Latest / Let's Talk Risk! Podcast / LTR 161: AI Governance by Design

Transcript

LTR 161: AI Governance by Design · Let's Talk Risk! Podcast

0:00
SPEAKER_00: Hi, everyone, and welcome to Let's Talk Risk!.
0:02
SPEAKER_00: I'm your host, Naveen Agarwal.
0:04
SPEAKER_00: Here we explore how leaders in medtech navigate uncertainty,
0:07
SPEAKER_00: make better risk-based decisions,
0:09
SPEAKER_00: and build safe and effective medical products
0:11
SPEAKER_00: to support sustainable business growth.
0:14
SPEAKER_00: Today, I'm super excited to welcome my guest, Ankita Mishra.
0:17
SPEAKER_00: She is the Digital Health Quality Director at Avinova.
0:21
SPEAKER_00: Now, I've known Ankita for some time now,
0:23
SPEAKER_00: and we also met recently at the IMSC 26 conference in Boston.
0:27
SPEAKER_00: By the way,
0:28
SPEAKER_00: if you guys did not attend that,
0:29
SPEAKER_00: we had a session about a couple of weeks ago where we had highlights
0:33
SPEAKER_00: with four key members of that conference.
0:36
SPEAKER_00: So take a look at that.
0:37
SPEAKER_00: The link is out there on my substack.
0:40
SPEAKER_00: But with that, I'm so pleased to invite Ankita today to talk to us about AI governance.
0:45
SPEAKER_00: I know it's on your mind as well.
0:47
SPEAKER_00: So we'll get into that.
0:48
SPEAKER_00: But as you are settling down,
0:50
SPEAKER_00: please know that this is just a conversation and you are part of it.
0:52
SPEAKER_00: So make sure you are sharing your comments and questions and thoughts in the chat
0:56
SPEAKER_00: and I'll bring them here in the discussion as much as possible.
0:59
SPEAKER_00: With that, Ankita, welcome to the show.
1:00
SPEAKER_00: How are you today?
1:01
SPEAKER_01: I'm good, Naveen, and thank you so much for this opportunity.
1:03
SPEAKER_01: Wonderful.
1:04
SPEAKER_00: So, Ankita, let's start by having you share with us very briefly your career journey.
1:08
SPEAKER_00: What has brought you to this point where you are now acting as
1:12
SPEAKER_00: the Digital Health Quality Director?
1:14
SPEAKER_01: Yeah, sure.
1:14
SPEAKER_01: Definitely.
1:15
SPEAKER_01: Yeah.
1:16
SPEAKER_01: I'm a biomedical engineer by training.
1:18
SPEAKER_01: However, I started my professional journey as a software developer.
1:22
SPEAKER_01: Then I transitioned to device development.
1:24
SPEAKER_01: I've worked on class two devices, threes, implantables, sensors, RPMs, CGMs.
1:29
SPEAKER_01: And around 2019,
1:31
SPEAKER_01: I transitioned from product development to process
1:35
SPEAKER_00: development in former companies.
1:36
SPEAKER_01: So that's been my progressive career pathway so far.
1:40
SPEAKER_01: And what that has provided me is the opportunity to predominantly work in
1:45
SPEAKER_01: the software space more than other areas,
1:47
SPEAKER_01: but also look at the diverse inputs that help shape the decision-making,
1:53
SPEAKER_01: how to be critical about risk,
1:55
SPEAKER_01: for example,
1:56
SPEAKER_01: or how to develop your products,
1:57
SPEAKER_01: how to be agile,
1:59
SPEAKER_01: how to work with ambiguity,
2:00
SPEAKER_01: how to move fast,
2:02
SPEAKER_01: yet make sure In a regulated space,
2:04
SPEAKER_01: you still have to be compliant to the regulations and requirements.
2:08
SPEAKER_01: So all that for the past almost two decades now has helped me transition into my current role.
2:14
SPEAKER_01: Which focuses more on supporting AI solution development in a GXP environment.
2:19
SPEAKER_01: And we are more aligned to GCP regulations.
2:22
SPEAKER_01: And so that's my background, my academic as well as professional.
