Latest / Becker’s Healthcare Podcast / The New Healthcare Access Layer: Connectivity, AI, and Real Operational Gains

Transcript

The New Healthcare Access Layer: Connectivity, AI, and Real Operational Gains · Becker’s Healthcare Podcast

• 0:00
  Hello, everyone. This is Erica Spicer Mason with
• 0:02
  Becker's Healthcare. Thank you so much for tuning
• 0:05
  into the Becker's Healthcare podcast series.
• 0:08
  Today, we're going to talk about how organizations
• 0:10
  are thinking about connectivity,
• 0:12
  where AI is delivering practical value, why health
• 0:15
  care requires a different approach to AI,
• 0:18
  and how patient access continues to drive operational
• 0:21
  efficiency
• 0:22
  initiatives.
• 0:23
  And joining me for this conversation is David
• 0:25
  Edwards, the chief technology officer at Relation.
• 0:29
  David, welcome to the podcast. Thank you so
• 0:31
  much for joining us today.
• 0:32
  Thanks, Erica. It's good to be here.
• 0:35
  It's good to have you here. And before
• 0:37
  we dive into the conversation,
• 0:39
  I wanted to ask you to share just
• 0:41
  a little bit more about yourself and the
• 0:42
  work that you're leading in the health care
• 0:44
  space.
• 0:45
  Yeah. Sure thing. So I am the chief
• 0:48
  technology officer, as you mentioned, at Relation.
• 0:51
  I've been in the company about one year.
• 0:55
  We're essentially a patient access company that specializes
• 0:58
  in intelligent scheduling.
• 1:01
  My career started in 1991
• 1:04
  that certainly dates myself,
• 1:06
  with Cerner,
• 1:08
  and I started as a software engineer.
• 1:11
  I spent a couple decades there.
• 1:13
  Most of my time was in leadership roles.
• 1:17
  Eventually, at one point, I became an engineering
• 1:19
  fellow at Cerner. And then about ten years
• 1:22
  ago, I stepped away
• 1:24
  from health care to really cut my teeth
• 1:27
  on my first CTO job. And I actually
• 1:29
  found that quite
• 1:31
  useful to really step away from health care
• 1:33
  after twenty five years,
• 1:35
  do something different, focus on
• 1:37
  really,
• 1:38
  improving operational skills, running a technology organization,
• 1:42
  learning more about the business,
• 1:45
  working with boards of directors, and so on.
• 1:48
  And really stepping back into health care,
• 1:51
  having all that sort of experience over the
• 1:53
  last ten years has really helped me become
• 1:55
  a better CTO in the context of
• 1:58
  of doing the health care industry.
• 2:01
  Well, great to learn that about you, David.
• 2:03
  And, certainly, some time away can help bring
• 2:06
  new perspective, new skills,
• 2:09
  to to what you're doing currently. And I
• 2:11
  think that's so important in an industry as
• 2:13
  complex as healthcare.
• 2:15
  And we know this industry has long struggled
• 2:18
  with fragmented
• 2:19
  systems and siloed data. So I kind of
• 2:22
  wanted to start our conversation there.
• 2:24
  But from your perspective,
• 2:25
  how is health care's progress on connectivity
• 2:29
  reshaping
• 2:30
  what's possible right now? And where do you
• 2:32
  think the biggest gaps are still there?
• 2:35
  Yeah. So when I when I stepped away
• 2:37
  from health care in 2016,
• 2:40
  the the FHIR specification, I think most of
• 2:43
  your listeners probably know what that means. But
• 2:46
  if not, it's the FHAS Healthcare
• 2:48
  interoperability resources
• 2:50
  specification. It's essentially
• 2:52
  was a standard that was starting to really
• 2:54
  gain a lot of traction.
• 2:56
  And at the time,
• 2:58
  you know, Cerner had implemented it. I think
• 3:00
  Epic at the time was sort of dabbling
• 3:02
  in it.
• 3:03
  And then sort of fast forward, I come,
• 3:05
  you know, back into health care a year
• 3:07
  ago
• 3:08
  and starting to look at where that specification
• 3:11
  has has has emerged.
