Latest / Let's Talk Risk! Podcast / Executive Brief: Scaling Software as a Medical Device

Transcript

Executive Brief: Scaling Software as a Medical Device · Let's Talk Risk! Podcast

• 0:00
  SPEAKER_01: Imagine you've just spent, I don't know,
• 0:02
  SPEAKER_01: three years and tens of millions
• 0:03
  SPEAKER_01: of dollars developing this brilliant AI-driven diagnostic tool.
• 0:07
  SPEAKER_00: Right, a massive investment.
• 0:09
  SPEAKER_01: Exactly. And clinically, it works flawlessly.
• 0:12
  SPEAKER_01: But because of how your internal teams are structured,
• 0:14
  SPEAKER_01: it's going to take another 10 to 12 months just to clear the internal paperwork
• 0:18
  SPEAKER_01: and validation processes.
• 0:19
  SPEAKER_00: Which is just an incredibly frustrating bottleneck.
• 0:22
  SPEAKER_01: Oh, it's brutal.
• 0:23
  SPEAKER_01: By the time you actually ship it, your primary competitor has already captured the market.
• 0:28
  SPEAKER_01: So let's get straight to the strategic bottom line here.
• 0:31
  SPEAKER_01: If you are sitting in the C-suite or holding a senior leadership role
• 0:34
  SPEAKER_01: in medtech right now, you know this reality.
• 0:37
  SPEAKER_01: Hardware margins are under intense, unrelenting price pressure.
• 0:41
  SPEAKER_00: Yeah, the landscape has completely shifted on us.
• 0:43
  SPEAKER_01: It really has.
• 0:44
  SPEAKER_01: Software is now your primary engine for differentiation, for recurring revenue,
• 0:49
  SPEAKER_01: and really for dictating your market velocity.
• 0:51
  SPEAKER_01: But that engine only works if you can fundamentally scale how fast you deliver it.
• 0:56
  SPEAKER_00: Right, which requires a massive paradigm shift,
• 0:59
  SPEAKER_00: especially from a technical and regulatory perspective.
• 1:02
  SPEAKER_00: I mean, historically, organizations have viewed regulatory milestones
• 1:06
  SPEAKER_00: as these downstream hurdles, you know.
• 1:08
  SPEAKER_01: Like a roadblock at the end of the journey.
• 1:10
  SPEAKER_00: Exactly.
• 1:11
  SPEAKER_00: Like these massive final exams that you just cram for at the very end of a project.
• 1:16
  SPEAKER_00: To achieve true market velocity,
• 1:17
  SPEAKER_00: we have to start treating compliance and quality
• 1:20
  SPEAKER_00: as routine daily operational opportunities.
• 1:23
  SPEAKER_01: Which means fundamentally changing how the teams work together.
• 1:26
  SPEAKER_00: Right.
• 1:27
  SPEAKER_00: We have to solve them collaboratively within agile workflows rather than, you know,
• 1:32
  SPEAKER_00: just throwing them over the wall to a quality department at the finish line.
• 1:35
  SPEAKER_01: Well, that is the core of our deep dive today.
• 1:37
  SPEAKER_01: We are unpacking a July 2026 McKinsey Life Sciences Practice Report.
• 1:42
  SPEAKER_01: It's titled, The Future of Delivering Software as a Medical Device.
• 1:46
  SPEAKER_00: And it is a fascinating read.
• 1:47
  SPEAKER_01: It really is.
• 1:49
  SPEAKER_01: Our mission today is to give you, the senior leaders in R&D, quality, regulatory,
• 1:53
  SPEAKER_01: and medical affairs, a definitive blueprint for modernizing software
• 1:57
  SPEAKER_01: as a medical device, or SAMD.
• 1:59
  SPEAKER_00: Yep, because we're looking at how to boost your developer capacity and capture real ROI.
• 2:05
  SPEAKER_00: All without compromising patient safety in the process.
• 2:08
  SPEAKER_01: Right. Safety is obviously paramount.
• 2:10
  SPEAKER_00: Always.
• 2:10
  SPEAKER_00: And it's arguably the most critical capability gap in the medtech industry today.
• 2:14
  SPEAKER_00: I mean, the ability to deliver regulated software at speed isn't just a nice
• 2:19
  SPEAKER_00: to have bonus feature anymore.
• 2:21
  SPEAKER_00: It's the dividing line between companies that will dominate the next decade
• 2:25
  SPEAKER_00: and those that are just going to be commoditized.