2:26
SPEAKER_00: Wonderful.
2:26
SPEAKER_00: So you have seen it all.
2:28
SPEAKER_00: And I think you bring up an interesting point that we have to move fast,
2:32
SPEAKER_00: but we also have to do it responsibly.
2:34
SPEAKER_00: And we have to make safe and effective devices.
2:36
SPEAKER_00: It's kind of, in my view, a little bit of a conflict that we face every day.
2:40
SPEAKER_00: And now with AI coming onto the scene and we are talking about...
2:44
SPEAKER_00: So I would agree it is a difficult problem, but it's not impossible.
2:57
SPEAKER_01: The way I look at it, impossible says I am possible.
3:00
SPEAKER_01: So that's my outlook on everything.
3:02
SPEAKER_01: I'm a big practitioner of systems engineering.
3:05
SPEAKER_01: And what systems engineering offers you is to look at things critically.
3:09
SPEAKER_01: And what critical thinking supports is to work with ambiguity,
3:13
SPEAKER_01: to be fast,
3:14
SPEAKER_01: but not ignore the other constraints that we often have to work with in
3:19
SPEAKER_01: a regulated space,
3:20
SPEAKER_01: whether it's high risk, it's low risk or a different risk profile altogether.
3:24
SPEAKER_01: So with that approach in mind, Thank you so much.
3:47
SPEAKER_01: Oh, wow.
4:04
SPEAKER_00: Yeah, I just love that perspective,
4:06
SPEAKER_00: how you have positioned this as a key role for critical thinking.
4:10
SPEAKER_00: And I think that is where all of us have a role to play using our knowledge
4:14
SPEAKER_00: and experience and judgment and applying critical thinking.
4:17
SPEAKER_00: We hear so much about AI is doing this, AI is doing that.
4:20
SPEAKER_00: But what I'm hearing you say is that in the middle of it,
4:23
SPEAKER_00: there is a big need for applying risk-based critical thinking.
4:27
SPEAKER_00: We are going to be even more,
4:28
SPEAKER_00: I would say, important in the whole ecosystem driven by AI, not less important.
4:33
SPEAKER_00: That's what I feel, right?
4:34
SPEAKER_01: No, I absolutely agree.
4:35
SPEAKER_01: Yeah.
4:36
SPEAKER_00: Yeah.
4:36
SPEAKER_00: So let's jump into our conversation today.
4:38
SPEAKER_00: So many people talk about AI governance.
4:40
SPEAKER_00: And the first thing that comes to mind when I hear the word governance,
4:43
SPEAKER_00: oh my God, more procedures, more rules.
4:44
SPEAKER_00: And I know a lot of people might be thinking about that.
4:47
SPEAKER_00: How do we frame this conversation about AI governance,
4:49
SPEAKER_00: not Not as a barrier to innovation,
4:52
SPEAKER_00: but as a catalyst for innovation, which helps us be responsible.
4:56
SPEAKER_00: Can you help us frame that?
4:59
SPEAKER_01: I will echo your sentiment in the sense that AI governance is always perceived as a checklist.
5:05
SPEAKER_01: I have like these hundred things to do.
5:07
SPEAKER_01: This is a paper exercise, but let me present a different view on it.
5:11
SPEAKER_01: Think of AI governance by design as a concept.
5:14
SPEAKER_01: We are aware of quality by design, security by design, privacy by design.
5:18
SPEAKER_01: So let's think of it as AI governance by design.
5:22
SPEAKER_01: When you stop thinking or looking at it from that perspective, it offers a different outlook.
5:27
SPEAKER_01: So I'm focusing more on what I need to do with my existing Mm-hmm.
6:01
SPEAKER_00: Yeah.
6:02
SPEAKER_00: So what I'm hearing you say that is,
6:04
SPEAKER_00: let's not think about AI governance as an add-on requirement
6:09
SPEAKER_00: after things have been done,
6:10
SPEAKER_00: but think of it as early on how we can create a frame
6:14
SPEAKER_00: where AI can be deployed responsibly using a risk-based approach.