• 3:13
  You're starting to see, I think, really a
• 3:15
  critical mass adoption, which is fantastic
• 3:19
  when it comes to improving the connectivity
• 3:21
  and interoperability of systems.
• 3:24
  So
• 3:25
  what's what's really exciting to me
• 3:27
  about
• 3:29
  the
• 3:30
  the Firespec
• 3:32
  is
• 3:33
  that when you get critical mass adoption we're
• 3:35
  not there yet, by the way, and I'll
• 3:37
  I'll make I'll I'll I'll comment on that
• 3:38
  in a bit.
• 3:40
  When you get a critical mass there,
• 3:42
  it starts to open up all kinds of
• 3:45
  opportunities on the edge of these systems to
• 3:47
  innovate. And I think that's really,
• 3:49
  for me, what's exciting.
• 3:51
  And so if you layer
• 3:52
  if you think about what's happening in the
• 3:54
  AI space, particularly in the agentic
• 3:57
  realm of AI,
• 3:59
  Once you open these systems through open standards
• 4:01
  and liberate that data,
• 4:03
  there's all kinds of interesting things you can
• 4:05
  do on the edges.
• 4:07
  And we'll talk about a little bit about
• 4:08
  this, but voice AI is is an area
• 4:11
  I think that's pretty exciting,
• 4:13
  and and creates a lot of value there.
• 4:16
  But in terms of some of the gaps,
• 4:18
  I think
• 4:19
  despite having sort of critical mass around
• 4:22
  a standard like FHIR,
• 4:26
  there is
• 4:27
  still an adoption issue.
• 4:30
  And just to illustrate my point,
• 4:32
  Relation integrates with about 10 different EMR and
• 4:35
  practice management systems today.
• 4:37
  Only a few of them actually support the
• 4:40
  FHIR specification.
• 4:41
  So despite the fact that we've got
• 4:44
  a lot of momentum, the reality is broad
• 4:46
  adoption just isn't there.
• 4:49
  There's a variety of reasons for that. There's
• 4:51
  cost, time, complexity, and so on. I think
• 4:51
  that's sort of the anchor that's preventing,
• 4:55
  I think that's sort of the anchor that's
• 4:57
  preventing sort of, more and more, let's say,
• 5:00
  acceleration of of innovation on the edges of
• 5:03
  these systems.
• 5:04
  Mhmm.
• 5:05
  Yeah. So, David, it sounds like since you
• 5:08
  have come back into into the health care
• 5:10
  space, you've seen some progress in connectivity,
• 5:12
  especially with the FHIR specs as you mentioned,
• 5:15
  but still some gaps in adoption,
• 5:17
  room for growth.
• 5:19
  And it sounds like there's definitely some opportunities
• 5:21
  and excitement behind
• 5:23
  AI, whether we're talking agentic AI, voice AI.
• 5:27
  So I'd like to bring our conversation there
• 5:29
  next. There's certainly no shortage of AI hype
• 5:33
  in the industry,
• 5:34
  but leaders are are being asked to make
• 5:36
  very real and heavily scrutinized
• 5:38
  decisions around their AI investments.
• 5:41
  So where are you seeing the technology deliver
• 5:43
  genuine and measurable value right now, and how
• 5:46
  do you distinguish that from the noise?
• 5:49
  Yeah. There's certainly a lot of hype out
• 5:51
  there.
• 5:52
  I think I can certainly appreciate
• 5:55
  the the c suite and the pressure
• 5:57
  that they're getting from their boards,
• 6:00
  their owners to, to adopt
• 6:03
  AI. And look, there's
• 6:04
  despite the hype, there is a lot of
• 6:07
  very interesting problems that are being solved out
• 6:09
  there. So I think cutting through that noise
• 6:12
  is important.
• 6:14
  And some of the areas that
• 6:16
  we are
• 6:17
  seeing very interesting results in is, as I
• 6:20
  mentioned, was the voice AI. We can talk
• 6:22
  about that here in a little bit.
• 6:24
  But the key thing is, I think, for
• 6:26
  all c suite leaders,
• 6:28
  the important thing to remember is go back
• 6:30
  to
• 6:32
  ask yourself this question, what business problem are
• 6:34
  we solving? What are the pain points that
• 6:37
  we feel as an organization?
