Our guest this episode was Alex Mitchell, who joined Reuben Hall to discuss the role of artificial intelligence in improving mental health care. Alex shares insights on the current challenges within the mental health system, the inefficiencies of traditional intake and triage processes, and how AI-driven tools can help streamline access to care.
“Three years ago, nobody really wanted to adopt these technologies. Now, there are 10 new companies entering the space every week.” Alex Mitchell on the rapid shift from skepticism to competition in AI-driven healthcare.
Find Moving Digital Health on Apple Podcasts and Spotify, and subscribe to the MindSea newsletter to be notified about future episodes.
Read Transcript:
Reuben (00:01)
Welcome to the MindSea Podcast series, Moving Digital Health. Our guest today is Alex Mitchell, surgeon and chief operating officer at Brightlight Health. Thanks for joining us today, Alex.
Alex Mitchell (00:13)
Thanks for having me, Ruben.
Reuben (00:15)
Maybe you can start by telling us a bit about your background.
Alex Mitchell (00:19)
Sure. I’m a general surgeon by training, still practice part time.
During my surgical career, I got into healthcare leadership and into healthcare infrastructure. I spent quite a number of years in executive leadership in the healthcare infrastructure space in Nova Scotia. And in addition on the side, have also been a serial entrepreneur, real estate investor, and have started a couple of other small companies, and currently, one of my ventures is participating as the Chief Operating Officer at a company called Brightlight Health.
Reuben (01:08)
Cool, so practicing surgeon, serial entrepreneur, healthcare leadership, that’s a lot. You must be a busy guy. Maybe tell us what a typical day looks like for you.
Alex Mitchell (01:25)
Yeah, so now I left my former public service role so I don’t work for the province of Nova Scotia any longer. So since June, a typical day for me depends on whether it’s a clinical day or a non-clinical day. A clinical day is off to do things at the hospital. Non-clinical days are get up super early. I’m an early riser. I believe strongly in exercise and quiet time before my day begins. So usually I get up, quiet cup of tea and a solid work out and then start driving. I’ve got four kids. Everybody heads off in different directions and one of my kids is usually my responsibility to get them into town and I will either take him to school or come home and go to our Dartmouth office or I’ll come home and work from home online and juggle between the various different companies.
healthcare consulting company that I’ll spend a few hours with every single day and better part of half the day probably most days with Brightlight related activities and I’ve also got a septic engineering company that usually runs pretty smoothly but a few times a week needs a few hours a day.
Reuben (02:50)
Okay, I think I got all that. So how many days a week are you kind clinical versus non-clinical?
Alex Mitchell (02:59)
Presently about a day a week clinically and in the near future that’ll go up to two and a half days a week.
Reuben (03:06)
Okay, and maybe you could tell us a bit more about Brightlight Health, how that came to be and more what it’s all about.
Alex Mitchell (03:17)
Yes, so Brightlight Health, I’m not the founder. I came on, I think about 18 to 20 months after it was founded. Founded in 2019 by two individuals, one Grant Betts and Om Agarwal.
Grant and I actually are childhood friends and hadn’t seen each other in quite some time. Our kids ended up at the same school and we ended up connecting through the kids, which is where I discovered that he was working on machine learning solutions in mental health. And, you know, and we linked up and I saw what they were working on. And with my background in small business and in healthcare leadership and healthcare design.
There was an excellent alignment of what they needed at the time. They had some early machine learning solutions without much of a business model around it and were working to mature their solutions and start trying to become market ready with those. Brightlight initially was working with some early machine learning solutions, ultimately transitioning over to the large language model space as that space is matured. And Brightlight has been working with large language models to improve the provision of mental health care, really to improve access to care, reduce burdens to care, reduce documentation administrative burden to more efficiently streamline the mental health care workflow right across the entire continuum of care.
Reuben (05:05)
That’s a great story. It’s funny, as parents, how much of our social life, and even professional in some cases, can happen through the connections of our children. There’s one girl on my daughter’s basketball team that jokes that her dad would have no friends if it wasn’t for all the activities and sports she was involved in. It’s interesting how those connections can happen sometimes.
Alex Mitchell (05:38)
Yeah, definitely. I’ve got five kids and I’d say the majority of our social network and or everything else revolves around what they’re up to and who they’re up to it with.
Reuben (05:50)
Cool, so let’s talk about Brightlight Health a little bit more. I you mentioned the intake process for mental health is one of the solutions that you’re focused on. Maybe talk about what’s the status quo and how does Brightlight improve on the typical existing intake process.
