s a monologue on it. But I’m very lucky to have grown up during this time. Now I feel like our generation needs to help sponsor the new folks coming in to fix the problems that we’ve left: climate, healthcare. I just came out of an MIT session with the Ocean MBAs, looking at climate tech. It’s a great place to be, and we’re trying to make these tools available ultimately worldwide, which is consistent with MIT’s mission.

    Ayush Jain [05:47]

    Right. So, very interesting, Doug. I’ve seen the versatility in your experience, and I think the agility to change careers that you have shown so well throughout your career is really commendable, because making those choices and going with the possibilities is equally important for one to grow. Just like now, you have joined the hotbed of GenAI, from what I can see, at Orbit, MIT’s entrepreneurship platform. So tell us more about that, and then also elaborate on your view of GenAI, because a lot is happening with AI today, and we would love to learn your insights into it and what’s coming in the future.

    Doug Williams [06:48]

    Yeah. I have had the good fortune to work across a number of successful ventures. I feel like I’ve been able to assist, but essentially I’ve been able to find people who had mission-driven startup companies, and what I’ve been able to do is pattern match really, really well. I’ve basically been able to find places that really felt like they could go forward, and then my main goal was to help keep pushing the opportunity up and to the right. That’s made it easy to move around. So my agility is basically knowing a lot of other smart people and collaborating with them.

    Certainly that was the case here at MIT. When the GenAI wave hit, we were all trying to figure out where to get our feet on the ground. When I feel like I’m a little unmoored, I usually begin to think who can help me get moored, and I began reaching out. Certainly at MIT there are a lot of people, so you look for people you want to work with who can help. And now I’ve found a huge group of people that we can help.

    But the critical part that I would impart to anybody is: think about what problems you’re trying to solve, and then how the technology can work toward solving those problems, as opposed to getting enamored with the technology, developing it, and then later trying to figure it out. We just came out of a session; we have a week-long bootcamp for Ocean MBAs, and we emphasized really thinking about who’s your target market, who’s your beachhead market. Essentially, with any new technology, where can you be great, where can you make a difference? And this is true at large companies. If you’re doing a new initiative: who’s your customer, what problems are you solving, how is this going to improve on the status quo? I know as you work with your clients, you try to steer them to the same thing, and you look for clients that are thinking that way. That’s really the whole game, and then the technology comes afterwards.

    Ayush Jain [09:53]

    Right, right. So how you’re making a difference, really. You cannot start with the technology; you have to start with the problem, and then ask whether there is technology, or whether you can make a technology, that can solve it.

    Doug Williams [10:08]

    Usually, always, at least for me: starting with what the problem is, and then figuring out how the technology can make a difference, and riding along as, say, mobile applications come to the forefront: how could we apply those? So I keep an outlook for people who are doing things like that.

    For me, the first piece is: do I care about the mission of the company? I would strongly encourage people to ask that. Do I care about it? Will my friends care about it? If I’m helping recruit others, would I be proud to bring them in? I think those are critical. And do I think it can be successful? If you bring all your friends to a place that can’t be successful, that’s hard for everybody. So I get an assessment of what I think the prospects are, how the entity can make money, and whether I agree with the mission. That’s first.

    Then I figure out: do I like the people involved? Is this somebody I want to work with? Because you can be with people you want to work with and be going down a path where you’re five years too early. People often say, “I had that idea two years ago.” My answer is inevitably: that’s great, it’s all about timing. So is the timing right? The right people, the right timing, and then you can build around those things. That’s what I’m looking for: that fit.

    I was a runner, so RunKeeper was a natural fit. That was also a kid-in-the-candy-store situation. I’ve been a startup person, and working here at the Trust Center, providing tools at MIT but also beginning to make them available worldwide, is super appealing. Now we’re at the stage of: are they really giving useful outputs? And the fun thing for us is they are beginning to give very useful outputs. I can speak more about what we’re doing, but that’s basically the recipe I follow. I think you’ve done similarly. You’ve steered more to healthcare because you see the need and the expertise. You can’t be great at everything. When you do work, you want you and your company to be great.

    Ayush Jain [12:43]

    And similarly, in order to go deep, you have to focus, because otherwise you are just spread too thin.

    Doug Williams [12:51]

    Yeah, and you can’t be great. If you’re just coding, you’re just okay across things. Like you all have focused on healthcare: you can get great in it. That’s what I’ve tried to do, and that’s what I always recommend for startups: where are you going to be world class, and do you see a path to it? You might not start there, but you should have a path to it.