• 2:28
  SPEAKER_01: Let's ground this before we get into the actual mechanics of how to build faster.
• 2:32
  SPEAKER_01: We need to talk about what we are actually building toward because we are witnessing
• 2:36
  SPEAKER_01: a huge structural shift.
• 2:38
  SPEAKER_00: We really are. The whole business model is changing.
• 2:40
  SPEAKER_01: Exactly. The days of selling a standalone piece of capital equipment,
• 2:44
  SPEAKER_01: shaking hands and just walking away, those days are ending.
• 2:47
  SPEAKER_01: Standalone hardware is maturing.
• 2:49
  SPEAKER_01: The future of customer stickiness really lies in new recurring revenue models.
• 2:54
  SPEAKER_00: You're talking about things like premium analytics layers, right?
• 2:57
  SPEAKER_01: Right. Analytics, subscription-based services,
• 3:00
  SPEAKER_01: and remote monitoring that extends patient care far beyond the four walls
• 3:04
  SPEAKER_01: of a hospital.
• 3:05
  SPEAKER_00: And that shift forces us to completely redefine how we calculate value.
• 3:10
  SPEAKER_00: Because value isn't created by a piece of software just sitting in a vacuum.
• 3:13
  SPEAKER_01: No, it has to connect to something.
• 3:14
  SPEAKER_00: Precisely. It is generated through integration.
• 3:18
  SPEAKER_00: The real goal is placing regulated SAMD into broader digital ecosystems
• 3:22
  SPEAKER_00: that handle complex workflow integration and data aggregation.
• 3:26
  SPEAKER_01: And the McKinsey Report highlights some really brilliant real-world exemplars
• 3:30
  SPEAKER_01: of this in action, right?
• 3:32
  SPEAKER_00: Yeah, they do.
• 3:33
  SPEAKER_00: Take Medtronic's GI Genius, for example.
• 3:36
  SPEAKER_00: It uses real-time AI to improve polyp detection during colonoscopies.
• 3:40
  SPEAKER_01: Which is incredible technology on its own.
• 3:43
  SPEAKER_00: It is, but the key is that it's not just a standalone app living on a tablet somewhere.
• 3:47
  SPEAKER_00: It is integrated directly into the procedure,
• 3:50
  SPEAKER_00: providing core clinical differentiation right at the point of care.
• 3:53
  SPEAKER_01: It fundamentally upgrades the capability of the physician while they are actively working.
• 3:58
  SPEAKER_00: That is precisely the point.
• 4:00
  SPEAKER_00: Another standout example they mentioned is Johnson & Johnson's polyphonic ecosystem.
• 4:04
  SPEAKER_01: Oh yeah, the surgical robotics one.
• 4:05
  SPEAKER_00: Right. They are taking physical robotics and instrumentation and pairing them
• 4:09
  SPEAKER_00: with a digital layer that integrates surgical video. Okay.
• 4:12
  SPEAKER_00: Let's unpack this.
• 4:14
  SPEAKER_00: Because I look at those examples from an executive perspective,
• 4:17
  SPEAKER_00: and the temptation is to
• 4:18
  SPEAKER_00: just say, It is in these scenarios,
• 4:21
  SPEAKER_00: the software acts as the connective tissue.
• 4:23
  SPEAKER_00: It basically makes the underlying hardware exponentially more valuable. Okay.
• 4:27
  SPEAKER_01: Let's unpack this because I look at those examples from an executive perspective
• 4:33
  SPEAKER_01: and the temptation is to just say,
• 4:35
  SPEAKER_01: well fantastic let's hire some developers and write some code for our next device.
• 4:40
  SPEAKER_00: Right. The classic trap.
• 4:41
  SPEAKER_01: Yeah, exactly. Treating dynamic software like it's just a mere accessory
• 4:45
  SPEAKER_01: to a traditional hardware timeline, I mean, it seems like a recipe for disaster.
• 4:50
  SPEAKER_01: It feels like trying to run a high-speed commuter train on a 19th century freight rail schedule.
• 4:55
  SPEAKER_00: That is a great way to put it.
• 4:57
  SPEAKER_01: Right.
• 4:58
  SPEAKER_01: If your rigid five-year hardware development cycle is dictating your software
• 5:02
  SPEAKER_01: release cycle, well, all that dynamic software value is just trapped.
• 5:07
  SPEAKER_00: Yeah, that operational mismatch is exactly where the bottleneck happens.