6:18
SPEAKER_00: So start from the beginning, not in the end.
6:20
SPEAKER_00: And I think that's a mindset people have.
6:22
SPEAKER_00: We talk about do risk management early on.
6:24
SPEAKER_00: We talk about human factors early on.
6:27
SPEAKER_00: And we intuitively understand that,
6:29
SPEAKER_00: but I just feel like it's very hard for us to implement, right?
6:32
SPEAKER_00: First, I want to confirm with you my understanding of what we were just talking about.
6:35
SPEAKER_00: And second, maybe reflect on why we find it so hard to implement anything by design.
6:40
SPEAKER_01: I would say part of it is, do we trust what we are doing and will it have the intended outcome?
6:46
SPEAKER_01: If you really start thinking in that, I mean, is it worth the investment?
6:49
SPEAKER_01: Am I going to get the ROI?
6:50
SPEAKER_01: Because when we start shifting left,
6:52
SPEAKER_01: it also means We are putting more thought process up front.
6:56
SPEAKER_01: And that's what,
6:57
SPEAKER_01: if you go back to my earlier comment on systems thinking and critical thinking,
7:02
SPEAKER_01: what systems engineering is asking us to do is spend some time on the planning phase
7:06
SPEAKER_01: if you want to avoid rework later downstream.
7:09
SPEAKER_01: So that's where we have to think about it critically in the sense
7:12
SPEAKER_01: that what does my trust translate into in terms of,
7:15
SPEAKER_01: let's say, product requirements or functional requirements?
7:18
SPEAKER_01: It all starts with defining those requirements up front.
7:21
SPEAKER_01: How much risk is mild?
7:23
SPEAKER_01: Is my intended use going to translate for my end user?
7:26
SPEAKER_01: What mitigations am I going to have in place?
7:29
SPEAKER_01: And what's interesting about use of AI in a responsible manner is it's not just
7:34
SPEAKER_01: about the technology.
7:35
SPEAKER_01: Technology is the least of our concerns.
7:36
SPEAKER_01: There are all these peripheral necessities, like are we in compliance with privacy regulations?
7:43
SPEAKER_01: Are we making sure the security breaches do not introduce other unintentional consequences?
7:49
SPEAKER_01: So there's a lot of this cross-collaboration input from other areas
7:54
SPEAKER_01: of product development,
7:55
SPEAKER_01: which is very much aligned to how you would think of when you're considering
7:59
SPEAKER_01: a system and its components.
8:01
SPEAKER_01: So the concepts are not new.
8:02
SPEAKER_01: It's just looking at it with a different lens.
8:04
SPEAKER_01: Yeah, absolutely.
8:05
SPEAKER_01: Now, you made several key points, which I want to make sure I emphasize a little bit.
8:08
SPEAKER_01: What I'm hearing you say is thinking about AI itself,
8:22
SPEAKER_00: In a broader sense,
8:23
SPEAKER_00: not as a governance model to be imposed later on,
8:27
SPEAKER_00: but within the context of the system itself,
8:29
SPEAKER_00: AI is the enabler of what we are trying to achieve, the intended use and indication.
8:33
SPEAKER_00: And the basic question to ask is, do we even need it?
8:35
SPEAKER_00: Just because our top management is saying,
8:37
SPEAKER_00: oh, we must do something with AI, doesn't mean we have to do it.
8:40
SPEAKER_00: We have to be responsible.
8:41
SPEAKER_00: So listen, this is fantastic setup for our conversation today.
8:45
SPEAKER_00: Very good context.
8:46
SPEAKER_00: And I love the point you made that technology is not the main problem here.
8:50
SPEAKER_00: The main problem here is understanding and applying that critical thinking
8:54
SPEAKER_00: and judgment in a systems engineering framework.
8:57
SPEAKER_00: So let's talk about a little bit.
8:58
SPEAKER_00: You mentioned briefly digital health lifecycle before.
9:01
SPEAKER_00: There's a framework.
9:02
SPEAKER_00: And of course, people in our industry understand the SDLC model, the digital health model.
9:06
SPEAKER_00: But help us, you know, dig into that a little bit deeper.