    Ayush Jain [13:18]

    Right. That makes so much sense, Doug. But I’d like to touch upon the Ocean MBA. I would like to learn more about what’s happening there, especially around climate, because we know we’re sitting on a time bomb, and most of us really don’t have any control over it or know what we can really do. So tell us more, since you have been deep into that recently.

    Doug Williams [13:48]

    So I’m at the Martin Trust Center for MIT Entrepreneurship, the entrepreneurship center, and our thinking is that we’re going to need entrepreneurs, entrepreneurs like you, Ayush, and entrepreneurs like the Ocean MBAs, to solve our biggest problems. We’re going to need that level of innovation. So we’re really encouraging mission-driven startups wherever possible. Climate is one of the big initiatives; MIT has just launched a climate center, for exactly the reason that it’s an existential threat. These are folks with a whole range of different technologies: biotech, microalgae for more efficient aquaculture, new battery technologies, carbon capture, a whole range of different solutions. We try to segment those…

    [Sponsor message] [14:53]

    This episode of the HealthTech with Purpose podcast is brought to you by Mindbowser, a product engineering and digital transformation company focused on health. At Mindbowser, we enable health companies to build the future of health with accessible, equitable [inaudible] patient care. We strongly believe that technology can empower a healthier world, and that’s why we are partnering with healthcare experts like you to make it happen.

    Hi guys, thanks for listening to HealthTech with Purpose. Make sure to subscribe to us on Spotify, Apple, and YouTube for more expert insights on industry innovation and transformation. And remember: share the knowledge.

    Doug Williams [15:33]

    …solutions, and then ask them to further segment it. Our role, we feel, is to give them the best tools. I just presented our AI tools to them, and we’re beginning to refine our tools to specialize for things like climate and healthcare. So we’re trying to give them the best education we can, and then begin to build communities. This Ocean MBA is a bootcamp for people in climate startups to meet each other and investors and get MIT-style training. And as we work on this, it’s a worldwide problem. MIT’s mission is to serve MIT and the world, so we’re looking at where we can take our learnings and make them available worldwide, which is why I’m happy to talk to you. You’ve got a worldwide perspective.

    Ayush Jain [16:46]

    Right, right, true. Like you said, it’s not a problem of a particular area or a country; it’s the world’s biggest problem, in fact. Just to share: we as a company just became carbon neutral certified this year. That’s one of the values I took on when I went to Berkeley: while work is important, you have to still do it in a way that is sustainable and good to the environment and good to the people. All these values may look small, but they make a lot of difference to how you live your life, build your company or build things. So I’m always a big proponent of the whole cause, because I can see it’s such a big problem that has to be solved, and we can at least do small bits wherever and whenever possible.

    Doug Williams [17:46]

    That’s right. And we encourage the students to be thinking that way: how can they make a difference in the world? I think each of us has to do our part. And then we’re also looking for these really smart innovators to come up with ideas that we may not have thought of. Those are the kinds of things we’re doing with the climate MBAs, for instance.

    Ayush Jain [18:17]

    Right. What do you see as the recipe for making these startups really make a mark? Of course, everybody starts with a promise of making a world-class product or making a dent in the world. But what could be the repeatable lessons that you would like to share for student startups, not only at MIT but throughout the world? Like you said, we all need new entrepreneurs and new thinking, and students and young people are our hope. So how can we all do better at it, how can we create that ecosystem, or what lessons do we need to instill in the young ones?

    Doug Williams [18:58]

    Yeah, great, great question, and we think about that a lot, because we’re the center here at MIT and we feel a responsibility here and beyond. We’re very lucky. You were at Berkeley; that’s another hotbed where you’ve got all the elements. But how do we make those available worldwide? I’ve worked with people in Pune and Shanghai; those are other hotbeds. The US doesn’t have a monopoly on these things, but we do want to share best practices of what we learned.

    The elements: certainly leadership. Ayush, you provide that for your organization. I think having strong, committed leadership, ideally more about the mission and less about the ego, is critical. That’s what investors are looking for. Folks can run the risk of looking at the technology first. Often, with today’s technology, a lot of problems can be solved much less expensively. That’s not always true in climate, which is what we’re trying to help sponsor, but it’s certainly true in software, and I expect you’ve seen that.