• 6:39
  And can
• 6:41
  can we reimagine
• 6:43
  solving those problems in a different
• 6:46
  way using AI as sort of a a
• 6:48
  first step in that process?
• 6:51
  And I think that that's the really hard
• 6:53
  transition that I found. I mean, particularly, you
• 6:56
  know, I I work in the software development
• 6:58
  side of the defense here, but
• 7:00
  there's a lot of sort of institutional fear
• 7:03
  and resistance
• 7:04
  of,
• 7:05
  you know, what does this mean and how
• 7:07
  does this work? There's just a lot of
• 7:09
  sort of, like, change management that has to
• 7:12
  happen.
• 7:13
  Don't make poor decisions around AI projects that
• 7:16
  don't actually return value to you. So, again,
• 7:18
  I go back to sort of what business
• 7:21
  problem are we solving,
• 7:22
  try to avoid the sort of fear of
• 7:24
  missing out
• 7:25
  that would lead to poor decisions around a
• 7:28
  project.
• 7:29
  But
• 7:30
  areas that we see in particular
• 7:33
  that are driving, like, clear value is think
• 7:35
  of
• 7:36
  g and a and sales and marketing functions
• 7:38
  inside the company.
• 7:40
  We're seeing a lot of our those organizations
• 7:43
  start to really lean in and start with
• 7:46
  first principles
• 7:48
  and reimagining how they can get their work
• 7:50
  done. So a lot of times, it's an
• 7:52
  efficiency and productivity improvement.
• 7:54
  On the product side, again, VoiceAI, we could
• 7:57
  talk about that a little bit more in
• 7:58
  detail, but,
• 8:00
  I think there's there's really, really interesting
• 8:03
  fantastic work that's going on there.
• 8:06
  On the support side of the organization,
• 8:08
  what we're finding is that the AI is
• 8:10
  helping us resolve,
• 8:13
  issues a lot faster than we could in
• 8:15
  the past. So we're getting massive productivity gains
• 8:18
  there as well.
• 8:19
  Yeah. Really interesting to hear where you're seeing
• 8:22
  some of these measurable gains happening, David,
• 8:25
  and also appreciate you highlighting some of the
• 8:28
  parameters that you'd encourage leaders to think about
• 8:31
  when assessing AI tools. So, you know, staying
• 8:34
  really grounded in the business problem they're solving,
• 8:37
  reimagining solving those problems a different way.
• 8:40
  But you you highlighted voice AI a few
• 8:42
  times, so I wanna go a little bit
• 8:44
  deeper there. This is something that seems to
• 8:46
  be gaining traction in health care. So I'd
• 8:49
  love to know what you think is driving
• 8:51
  greater adoption of voice AI, and what should
• 8:54
  leaders realistically expect from it?
• 8:57
  Yeah. And this is this is a great
• 8:58
  question. I and and this is an example
• 9:00
  of something that is solving a real business
• 9:02
  problem, you know, tying it back to my
• 9:04
  earlier point.
• 9:06
  A lot of these and particularly the organizations
• 9:08
  that we sell into,
• 9:10
  they have call centers with very high turnovers,
• 9:13
  30 to 40%,
• 9:15
  long wait times for patients,
• 9:17
  eight to five operations, and so on.
• 9:20
  And it
• 9:22
  ultimately creates a
• 9:25
  a less than ideal experience
• 9:27
  for
• 9:29
  the patient when they're trying to interact with
• 9:31
  the,
• 9:32
  the the practice to to reschedule or do
• 9:35
  any kind of activity and so on. And
• 9:37
  so
• 9:38
  the what the voice AI is doing essentially
• 9:40
  is is creating a new system of engagement.
• 9:43
  So
• 9:44
  it is think of it as an agent
• 9:46
  that is
• 9:47
  twenty four seven, never quits their job,
• 9:51
  always there to help.
• 9:52
  And there's turns out that there's a large
• 9:55
  number of
• 9:56
  routine calls that happened, things like
• 9:59
  verifying
• 10:00
  appointments, rescheduling appointments, canceling appointments.