• 5:10
  SPEAKER_00: You've got two entirely different development metabolisms being forced
• 5:14
  SPEAKER_00: into the exact same pipeline.
• 5:15
  SPEAKER_01: Which just doesn't work.
• 5:17
  SPEAKER_00: No, it doesn't.
• 5:17
  SPEAKER_00: And when we ask why the industry's software delivery speed is stalling out,
• 5:21
  SPEAKER_00: despite the obvious market demand in ROI,
• 5:23
  SPEAKER_00: it usually comes down to that exact friction.
• 5:26
  SPEAKER_01: So let's dig into that friction because it is causing serious business case erosion.
• 5:31
  SPEAKER_01: Traditional phased development approaches have been the backbone of MedTech for,
• 5:36
  SPEAKER_01: I mean, decades.
• 5:37
  SPEAKER_00: Oh, yeah. The classic waterfall model.
• 5:39
  SPEAKER_01: Right. For the hardware leaders listening,
• 5:41
  SPEAKER_01: we are talking about that traditional waterfall model where you lock
• 5:45
  SPEAKER_01: in all your requirements at step one, you build it at step two,
• 5:48
  SPEAKER_01: and you don't even test it until step three.
• 5:50
  SPEAKER_00: Which is so dangerous for software.
• 5:52
  SPEAKER_01: It really is.
• 5:54
  SPEAKER_01: Applying that to software is extending timelines by 20 to 30 percent.
• 5:58
  SPEAKER_01: And when you look at the financials,
• 6:00
  SPEAKER_01: organizations are spending roughly 7 to 10 percent of their total sales just
• 6:04
  SPEAKER_01: on quality related activities.
• 6:07
  SPEAKER_00: That is a massive chunk of revenue.
• 6:10
  SPEAKER_01: Massive.
• 6:11
  SPEAKER_01: When you have late stage validation and just mountains of rework at the end
• 6:15
  SPEAKER_01: of a multi-year cycle, you drain your R&D budgets
• 6:18
  SPEAKER_01: and severely delay your market entry.
• 6:21
  SPEAKER_00: And, you know, the human cost of that bottleneck is just as severe.
• 6:24
  SPEAKER_00: Think about the talent acquisition struggle right now.
• 6:26
  SPEAKER_01: Right. Trying to hire good engineers.
• 6:28
  SPEAKER_00: Exactly. How do you attract top tier tech talent,
• 6:31
  SPEAKER_00: say a brilliant software engineer from a major tech firm,
• 6:35
  SPEAKER_00: when they are used to pushing code to production daily?
• 6:38
  SPEAKER_01: They're used to moving fast.
• 6:39
  SPEAKER_00: Right. If they come to a legacy med tech firm, they might spend, I don't know,
• 6:43
  SPEAKER_00: nine months stuck in a documentation loop.
• 6:46
  SPEAKER_00: Waiting for change control approvals before a single line of their code even sees a user.
• 6:52
  SPEAKER_01: Which has to be incredibly demoralizing for them.
• 6:54
  SPEAKER_00: It is. The root cause is a fundamental clash in methodologies.
• 6:58
  SPEAKER_00: Iterative agile software development simply breaks down when you force it
• 7:02
  SPEAKER_00: through rigid waterfall regulatory gates.
• 7:05
  SPEAKER_01: And beyond just speed, there is a very real security constraint here too.
• 7:10
  SPEAKER_01: Look at the reality of modern cybersecurity.
• 7:13
  SPEAKER_01: It requires continuous monitoring and continuous patching.
• 7:16
  SPEAKER_00: Yeah, the threat landscape is constantly shifting.
• 7:18
  SPEAKER_01: Exactly.
• 7:19
  SPEAKER_01: We are seeing AI-driven vulnerability discovery massively compress the time
• 7:24
  SPEAKER_01: between when a vulnerability is found and when it's actively exploited in the wild.
• 7:29
  SPEAKER_00: It's scary fast now.
• 7:30
  SPEAKER_01: It is. If your operating model relies on a static once-a-year release cycle,
• 7:36
  SPEAKER_01: you structurally cannot secure modern software.
• 7:38
  SPEAKER_00: Which means the traditional funding and planning models are, well,
• 7:42
  SPEAKER_00: they're essentially obsolete for digital products.
• 7:45
  SPEAKER_01: And that brings up a really hard strategic question for the enterprise level.