9:09
SPEAKER_00: What does that mean?
9:10
SPEAKER_00: And maybe you can give us an example.
9:11
SPEAKER_01: Sure.
9:12
SPEAKER_01: So the way I like to visualize things in my head,
9:15
SPEAKER_01: Naveen, is let's say we understand the product portfolio.
9:18
SPEAKER_01: Any organization would have that product portfolio.
9:21
SPEAKER_01: For example, let's give some examples, but in the digital health context.
9:45
SPEAKER_00: Let's say we have an app that is just collecting data and displaying data like
9:49
SPEAKER_00: a glucose monitoring app.
9:51
SPEAKER_00: That could be different than an app which monitors data and actually predicts what
9:54
SPEAKER_00: the level would be and allows the pump to inject insulin.
9:57
SPEAKER_00: There's two different scenarios, correct?
9:59
SPEAKER_01: Right.
10:00
SPEAKER_01: Yes.
10:00
SPEAKER_00: So in this framework, let's talk about that.
10:02
SPEAKER_01: So let's say we talk about just collection and display versus making,
10:06
SPEAKER_01: helping make some decision making.
10:08
SPEAKER_01: I mean, it's involved in decision making, right?
10:10
SPEAKER_01: Let's look at these two separate profiles.
10:13
SPEAKER_01: We'll categorize data collection display as low risk and the other case as a high risk.
10:18
SPEAKER_01: So we have different risk profiles here.
10:20
SPEAKER_01: Let's say,
10:21
SPEAKER_01: and just this is hypothetically speaking,
10:23
SPEAKER_01: one of them is non-regulated, the other one is regulated.
10:26
SPEAKER_01: So you see there's a matrix there wherein we have low-risk, non-regulated, high-risk, regulated.
10:34
SPEAKER_01: If now we look at this from a framework perspective and let's say we have procedures
10:39
SPEAKER_01: that are not static but offer the flexibility to define the rigor based on
10:45
SPEAKER_01: which cell of this matrix I select,
10:48
SPEAKER_01: I only focus on that part of my product development requirements, for example.
10:53
SPEAKER_01: Not everything needs to apply everywhere.
10:55
SPEAKER_01: So your whole QMS could become like a framework rather than a static set
11:01
SPEAKER_01: of requirements that every single product has to follow to the code.
11:04
SPEAKER_00: Wow, this is amazing.
11:05
SPEAKER_00: But just to clarify for everybody what you said,
11:08
SPEAKER_00: Risk on one axis, regulated or not on the other axis.
11:12
SPEAKER_00: And in regulation,
11:13
SPEAKER_00: we are talking about whether FDA is regulating our digital health product as
11:16
SPEAKER_00: a device or not, right?
11:17
SPEAKER_00: We can understand that.
11:18
SPEAKER_00: So by risk, we mean lower.
11:21
SPEAKER_00: And let's just not get too stuck on terminology or granular, but keep it high level.
11:27
SPEAKER_00: And what I'm hearing you say is that QMS should be flexible in applying
11:31
SPEAKER_00: the requirements to this portfolio.
11:33
SPEAKER_00: And the same application could have software functions,
11:36
SPEAKER_00: which are Low risk and not regulated and high risk and regulated.
11:40
SPEAKER_00: So we could think about that.
11:41
SPEAKER_00: You bring up a point where,
11:42
SPEAKER_00: you know,
11:43
SPEAKER_00: we have been talking about this concept for a while on Let's Talk Risk is that
11:47
SPEAKER_00: the QMS itself We should start looking at that as a system.
11:51
SPEAKER_00: And with QMSR,
11:52
SPEAKER_00: now FDA expecting us to apply risk-based decision and judgment throughout
11:57
SPEAKER_00: the system.
11:58
SPEAKER_00: What we are talking about here becomes very relevant, is it not?
12:01
SPEAKER_01: Absolutely.
12:01
SPEAKER_01: And look at any other framework, at least in the AI space as well.
12:06
SPEAKER_01: I mean, you look at the NIST framework,
12:07
SPEAKER_01: whether it's security or privacy or AI, everything is risk-based.