Alex Mitchell (06:16)
Yeah, so I mean, I think we’ll go a little bit higher, you know, and just say like the status quo, I think most would understand is terrible. Right, that the overwhelming burden of illness globally is horrendous. Yeah, over a billion people on the planet with documented mental illness. In Canada, one in two by the age of 40 suffered with some form of mental illness and about half a million Canadians out of work every single week in Canada related to mental illness related disability or dysfunction. $51 billion hit on the Canadian economy each year.
And presently, the health system manages the mental health burden through a number of different ways. So on the private side, we have employee assistance programs, among other things, but they run relatively similar to the public side, where people are documenting and noting that they’ve got challenges, usually sent to a phone call center to be, you know, to undergo an intake interview with some form of health professional. Many people just simply directed to online self-help resources, not human-based. In the public health system, very similar. separate out mental health crisis, right? But still it’s a either see family physician or call one of the available intake phone lines usually leave a message and receive a phone call back whatever, it’s staffed by human beings, usually from eight till four p.m. local time, wherever you’re at, sometimes not every day, and no way really to triage who needs the call first, it’s sort of first in first out. And then so a really slow process of filtering everybody only to enter into a system that even if we did have everybody filtered, there’s really no, you know, there’s really no exit plan.
There’s really nowhere for everybody to land in a large volume of people sent to self-help. Mental health crisis, quite a different model of course, like still crisis help lines, but a whole variety of different doors into that system, which include emergency health services, police, other different elements to how patients access care. And we all know that a large amount of that is still driving through primary care, urgent care and emergency departments.
Brightlight and I guess in more in general terms, I think the advent of the large language models affords an incredible opportunity to offload and augment a lot of the human tasks. Artificial intelligence can work 24 seven. It doesn’t have to shut down at four o’clock. It can be making sense of information and having AI driven conversations with people at any time of day and collecting information, summarizing that information into efficient and compact stories and providing potentially an overlying triage layer and other things trying to streamline and make more efficient that intake process and augment the existing human staff and potentially allow for more highly trained healthcare professionals to get off the phone lines and to participate in active care of patients instead of just information collection. And that just as a starting, that’s just the first scratch. There’s so many other ways that the use of artificial intelligence can help streamline and make healthcare more efficient and free up those human resources to do the human elements of care.
Reuben (10:25)
Yeah, and there’s so many different humans or stakeholders involved here, right? There’s the patients and their family and supports. There’s the clinicians and then the administrators and other staff involved in providing care. Maybe let’s start with the patients. Like, how do you feel the patient’s attitude towards interacting with AI tools? Are they very receptive? Is there kind of resistance or mistrust there? What’s been your experience?
Alex Mitchell (11:00)
Yeah, think generally it’s early, right? I think the deployment of many of these technologies is still so early that it’s hard yet to tell. I think that outside of healthcare, the use of AI-driven trap bots, among other things, has had fairly wide adoption. And I think all the reviews are certainly not in, all the reviews are certainly not glowing.
What I would continue to challenge us on is not just the user experience at this moment, I believe it’s important, but I think we do need to compare it against the current contrast of just nobody to talk to. There’s no way to get my story in. so I think that clients, when given the opportunity, are very grateful for any opportunity to share their story in any means that helps move the needle and helps move their particular case and scenario and situation through the system towards a moment at which they might actually receive care.
Reuben (12:07)
And how about the administrators and staff and clinicians? What has there been response been to these types of tools?
Alex Mitchell (12:19)
Initially, healthcare has been terrified of the advent of AI. I heard a quote recently from a group of Canadian healthcare CEOs that when asked, openly admitted that they were standing on the sidelines waiting to see what was going to happen, which is unfortunate and concerning healthcare is a laggard when it comes to adopting new technological change and certainly Canadian healthcare has been slow in other technologies and will continue I believe to be slow in adoption of this. That said on a sort of more individual clinician and smaller divisional and group situation initially people are at first interested, curious, apprehensive
And then once they have an opportunity to see it work, to see the nature of it and to thoroughly understand that this is not full automation, handing the reins entirely over to a computer or software that this really is augmenting their usual work and trying to replace some of the functions that they know aren’t necessarily the most valuable things they do each day, we do get a fair amount of support and understanding and a willingness to use. I think part of the challenge we’ve had is really just deploying and letting people try it and use it. And once they do and see how helpful it can be, then they’re usually very excited to use it more and go deeper.
Reuben (13:59)
So with the Brightlight, I there’s a few different solutions or tools and integrations you’re working with. Where have you seen the biggest adoption from people being willing to try it out and make the transition?