• 7:48
  SPEAKER_01: Are medtech portfolios fundamentally misaligning their investment clocks?
• 7:53
  SPEAKER_00: That's the million dollar question.
• 7:54
  SPEAKER_01: Right.
• 7:55
  SPEAKER_01: Are we treating dynamic software development like a rigid multi-year capital expenditure?
• 8:00
  SPEAKER_01: Because if you fund a digital ecosystem the exact same way you fund the R&D
• 8:04
  SPEAKER_01: for a titanium implant, you are going to get the market velocity
• 8:08
  SPEAKER_01: of a titanium implant.
• 8:09
  SPEAKER_00: Yeah. You're locking in funding and requirements three years in advance,
• 8:13
  SPEAKER_00: which basically means you are flying blind.
• 8:15
  SPEAKER_01: Completely blind.
• 8:16
  SPEAKER_00: You ignore early market feedback.
• 8:18
  SPEAKER_00: You negate software's unique ability to actually iterate.
• 8:21
  SPEAKER_00: And you almost guarantee that your final product is going to feel dated
• 8:25
  SPEAKER_00: the moment it finally clears regulatory approvals.
• 8:28
  SPEAKER_01: So if the old bottle is completely broken, how do we fix this investment clock misalignment?
• 8:33
  SPEAKER_01: Because the McKinsey report makes it very clear that we can't just tweak
• 8:37
  SPEAKER_01: a few agile sprints and call it a day.
• 8:39
  SPEAKER_00: No, it requires a much deeper change.
• 8:41
  SPEAKER_01: Right. Organizations have to completely redesign their operating models.
• 8:45
  SPEAKER_01: Structurally, that starts with transitioning to continuous business case reviews.
• 8:49
  SPEAKER_00: Which is a huge shift for finance and R&D leaders.
• 8:52
  SPEAKER_01: It is. We have to start treating software as a dynamic investment portfolio.
• 8:57
  SPEAKER_01: Resources should be allocated and reallocated based on active customer feedback
• 9:01
  SPEAKER_01: and actual sprint velocity rather
• 9:03
  SPEAKER_01: than just blindly following some multi-year upfront funding plan.
• 9:07
  SPEAKER_00: And what's fascinating here is the technical execution of that strategic shift.
• 9:12
  SPEAKER_00: It's what the report calls shifting left.
• 9:15
  SPEAKER_01: Shifting left.
• 9:16
  SPEAKER_01: I hear that phrase a lot.
• 9:18
  SPEAKER_00: Yeah, it's becoming a key concept.
• 9:19
  SPEAKER_00: It basically means embedding quality, regulatory,
• 9:23
  SPEAKER_00: and risk management activities into the earliest possible stages of development.
• 9:27
  SPEAKER_01: So moving it away from the end of the process.
• 9:29
  SPEAKER_00: Exactly. Instead of the quality department acting as this sort of policing function
• 9:34
  SPEAKER_00: at the very end of the line,
• 9:35
  SPEAKER_00: quality engineers become embedded active partners in the daily workflow.
• 9:40
  SPEAKER_01: Let's paint a picture of what that actually looks like on the R&D floor.
• 9:43
  SPEAKER_01: How does a day in the life of an engineer change in this shift left model?
• 9:49
  SPEAKER_00: Well, in the traditional model, an engineer writes code for a month,
• 9:52
  SPEAKER_00: throws it over the proverbial wall to the quality assurance team,
• 9:55
  SPEAKER_00: and then just waits weeks for them to flag a compliance issue.
• 9:58
  SPEAKER_01: Which is terribly inefficient.
• 10:00
  SPEAKER_00: Right.
• 10:00
  SPEAKER_00: But in the new model, we move design controls directly into the backlog refinement process.
• 10:06
  SPEAKER_00: When the team does their daily stand-up,
• 10:07
  SPEAKER_00: the quality engineer is actually right there in the room with them.
• 10:10
  SPEAKER_01: Oh, wow. So they're part of the core team.
• 10:12
  SPEAKER_00: Exactly.
• 10:13
  SPEAKER_00: We're conducting continuous risk assessments within the Agile sprints themselves,
• 10:17
  SPEAKER_00: rather than waiting for a separate, massive end-stage hazard analysis.
• 10:23
  SPEAKER_00: Documentation is generated and updated in parallel with the actual development.
• 10:27
  SPEAKER_01: It's like making the nutritional label right into the recipe as you cook.
• 10:30
  SPEAKER_00: I love that analogy. Thanks.