    So we really work on focusing on the customer solution first and the building of it later. From a technology standpoint, I’m pretty confident, and I think you are too, Ayush, that if we’re asked to build something, we can build almost anything. It’s really just the amount of time and resources, and more resources isn’t always better. So it’s a tight focus and really understanding the problem to be solved and the priorities of the problem to be solved. If you have a big problem, you can’t eat an elephant in one bite; you have to eat it a step at a time. Those basic understandings, which in software I think we understand, really help.

    As you’re solving problems, understanding investors who are also interested in the idea is critical; that does become the fuel. So having early conversations and relationships is important if you’re building a company on the innovation side, but also look at how far you can get on your own investment; you’ll have more control. At RunKeeper, Jason was able to build a lot out small, and so he ultimately had a lot of final say until we sold it to ASICS. You must be thinking that way too: how much you can self-fund. But it all comes back to looking at the problem, then looking at the market opportunity, and providing an economic map along with a technology map. That’s the key.

    Ayush Jain [22:21]

    Right, right, true that. We often say in technology that building is easy, but finding the market fit, or the use case benefit, is what’s important and really becomes the key to finding success. Technology today is definitely far easier, if I can say so, compared to how difficult it was to build something 10 or 20 years back. But you still have to make sure that what you’re building is really fit for the audience, or for the problem that you’re solving.

    Doug Williams [23:02]

    And we still get surprised all the time, and that’s the risk. It’s fairly easy to build an app, so you get started before you’ve focused on it. RunKeeper would be a good example. Obviously the early focus was running. As we began to get bigger, we began adding other sports, to the point of biking, since a lot of runners do cross-training. But we found ourselves developing more of a general health app: we were a good general health app as opposed to a great running app. So we had to refocus.

    We found the employees at RunKeeper tended to be former runners. Their idea of running is six-to-eight-minute miles. I started there, but as I’ve gotten older, I’ve evolved into our target market, which we found was actually beginner to intermediate runners. When we finally refocused ourselves on running, that was the market. If you’re going to carry a phone, you’re more of a beginner to intermediate runner, and we became a gateway to running, which was a fabulous place to be. That’s what ASICS was most interested in: that broader segment. So even at RunKeeper we had to redouble our efforts and refocus on running, and that’s a message I think you would want for your clients as well. It can sometimes be too easy to build, and the action of building becomes a substitute for real market knowledge.

    So when we focused back on running, we had to go back out. We actually accosted people on the street and asked them, which is commendable for our user experience department. Here we’re always talking about focusing on the customer. You don’t get smarter talking to your own team; you get smarter talking to customers.

    Ayush Jain [25:19]

    So my last question for today, Doug: since you are in such a perfect place in terms of where innovation, intelligence and technology are coming together, what are your big bets or big ideas that you’re counting on, or the great themes that you see coming in the next three to five years to make the world better?

    Doug Williams [25:48]

    Yeah. In the near term, we’ve been really heartened. Everyone’s looking at the large language models (LLMs) and trying them out, and they become tools available for all of us to build on. Just like the mobile technologies: we could build on Android and iOS, we could count on those and the associated hardware, and we could count on the cloud infrastructure. We can now count on the large language models, which will be in an arms race to build the best possible. So for those of us who are not at the large tech companies, we can apply the technology, and it really comes down to thinking through the same things we’ve been talking about: where you apply it.

    We found great success. We have a 24-step entrepreneurship framework. It’s a lot of steps, but it’s the right steps: thinking about the customer, thinking about the finance. I think you’ve looked at it. We’ve pulled the best practices across a range of books and research. What we’ve been surprised at is how well the models work when we focus them step by step by step. When we were blending steps, they got confused, but our framework really helps. So think about problems like that. We have an alumnus working in the Australian railway system; they could do the same things. They have processed documents they could automate.

    In my mind, particularly on the AI side, it’s about thinking how to assist a human rather than replace a human. A lot of us have been surprised at the push for self-driving cars, but the technology to assist drivers is fantastic. Taking the driver out of it, as a software engineer, concerns me. The only software without bugs is software that’s not used. But you can have bugs as long as a human is there to catch them, and you can push the boundaries. So when we provide AI assistance for entrepreneurs, there’s no substitute for the entrepreneur doing their own research, but boy, you can provide assistance. I would encourage people to think about where they could provide assistance rather than replacing people: all the processed documents you have at large corporations.