Alex Mitchell (14:21)
Yeah, have two of the two easiest areas are the easiest first is AI scribe to deploy an AI scribe and have people try it and use it is a relatively straightforward low requires very little in the way of integration or other technology challenges and it’s been the easiest thing to get people to try and start with is the AI scribe technology. We also we have an EMR and workflow platform that works very well in allied health care and mental health care situations in those scenarios where we have private third party providers of allied health or mental health especially when they’re new and just opening and trying to decide on adoption of other platforms and that’s been an area where we’ve certainly had some strength. And now currently, you know, we think the largest impact area is on AI driven intake and triage and that is the area that’s most challenging to get approval to deploy and get people to consider.
Reuben (15:42)
So how does the triage part of the intake work? you know, let’s say I’m, you know, an administrator comes to my desk first thing in the morning, there’s been all this, you know, these calls, information that has come in overnight or off hours. You know, what does that look like to me from a workflow perspective?
Alex Mitchell (16:05)
Yeah, so I think the biggest difference is that, you know, from, I guess, you know, so if your call center finished at 4 p.m., so from 4 p.m. onwards through till 8 a.m. this morning, what happened is that, well, anybody that actually depends on whether they’re just coming through or whether we’ve actually deployed. So in some situations, some of our clients would identify potential clients, identify people who they need to collect information from and then deploy the intake chat bot to that individual specifically to then collect all of the information. Or in some situations, it may be the first pass digital front door and triage mechanism. In any event,
Reuben (16:44)
Right, like inbound versus outbound basically,
Alex Mitchell (16:47)
Yeah. In either circumstance, an AI driven chatbot, depending on the situation, would have its conversation thread at the way it speaks, ask questions, the kinds of questions it asks, the way it asks its questions has been co-designed with the client so that it is doing what they want it to do and want it to say. so that, you know, client engages with the chatbot after hours and, you know, it can take a 30 to 60 minute interview and a large amount of information and consolidate it down into a cohesive one to two paragraphs. We can tie that to the DSM-5 criteria or to ICD-11 criteria and pull out a variety of symptoms which can be highlighted and then we can overlay that whatever triage score the organization wishes to apply to it.
So if there’s particular things that they’re looking for that would trigger, hey, this is somebody I’ve got a call first thing this morning, this is somebody I want to send to this place, this is the information I think I want from that patient, whatever it is as part of the information gathering, we can pull that out, highlight it and allow to create a more streamlined bucketing of the various different cohorts of clients as they come through, know one particular One particular organization with a triage level of one through five five being, you know really acute mental health crisis requiring fairly urgent intervention and admission and one being people with very low level mental health issues who may be best directed at self-help and You know web-based self-help tools and it allows them to see the spectrum of what came in and meaningfully direct some of those low acuity stuff to self-help, medium acuity to licensed clinical therapy or clinical psychologists and those at the upper realm, know, fairly rapid contact and progress through the acute mental health system.
Reuben (18:59)
Sounds amazing to me and it would save a lot of time in the process for sure. So what does the onboarding experience look like for a new customer? Like for the intake solution for example, you know, they have their existing system, they want to try out Brightlight. How do they go from what they have now to migrating to a new solution?
Alex Mitchell (19:28)
Yeah, so I mean, I guess I challenged a caution to say it’s not really migration, right? It’s like very few organizations already have a chat bot or something else or even a web based forum at their front end. You know, it’s pretty much human driven. They’ve got usually people on the phone or providing virtual telemedicine care. What it would look like at first is we have a
mature, rapidly deployable intake chatbot. And I think one of our differentiators and value statements is we have an interesting user interface that allows for fairly rapid customization of the conversation threads of the way that the chatbot has its conversations that we don’t, you one size does not fit all. So.
you if you want to do a depression questionnaire, don’t have to take our, like, here’s our depression questionnaire. Our chat bot asks questions about depression this way every time. Well, no, you know, there’s a lot of different situations with a lot of different subtle needs. One group may be working predominantly with substance use issues and looking to triage people pre-detox treatment. Some people may be trying to better differentiate a PTSI population into who’s going to self-help, who’s going to licensed clinical therapy. So when we engage with a client, most importantly is try to understand what’s their business model, what’s their organization doing, how are they doing it, what’s the clientele that they’re targeting and who are they likely seeing.
and then work with them to try to understand, okay, what do you do today? What are the questions you ask? What’s the information that’s most important to you? And then we try to blend that and shape that into the way that our chatbot intake.
interviews so that really our chatbot is functioning as good or better as whatever the human being was doing eight till four for them on the front end. And that varies in many situations, but the first step is understanding who they are, who they’re seeing, what questions they typically ask and what information they really need. And then we make sure that our chatbot can do that.