• 10:32
  SPEAKER_01: I mean, you are measuring the ingredients as they go into the bowl rather
• 10:36
  SPEAKER_01: than sending the finished cake to a lab
• 10:37
  SPEAKER_01: and just crossing your fingers hoping it passes the test.
• 10:40
  SPEAKER_00: That is exactly how you should think about it.
• 10:43
  SPEAKER_00: And it extends to the architecture of the software itself, too.
• 10:46
  SPEAKER_00: You start embedding self-validation rules into the code.
• 10:50
  SPEAKER_01: Self-validation rules.
• 10:51
  SPEAKER_00: Yeah. For instance, if you have an adaptive self-learning algorithm,
• 10:55
  SPEAKER_00: you define clear automated parameters for when it needs to be revalidated
• 10:59
  SPEAKER_00: because its operating environment changed.
• 11:01
  SPEAKER_00: The system literally monitors itself.
• 11:04
  SPEAKER_01: That is wild.
• 11:05
  SPEAKER_01: And the evidence benchmarks in the McKinsey case studies for this are just staggering.
• 11:10
  SPEAKER_01: When leadership teams actually commit to this agile quality model,
• 11:13
  SPEAKER_01: organizations are seeing a 50% reduction in testing time. Which is huge. Huge.
• 11:18
  SPEAKER_01: Huge.
• 11:19
  SPEAKER_01: And they are achieving 70% fewer administrative tasks in the release process
• 11:24
  SPEAKER_01: and a 70% cut in change control cycle time.
• 11:28
  SPEAKER_00: That's a massive amount of friction just eliminated.
• 11:30
  SPEAKER_01: Right. And by eliminating all that friction,
• 11:33
  SPEAKER_01: the report notes that you unlock a 10% to 15% increase in pure developer capacity.
• 11:39
  SPEAKER_01: That is the equivalent of getting a significantly larger engineering team
• 11:42
  SPEAKER_01: for free just by changing your operational metabolism.
• 11:46
  SPEAKER_00: It is a massive efficiency gain, but we do need to ground this in reality for a second.
• 11:51
  SPEAKER_01: Okay, let's hear it.
• 11:52
  SPEAKER_00: Embedding quality checks, continuous risk assessments,
• 11:56
  SPEAKER_00: and traceability matrices across every single sprint,
• 11:59
  SPEAKER_00: it is incredibly resource intensive if you try to do it all manually.
• 12:04
  SPEAKER_01: There's just so much data to track.
• 12:06
  SPEAKER_00: Right.
• 12:07
  SPEAKER_00: As the complexity of your SAM'd ecosystem scales up,
• 12:10
  SPEAKER_00: human teams alone simply cannot keep up with the documentation burden
• 12:14
  SPEAKER_00: of a true Agile quality model.
• 12:16
  SPEAKER_01: I hear that. And it leads right into the next major recommendation in the report.
• 12:21
  SPEAKER_01: To scale this new operating model, organizations have to look toward automation.
• 12:26
  SPEAKER_00: Yes, this is where it gets really interesting.
• 12:29
  SPEAKER_01: Here's where it gets really interesting. Exactly.
• 12:31
  SPEAKER_01: Because we are specifically talking about the introduction of agentic AI
• 12:36
  SPEAKER_01: into the software development lifecycle.
• 12:38
  SPEAKER_00: Which is a whole new frontier.
• 12:40
  SPEAKER_01: It is.
• 12:41
  SPEAKER_01: But I have to admit, my buzzword alarm immediately goes off when I hear agentic AI.
• 12:46
  SPEAKER_01: I mean, I've seen tech fads come and go.
• 12:49
  SPEAKER_01: How is an AI agent actually mechanically different from, you know,
• 12:52
  SPEAKER_01: the standard automated testing scripts that our engineering teams have been using
• 12:56
  SPEAKER_01: for a decade?
• 12:57
  SPEAKER_00: It's a totally valid skepticism to have.
• 12:59
  SPEAKER_00: The real difference here is autonomy and context.
• 13:03
  SPEAKER_01: Okay, explain that.
• 13:05
  SPEAKER_00: Well, traditional automated scripts only do exactly what a human programs them
• 13:09
  SPEAKER_00: to do, like running a predefined test over and over again.
• 13:13
  SPEAKER_00: Agentic AI refers to systems that can actually understand a broader goal,
• 13:18
  SPEAKER_00: interpret context, and make decisions to complete complex, multi-step tasks.