    Healthcare, which is the area both of us are focused on, is ripe for that. The world’s corpus of healthcare information has been put online. If you’re a nurse or a doctor, you’re doing Google searches today. That data is very accurate. All the coding systems, SNOMED codes for diagnosis, LOINC codes for lab results, are published and continually published. AI can do a great job of pulling that together, as long as we’re thinking that there’s a human then examining the results and finding ways to continue to make the results better. That will happen somewhat automatically as we train the models, but also as you’re doing projects in healthcare, you want to keep thinking about how you help a human help a patient. In those areas, I think we can be fairly aggressive.

    I’ve been involved in the FHIR work for standardizing healthcare, beginning to think about how we can do automation, but all with the idea of making a better experience for the patient. If we can stay patient focused and keep humans in there, not just trying to automate, I think we’ll avoid mistakes. I’ve done a bunch of automation projects, literally in factories, and I’ve had a few where I had to back off the automation. I inherited the first line where we tried to take too many humans out, was asked to improve it, and I put more humans back in. So the goal should be to assist our fellow humans. If you do that, I think the new technologies can be used aggressively, because there’s someone looking at the outputs and keeping focus on how you’re making the world better with what you’re doing. That’s my take.

    Ayush Jain [31:00]

    Okay, well said. I think it connects back to the foundation piece: there are still these foundational problems that need to be solved, and from those, some of the great companies, startups or ideas will come out.

    Doug Williams [31:19]

    And you asked where I think things are going. We’ve got the big, well-known ones, the Anthropics or Perplexities. I personally am hoping for companies like Liquid AI, where people are beginning to look at how to make these large language models more efficient. I think that’s going to be important. It’s not just applying the technology; we need to make sure we’re applying it in ways that are good for the world. I think that’s one of the problems I have personally with Bitcoin: the energy consumption, much less the other issues. Large language models can help us, but we need to find ways to make them more efficient, and people like Liquid AI are working on that.

    Ayush Jain [32:07]

    Yeah, there was a Microsoft launch of a mini LLM, which can run on less capacity and can be used for specific tasks.

    Doug Williams [32:19]

    That’s right. And Ethan Mollick is talking about agents: smaller LLMs for specific tasks, I think, are really going to take off, and distributed, so our phones and our laptops can all run the large language models. We just need to think about it both in terms of efficiency at getting the job done and in terms of things like energy efficiency. Shame on us if we invent this new technology that doubles our energy consumption. We haven’t really moved the bar forward. So I’m hoping we see innovations in those kinds of areas.

    Ayush Jain [33:00]

    Yeah, that’s super awesome stuff, Doug. Thank you so much. I really enjoyed this conversation, and some points are well taken. Thank you so much; it was a pleasure hosting you.

    Doug Williams [33:14]

    Thank you, thank you, Ayush, and good luck to you and your team. Bye-bye.

    Ayush Jain [33:21]

    Thank you for joining us on the HealthTech with Purpose podcast. Before you leave, make sure to like and subscribe on Spotify, Apple, and YouTube. And if you would like to see someone on this podcast, do refer them to us; my contact details are in the description. Before I leave, I wish you: stay healthy, stay curious, and keep building.

About Doug Williams

At the time of recording (published June 2024), Doug Williams was Product and GenAI Lead for Orbit, the digital entrepreneurship platform of the Martin Trust Center for MIT Entrepreneurship, where his team builds AI tools for founders on MIT’s disciplined entrepreneurship framework. Before MIT, he served as CTO and in engineering and product leadership roles at several startups, including Zipcar, RunKeeper (acquired by ASICS in 2016) and the Family Education Network, an education company acquired by Pearson. He has also worked in healthcare technology, including work on FHIR-based standardization. LinkedIn: https://www.linkedin.com/in/dcwusa/

Got a healthcare problem worth solving and a hunch about the technology? Tell us about it, and we’ll start with the problem before we talk about the build.

Discover More Episodes

Let’s #Transform Healthcare,# Together.

Partner with us to design, build, and scale digital solutions that drive better outcomes.

Location

Global Tech Teams LLC, 525 Washington Blvd, Industrious at Newport Tower, Jersey City, NJ 07310, United States.

Contact

+1 408 786 5974
contact@mindbowser.com
BOOK A QUICK CONSULTATION

Have a Healthcare Project in Mind?

Let’s discuss your goals, workflows, and next steps in a focused consultation call.

Calendar icon Schedule a Call

Contact form