• 10:04
  And there's actually a lot of
• 10:07
  there's a kind of a long tail of
• 10:08
  interesting use cases there that we're finding as
• 10:10
  well. All of these can actually be done
• 10:13
  through an intelligent agent. Now what makes it
• 10:16
  really valuable
• 10:18
  is when the the voice AI system or
• 10:20
  the agent itself is integrated
• 10:23
  with the system of record.
• 10:26
  So unless you do that, you're not really
• 10:29
  getting the full benefit of that AI experience.
• 10:31
  And what I mean by that is
• 10:34
  I can call an agent that is connected
• 10:36
  into our scheduling system that knows when I'm
• 10:39
  scheduled for the appointment, and it knows when,
• 10:42
  for example,
• 10:44
  if I need to reschedule it, it knows
• 10:45
  what those open times are going
• 10:47
  to be. And so I can actually interact
• 10:49
  with this software agent
• 10:52
  entirely without a human and solve 50% of
• 10:55
  my calls and be handled by an agent
• 10:57
  now. And that's what we're finding is that,
• 11:00
  many of our clients are realizing that,
• 11:04
  the that that they're getting incredible value, better
• 11:07
  patient
• 11:08
  experiences as a result to that. Mhmm. So
• 11:11
  and look. The hard things go to a
• 11:13
  patient or they go to a human. Right?
• 11:15
  So when there's something that we can't solve,
• 11:17
  we route it to human, and and that's
• 11:19
  actually when you wanna talk to somebody. But
• 11:21
  if it's routine things, then, let let the
• 11:24
  software let the machines do what they're really
• 11:25
  good at doing.
• 11:27
  Yeah.
• 11:28
  It's a great example, David, and I appreciate
• 11:30
  you highlighting this particular use case.
• 11:33
  And I wanna make sure too that we
• 11:34
  touch a little bit more on,
• 11:37
  I guess, adoption
• 11:38
  and kind of the change management piece that
• 11:40
  you highlighted earlier,
• 11:41
  whether we're talking about clinical applications or administrative
• 11:44
  applications, like the example you gave,
• 11:47
  when it comes to AI, health care, of
• 11:49
  course, has higher stakes and concerns around things
• 11:51
  like safety, compliance.
• 11:53
  It's fundamentally different from other industries that might
• 11:56
  be adopting the same tools. So I'd love
• 11:58
  to know from your perspective,
• 12:00
  what makes getting AI right in this environment
• 12:02
  uniquely difficult but also imperative, and what does
• 12:06
  responsible adoption really look like?
• 12:08
  Yeah. That's a that's a great question. And,
• 12:11
  you know, health care has always been much
• 12:13
  more
• 12:14
  conservative
• 12:15
  in its adoption of technology and rightly so.
• 12:18
  I mean, we're talking about patient safety,
• 12:21
  highly regulated environment, so on. And so
• 12:25
  we we always tend
• 12:27
  to put a lot more controls and guardrails
• 12:29
  in place,
• 12:30
  around the systems we build, so much more
• 12:32
  conservative. But
• 12:34
  I I think it's when you look at
• 12:36
  AI, one thing that people need to realize,
• 12:39
  and I think we all know this, but
• 12:40
  we all tend to forget about this, is
• 12:42
  that
• 12:42
  these are probabilistic
• 12:44
  systems. Right? They make mistakes.
• 12:46
  You know, the AI industry calls it hallucination,
• 12:49
  but, ultimately, what it is that means is
• 12:51
  it's making mistakes,
• 12:54
  and,
• 12:55
  which is different than
• 12:57
  many of the software systems that we've built
• 13:00
  in the past, which are deterministic.
• 13:02
  Right? It always works the same way.
• 13:05
  And so
• 13:06
  with AI,
• 13:08
  the the the way that you mitigate
• 13:10
  you you have to think about the risk
• 13:12
  associated with what you're using the AI to
• 13:14
  solve. And the way that you mitigate that
• 13:17
  is by putting controls and guardrails in place.