• 13:24
  SPEAKER_01: I see.
• 13:24
  SPEAKER_00: Yeah, the report refers to this as an agile quality engine.
• 13:28
  SPEAKER_00: We are talking about a multi-agent AI system where specific AI agents take
• 13:33
  SPEAKER_00: on distinct compliance heavy roles to basically augment your human product teams.
• 13:38
  SPEAKER_01: Break down those roles for us.
• 13:40
  SPEAKER_01: If we are running a sprint to build a new remote monitoring feature,
• 13:44
  SPEAKER_01: what does this multi-agent system actually look like in practice?
• 13:48
  SPEAKER_00: Sure. The report outlines four key agents.
• 13:51
  SPEAKER_00: Let's walk through your hypothetical remote monitoring feature.
• 13:54
  SPEAKER_00: First, you have a requirements translator agent.
• 13:57
  SPEAKER_01: Okay, what does that do?
• 13:58
  SPEAKER_00: A product manager feeds it the raw clinical need.
• 14:01
  SPEAKER_00: Perhaps we need to track heart rate variability in real time.
• 14:04
  SPEAKER_00: The agent automatically converts that into structured user stories
• 14:08
  SPEAKER_00: for the developers, ensuring that regulatory language
• 14:11
  SPEAKER_00: and design control considerations are baked in from the very start.
• 14:14
  SPEAKER_01: So it is standardizing the input before the developer even starts coding.
• 14:18
  SPEAKER_00: Yes, exactly.
• 14:20
  SPEAKER_00: Then you have the quality engineer agent.
• 14:23
  SPEAKER_00: While the human developer is actively writing the code for that heart rate feature,
• 14:27
  SPEAKER_00: this agent is looking at the code and automatically generating the first drafts
• 14:30
  SPEAKER_00: of your technical and design documents in the background.
• 14:33
  SPEAKER_01: Wow.
• 14:34
  SPEAKER_01: It is doing all the heavy administrative lifting while the engineer actually engineers.
• 14:39
  SPEAKER_00: Right, which keeps them in a flow state.
• 14:40
  SPEAKER_01: That's incredible.
• 14:42
  SPEAKER_01: That brings us to the third one.
• 14:44
  SPEAKER_00: Third is the Risk Assessor Agent.
• 14:46
  SPEAKER_00: It acts as an analytical engine,
• 14:48
  SPEAKER_00: constantly reviewing the user stories and code specifications
• 14:51
  SPEAKER_00: to identify potential failure modes.
• 14:53
  SPEAKER_01: Proactively finding risks.
• 14:54
  SPEAKER_00: Exactly.
• 14:56
  SPEAKER_00: In our scenario, it might flag that this new heart rate feature could conflict
• 14:59
  SPEAKER_00: with existing pacemaker data protocols,
• 15:02
  SPEAKER_00: and it immediately feeds that risk data to the human systems engineers to solve.
• 15:06
  SPEAKER_00: And finally, you have the Traceability Specialist Agent.
• 15:10
  SPEAKER_01: Ah, traceability.
• 15:11
  SPEAKER_01: The endless spreadsheets that make half the engineering team threaten to quit every December.
• 15:16
  SPEAKER_00: Exactly. This agent replaces those spreadsheets.
• 15:20
  SPEAKER_00: It maintains the incredibly complex end-to-end links across requirements, risks, and test cases.
• 15:25
  SPEAKER_01: Just constantly running.
• 15:26
  SPEAKER_00: Yeah. It runs constantly in the background, proactively flagging gaps.
• 15:30
  SPEAKER_00: If a developer changes a piece of code,
• 15:32
  SPEAKER_00: the agent instantly highlights which test case now needs to be updated.
• 15:36
  SPEAKER_00: It keeps the entire project in a state of continuous audit readiness.
• 15:40
  SPEAKER_01: Okay, putting my executive hat back on here, I have to push back hard on this.
• 15:45
  SPEAKER_01: Handing over regulatory documentation and risk assessments to an AI sounds like
• 15:50
  SPEAKER_01: a massive vulnerability.
• 15:52
  SPEAKER_00: I get that fear. I really do.
• 15:53
  SPEAKER_01: I mean, we all know large language models can hallucinate.
• 15:57
  SPEAKER_01: They confidently invent facts.