12:10
SPEAKER_01: So the mindset has already shifted,
12:12
SPEAKER_01: even from a standards perspective or the state-of-the-art ways
12:16
SPEAKER_01: of implementing these technologies.
12:18
SPEAKER_01: So why shouldn't IQMS, for example?
12:20
SPEAKER_00: This is an important point, Ankita.
12:22
SPEAKER_00: Definitely it will require us to not only change our mindset,
12:25
SPEAKER_00: but I believe a big part of that, Naveen,
12:27
SPEAKER_00: the way I see it is the ability to justify and have a really good rationale
12:29
SPEAKER_00: behind your decision making.
12:30
SPEAKER_00: I do not believe our even device regulations, for example,
12:55
SPEAKER_01: restrict the manufacturers from implementing the requirements in the way it works
13:00
SPEAKER_01: for their device portfolio.
13:01
SPEAKER_01: But in my experience,
13:03
SPEAKER_01: what I have seen is there's often a hesitation to explain that in the right context.
13:09
SPEAKER_01: It's the justification that everybody's looking for.
13:11
SPEAKER_01: Why does something apply?
13:12
SPEAKER_01: And then that kind of naturally progresses to its defined output.
13:16
SPEAKER_01: However, when something does not apply, let's rationalize it.
13:20
SPEAKER_01: It's a critical thinking piece.
13:21
SPEAKER_01: And I can give you a quick example.
13:23
SPEAKER_00: Yeah, sure.
13:23
SPEAKER_01: During my early days off, Yeah.
13:45
SPEAKER_01: So these are the kind of things where let's use established science
13:48
SPEAKER_01: or clinical practices to rationalize where something can easily be rationalized,
13:53
SPEAKER_01: not overthink it,
13:54
SPEAKER_01: and focus more on areas that are of much higher risk and they require more rigor.
13:59
SPEAKER_00: Yeah, that's a great example.
14:01
SPEAKER_00: I really appreciate that, Ankita.
14:02
SPEAKER_00: Okay, so let's get back on track.
14:04
SPEAKER_00: We still want to talk about AI governance.
14:06
SPEAKER_00: Maybe you can help us So from
14:17
SPEAKER_01: my perspective,
14:19
SPEAKER_01: and again,
14:20
SPEAKER_01: I believe more like, okay,
14:21
SPEAKER_01: we have to think of AI governance as part of the design decision making itself.
14:26
SPEAKER_01: It's not a checklist.
14:27
SPEAKER_01: It's not just merely oversight.
14:28
SPEAKER_01: And what we need to make sure is it has a All the right people in the mix.
14:34
SPEAKER_01: Especially when we talk about AI and if we are procuring third-party AI solutions,
14:39
SPEAKER_01: supply chain becomes very critical.
14:41
SPEAKER_01: And not that it's not already critical in the device space,
14:44
SPEAKER_01: for example,
14:44
SPEAKER_01: or even in non-device spaces,
14:46
SPEAKER_01: but given the codependency with the vendors and the terms and conditions,
14:50
SPEAKER_01: for example,
14:51
SPEAKER_01: and the privacy issues and the security concerns and
14:54
SPEAKER_01: the infrastructure dependencies,
14:56
SPEAKER_01: it's very important to make sure we Really understand whether we are going
15:01
SPEAKER_01: to go in-house way or we are going to procure things and how does it reflect
15:05
SPEAKER_01: in our entire ecosystem of how we are developing our products and deploying them and
15:10
SPEAKER_01: then maintainability.
15:11
SPEAKER_01: It's not a one-time thing.
15:13
SPEAKER_01: Once we release or deploy our solutions, we have to maintain them.
15:17
SPEAKER_01: We have to constantly check for drifts and performance.
15:19
SPEAKER_01: So starting from the very beginning to all the way, how do we manage to maintain it?
15:24
SPEAKER_01: That all has to be part of the AI governance setup.
15:28
SPEAKER_01: And it's more than oversight.
15:29
SPEAKER_01: That's why I look at it more from an execution perspective than, okay, what do we have on paper?