Reuben (21:50)
And for those like frontline providers, do you find there’s much of a learning curve to working with the technology or is it pretty user friendly?
Alex Mitchell (22:00)
Yeah, the user interface for training it is very, very user friendly. So you do not need to be a software engineer or a coder or otherwise to manipulate this. There’s a few levels here where we do use some various prompt engineering steps that we would do ourselves internally with discussions, understanding that from the client, but it’s very straightforward.
Reuben (22:28)
Okay, and you mentioned a few of the team members that you work with there already. Maybe talk about some of the team at Brightlight and how you all work together to make this happen.
Alex Mitchell (22:44)
Yeah, so we have a fairly diverse background. starts with the two founders, Grant Betts, who one of the co-founders is an accountant by Original Training, also has a background in computer science and has been working in enterprise software support and development for much of his career. He co-founded and runs a enterprise software consulting
company in the US that helps large companies implement, make changes to and support their large, usually ERP platforms, SAPs, oracles, NetSuite, etc. Epicor, that sort of stuff. Agarwal is our other co-founder who was a young child prodigy who was doing machine learning coding and other interesting things.
you know in his early teens and he and Grant linked up during that time and Ohm is a know hardcore genius software engineer.
Reuben (23:53)
Well, it’s nice to have someone like that on your team.
Alex Mitchell (23:56)
Yeah, absolutely. Myself, we’ve already discussed. We also have Dr. Brydon Blacklaws, our Chief Medical Officer, who is a pioneer in the rural virtual emergency care sector and runs the rural virtual emergency care system in British Columbia and has had a lot of experience with telemedicine and provision of virtual care.
We also have Robert Zed, who’s a senior serial entrepreneur and consultant who functions as our board chair and Robert’s got
connections in both the private and public sector in healthcare leadership has started and exited a number of companies and provides us with ongoing governance oversight and also connection to a really large network of very successful businesses, business people and leaders that helps us along our way.
Reuben (25:02)
Excellent. Sounds like a pretty great team and a fun challenge to be working on, an important challenge to be working on. But we understand that as any startup, there are a lot of challenges and problems to solve on daily basis. What are the big challenges that Brightlight has to solve? Looking forward to really unlock the potential of the solution.
Alex Mitchell (25:35)
Yeah, I think that pendulum swings back and forth. Right, think on one day it’s all about trying to fertilize the field. It’s adoption of a very novel technology in a risk adverse sector that is slow to adopt technology. So it’s really finding those opportunities to deploy, to both prove its worth, but also to learn and adapt and fine tune.
And then the flip side of that is with each opportunity to deploy and successful opportunity to deploy for a small technology company is balancing the appropriate level of support, execution and whatnot to those deployments, making sure that they go well, that we meet the customer’s needs and that we take the opportunity to learn from each of those deployments and advance our knowledge and our process.
product. And you know, I think then the other pieces, this is a crazy moving space. So deployment and adoption might be relatively slow in this space, but the advancement and research and development of the products of the frontier large language models that underpin these is moving at light speed. So the technology is changing like really weekly,
like there are developments that occur almost every week that change how we see things working and how we might decide to, you know, to make our platforms work. And it also means that the space and the sector is filling up incredibly rapidly, you know, a space that we had no competition in. But we’re, you know, nobody really wanted to, you know, to adopt any of these technologies three years ago. We’re now in a space where there’s 10 new companies every week entering the space. So it’s a very, very interesting time.
Reuben (27:40)
Yeah, it really is an interesting kind of paradox, like you said, too, the slow movement in the sector and industry side and incredibly fast and changing technology on a regular basis. So I feel it’s like a fire hose sometimes trying to keep up with all the changes in AI and other
technology frameworks on a day-to-day basis. But as part of the fun of the field too is that there is really never a dull moment. It’s like constant challenges and I guess it’s what we live for at the end of it.
Alex Mitchell (28:23)
Yeah. And then, you know, I still love the book, The Lean Startup and the concept of innovation accounting and that it isn’t just about new customers and revenue and other things. It really is, how much did we learn? How fast can we learn and are we learning more faster? And that’s part of the game. if nothing else, part of the accounting that we do every day and every week is what we learn.
because this is a space where there is tons moving and tons to learn and adapt very quickly.
Reuben (29:04)
Yeah, it was a great way of looking at it and a good reminder of that perspective. I really appreciate your time today, Alex, joining me on the podcast. It was great to chat with you and hear about your stories from the frontier there.
Alex Mitchell (29:21)
Excellent, you’re very welcome. Thank you so much for having me. Always appreciate an opportunity to talk about artificial intelligence and healthcare and certainly any opportunity to tell everybody about Brightlight.