• 15:59
  SPEAKER_01: How do we ensure these agents don't just hallucinate a compliance gap or, worse,
• 16:04
  SPEAKER_01: invent a safety feature that doesn't actually exist in the code?
• 16:08
  SPEAKER_00: That is the single most critical question leadership must ask before piloting these tools.
• 16:13
  SPEAKER_00: And the answer is that this is strictly an augmentation strategy.
• 16:16
  SPEAKER_00: It is never a replacement strategy.
• 16:19
  SPEAKER_01: So humans are still running the show.
• 16:20
  SPEAKER_00: Absolutely.
• 16:21
  SPEAKER_00: The human experts remain completely essential in the loop.
• 16:24
  SPEAKER_00: The AI agents are just doing the repetitive high volume drafting and mapping,
• 16:29
  SPEAKER_00: the low value administrative work that, as you joked, burns engineers out.
• 16:33
  SPEAKER_01: Right.
• 16:33
  SPEAKER_00: But the human quality engineers and regulatory affairs leaders are reviewing,
• 16:37
  SPEAKER_00: validating, and making all the final strategic decisions.
• 16:40
  SPEAKER_00: The AI provides the scale and the velocity.
• 16:43
  SPEAKER_00: The humans ensure the clinical safety and the final regulatory compliance.
• 16:47
  SPEAKER_01: That makes perfect sense.
• 16:48
  SPEAKER_01: It's a co-pilot model, not an autopilot model.
• 16:51
  SPEAKER_01: You still need the pilot in the seat.
• 16:52
  SPEAKER_00: Exactly.
• 16:53
  SPEAKER_01: But before any of you listening decide to overhaul your entire R&D structure based
• 16:58
  SPEAKER_01: on these numbers, we really need to have a transparent,
• 17:01
  SPEAKER_01: objective look at the limitations of this study.
• 17:04
  SPEAKER_00: Yeah, that's crucial.
• 17:05
  SPEAKER_01: The McKinsey Report paints a highly compelling picture,
• 17:08
  SPEAKER_01: but executives need the full context to build a realistic five-year strategy.
• 17:13
  SPEAKER_01: First, the data in the impact metrics,
• 17:15
  SPEAKER_01: like that 70% reduction in change control time,
• 17:18
  SPEAKER_01: they rely on high-level proprietary benchmarks from
• 17:25
  SPEAKER_01: Your organization's mileage is going to vary wildly based
• 17:29
  SPEAKER_01: on your current digital maturity level.
• 17:31
  SPEAKER_00: And if you are starting from a highly fragmented legacy system,
• 17:34
  SPEAKER_00: your initial ROI might look very different during the transition phase.
• 17:38
  SPEAKER_01: Secondly, the agentic AI implementations we just discussed are in the very,
• 17:43
  SPEAKER_01: very early stages across the industry.
• 17:45
  SPEAKER_01: This isn't off-the-shelf software you can just plug into your QMS tomorrow morning.
• 17:49
  SPEAKER_00: No, definitely not.
• 17:50
  SPEAKER_01: It requires careful, custom integration and strict human oversight sandboxes.
• 17:55
  SPEAKER_01: And finally, the regulatory scope of this report is quite general.
• 18:00
  SPEAKER_01: It gives excellent structural advice,
• 18:02
  SPEAKER_01: but it doesn't dive into the granular device class specificities
• 18:05
  SPEAKER_01: or the regional differences between, say, the FDA pathways in the U.S.
• 18:09
  SPEAKER_01: versus the MDR in Europe.
• 18:11
  SPEAKER_00: And if we connect this to the bigger picture, honestly,
• 18:13
  SPEAKER_00: those limitations highlight where the real executive work lies.
• 18:17
  SPEAKER_00: Because the AI tools are still early stage and maturing,
• 18:21
  SPEAKER_00: the technological implementation isn't actually the hardest part
• 18:24
  SPEAKER_00: of this transformation.
• 18:25
  SPEAKER_01: No. What is?
• 18:26
  SPEAKER_00: The heaviest lift here is the cultural shift.
• 18:29
  SPEAKER_00: Change management is going to be the deciding factor between an organization
• 18:33
  SPEAKER_00: that achieves market velocity and one that just creates
• 18:36
  SPEAKER_00: a much more expensive AI-powered bottleneck.
• 18:39
  SPEAKER_01: How do you actually rewire a culture that has historically been rewarded
• 18:43
  SPEAKER_01: for being slow, cautious, and siloed?