• 13:19
  And a lot of times, that means human
• 13:21
  in the loop. Right? And in particularly in
• 13:24
  health care, I think always having a human
• 13:26
  in the loop is is a critical piece.
• 13:29
  But, again, you have to weigh the
• 13:31
  you have to weigh the risk associated with
• 13:34
  that use case. So let me give you
• 13:35
  an example of something where there's extremely low
• 13:38
  risk.
• 13:40
  One thing that we do is we actually
• 13:42
  use a secondary AI to
• 13:45
  evaluate
• 13:47
  in the sentiment
• 13:48
  of a call that a patient has with
• 13:51
  a voice AI agent.
• 13:53
  Well, that's a case where it's low risk.
• 13:55
  It just provides a score for us that
• 13:57
  the human can go look at if they
• 13:59
  want to. But that's a case where, you
• 14:01
  know, it it's it's it's something that we
• 14:03
  can tolerate. We don't need to put controls
• 14:05
  and guardrails in place around.
• 14:08
  If you look at something else like ambient
• 14:10
  listening,
• 14:11
  these are systems where a software agent will
• 14:14
  listen to
• 14:16
  the conversation between a clinician and a patient
• 14:19
  and then transcribe that into a clinical note.
• 14:22
  Now it's
• 14:24
  because the AI can make a mistake, it's
• 14:26
  extremely important
• 14:28
  for the clinician
• 14:29
  to be in the loop to verify
• 14:33
  that the AI actually produced something that was
• 14:35
  accurate. So, again, I I think at at
• 14:38
  the end of the day,
• 14:40
  the complexity and the criticality
• 14:43
  of safety
• 14:46
  in the health care space just requires us
• 14:48
  to have
• 14:49
  strong guardrails and controls in place to mitigate
• 14:52
  that risk.
• 14:53
  Mhmm. But, again, like, I tell everybody, like,
• 14:55
  don't don't be afraid of AI. Don't be
• 14:58
  afraid of what AI can do to help
• 15:00
  solve problems.
• 15:01
  Just make sure that you understand the risk
• 15:03
  and that you mitigate that appropriately.
• 15:06
  Yeah. And and I appreciate you sharing, David,
• 15:09
  how relation is approaching those extra
• 15:11
  checks and guardrails as you mentioned.
• 15:14
  Again, another helpful example for our listeners,
• 15:17
  and we've covered a lot of ground today
• 15:19
  already. But just wanted to check before we
• 15:22
  close, is there anything that we haven't covered
• 15:24
  that you think deserves more attention in this
• 15:26
  conversation?
• 15:27
  Yeah. You know, when
• 15:29
  I think it ties into my last point
• 15:31
  about
• 15:32
  leaning into
• 15:33
  AI and being curious. I think it's really
• 15:36
  important for
• 15:37
  for organizations
• 15:38
  to not
• 15:40
  fear the potential here. I would say lean
• 15:43
  into it, experiment,
• 15:45
  do things that are safe, low risk. What
• 15:48
  I often find is that if you just
• 15:50
  start small
• 15:51
  and get a win, and when I say
• 15:52
  a win, I mean go prove out that
• 15:54
  AI can help you solve a problem,
• 15:57
  then it starts to open up the aperture
• 15:59
  of possibilities that an organization would, you know,
• 16:03
  say, okay. Well, we got a whole new
• 16:04
  set of problems we could go solve. It
• 16:06
  starts to change the muscle memory
• 16:09
  of the organization
• 16:10
  and
• 16:11
  reduce the institutional resistance. So that's kind of
• 16:14
  my parting advice is just lean into it,
• 16:17
  be curious, and,
• 16:18
  experiment.
• 16:20
  Great parting advice, David. I wanna thank you
• 16:22
  again for the time that you've made for
• 16:24
  our conversation today. Thanks again for joining Becker's.
• 16:27
  Thank you, Erica. Appreciate the time. And, of
• 16:29
  course, we'd also like to thank our podcast
• 16:31
  sponsor for this episode, Relation.
• 16:34
  Listeners, be sure to tune into more podcasts
• 16:36
  from Becker's Healthcare by visiting our podcast page
• 16:39
  at beckershospitalreview.com.