• 18:47
  SPEAKER_01: I mean, you are asking engineers who naturally want to run fast to work daily
• 18:52
  SPEAKER_01: with quality leaders who are trained to be inherently risk-averse.
• 18:56
  SPEAKER_00: It's tough. It requires getting those two groups to transition from a combative,
• 19:02
  SPEAKER_00: documentation-heavy mindset to a unified, agile-quality mindset.
• 19:06
  SPEAKER_00: And that requires tremendous active leadership.
• 19:09
  SPEAKER_01: You can't just send an email.
• 19:10
  SPEAKER_00: Right. You cannot just issue a company-wide memo about going agile and step back.
• 19:15
  SPEAKER_01: It absolutely requires active sponsorship from the very top.
• 19:19
  SPEAKER_01: Senior leaders have to visibly champion these cross-functional product teams.
• 19:24
  SPEAKER_01: You have to show up in the key development forums.
• 19:26
  SPEAKER_00: Yeah, visibility is key.
• 19:28
  SPEAKER_01: And most importantly, you have to align their incentives.
• 19:31
  SPEAKER_01: If you tell your team to be agile,
• 19:32
  SPEAKER_01: but their annual bonuses are still tied to rigid legacy waterfall milestones,
• 19:37
  SPEAKER_01: nothing is going to change.
• 19:39
  SPEAKER_01: People do what they are incentivized to do. Spot on.
• 19:42
  SPEAKER_01: You have to clearly and repeatedly communicate how achieving software excellence
• 19:46
  SPEAKER_01: and velocity directly drives patient outcomes and tie that directly
• 19:50
  SPEAKER_01: to their performance metrics.
• 19:51
  SPEAKER_00: It is about equipping them with the right tools, yes,
• 19:54
  SPEAKER_00: but also training them on how to operate in a regulated agile environment.
• 19:58
  SPEAKER_00: It's about building a culture where the quality department isn't viewed
• 20:01
  SPEAKER_00: as the department of no, but as a critical design partner from day one.
• 20:06
  SPEAKER_01: So what does this all mean for you and your organization?
• 20:10
  SPEAKER_01: If we synthesize everything we've covered in this deep dive,
• 20:13
  SPEAKER_01: scaling SAMD and protecting your market position basically boils down
• 20:17
  SPEAKER_01: to three core mandates.
• 20:18
  SPEAKER_00: Let's hear them.
• 20:19
  SPEAKER_01: First, you must shift quality left,
• 20:22
  SPEAKER_01: embedding your design controls and risk assessments directly into your Agile sprints
• 20:26
  SPEAKER_01: so quality is built, not just inspected.
• 20:29
  SPEAKER_01: Second, you need to begin piloting AI, specifically multi-agent systems,
• 20:34
  SPEAKER_01: to shoulder the heavy administrative lifting of traceability and documentation.
• 20:38
  SPEAKER_00: And third?
• 20:40
  SPEAKER_01: You have to fundamentally change your organization's operational metabolism,
• 20:43
  SPEAKER_01: transitioning from static, multi-year funding to dynamic,
• 20:46
  SPEAKER_01: continuous business case reviews.
• 20:48
  SPEAKER_01: That is how you capture recurring ROI in a software-driven market.
• 20:53
  SPEAKER_00: It is a comprehensive overhaul of how MedTech operates, for sure,
• 20:56
  SPEAKER_00: but the cost of inaction is simply too high.
• 20:59
  SPEAKER_00: Organizations that master this continuous,
• 21:01
  SPEAKER_00: high-quality delivery model won't just survive the structural shift in our industry.
• 21:05
  SPEAKER_00: They are going to dictate the pace of innovation for everyone else.
• 21:09
  SPEAKER_01: Which leaves us with one final lingering question for you to ponder as you look
• 21:12
  SPEAKER_01: at your own R&D portfolios this week.
• 21:15
  SPEAKER_01: If agile SAM'd operating models eventually enable truly self-learning devices
• 21:19
  SPEAKER_01: and continuous over-the-air clinical updates,
• 21:23
  SPEAKER_01: how will your organization's definition of a finished product need
• 21:27
  SPEAKER_01: to fundamentally change in the next three years?
• 21:30
  SPEAKER_01: Thank you for joining us on this deep dive.
• 21:32
  SPEAKER_01: We hope it provided the strategic clarity you need to keep pushing the boundaries
• 21:36
  SPEAKER_01: of what's possible in medtech.
• 21:38
  SPEAKER_01: Until next time.