15:33
SPEAKER_01: And then we check all the check marks.
15:36
SPEAKER_01: We have to embed those considerations in our existing processes.
15:40
SPEAKER_01: We have to have the right people at the right time making the right decisions.
15:44
SPEAKER_01: Yeah.
15:45
SPEAKER_00: So in some ways, Ankita,
15:46
SPEAKER_00: then it is no different than thinking about The governance model for
15:50
SPEAKER_00: the quality system itself.
15:51
SPEAKER_01: Absolutely not.
15:52
SPEAKER_01: That's why I mean the underlying principles,
15:54
SPEAKER_01: the underlying fundamentals, they are not different.
15:57
SPEAKER_01: It's just focusing on the constraints and the risks the technology is introducing.
16:02
SPEAKER_00: Yeah, yeah.
16:03
SPEAKER_00: So I'm going to touch upon a slightly controversial point here.
16:08
SPEAKER_00: And at Let's Talk Risk!, we talk about these things in a very open way.
16:13
SPEAKER_00: A lot of people are already using And EQMS software, right?
16:17
SPEAKER_00: So many vendors out there.
16:18
SPEAKER_00: I'm nothing against any of these vendors.
16:20
SPEAKER_00: They do good technology, good work.
16:22
SPEAKER_00: Certainly you are confronted with an option,
16:25
SPEAKER_00: somebody coming and telling you,
16:26
SPEAKER_00: hey, we have this AI governance module in the EQMS software for us.
16:29
SPEAKER_00: What are some of the questions I
16:33
SPEAKER_01: believe we really need
16:34
SPEAKER_01: to understand what this solution is doing behind the scenes.
16:39
SPEAKER_01: Like for example, is it going to train on our data?
16:42
SPEAKER_01: How much invasive it will be with regards to our What does it mean in terms
16:47
SPEAKER_01: of how much disclosure is offered to our customers, for example, downstream?
17:02
SPEAKER_01: So it goes back to that trust factor again.
17:05
SPEAKER_01: And because of so much unknown and uncertainty,
17:08
SPEAKER_01: we don't have an acceptance criteria for everything yet.
17:10
SPEAKER_01: Regulations are still chasing.
17:12
SPEAKER_01: So is there a known way to do it?
17:14
SPEAKER_01: No.
17:15
SPEAKER_01: And that's why we have to look at it in the context of the risks involved with our products.
17:20
SPEAKER_01: With low risks, it might be an acceptable thing to move forward with.
17:23
SPEAKER_01: With high risks,
17:24
SPEAKER_01: we may need additional controls in place or additional agreements in place
17:28
SPEAKER_01: with our suppliers, for example, when onboarding.
17:30
SPEAKER_01: And again, going back to the very initial thing, is it worth the investment?
17:34
SPEAKER_01: How is this solution going to help me beyond a certain point than what we've been doing?
17:38
SPEAKER_01: Are there other ways to implement AI,
17:41
SPEAKER_01: still maintain the integrity of the output that we want to and the trust we want
17:44
SPEAKER_01: to offer to our different stakeholders and still meet what we're trying
17:48
SPEAKER_01: to achieve here?
17:49
SPEAKER_00: Wonderful.
17:49
SPEAKER_00: That is an excellent summary and really applying the principle of Flexible decision making.
17:56
SPEAKER_00: And if the software allows us to do that, fine, great.
17:58
SPEAKER_00: Yeah, it'll help us a lot.
18:00
SPEAKER_00: But if it doesn't,
18:01
SPEAKER_00: and if we cannot really have trust in the solution,
18:04
SPEAKER_00: let's hold off on that a little bit, maybe not jump into it.
18:07
SPEAKER_00: So I think these are the questions that practitioners like us should be asking
18:10
SPEAKER_00: before these decisions are made because we are shaping those decisions internally.
18:14
SPEAKER_00: Okay.
18:15
SPEAKER_01: Start small.
18:15
SPEAKER_01: Start small and see if it makes sense and then scale up.
18:18
SPEAKER_00: The sandbox type of an approach.
18:20
SPEAKER_00: I love that.
18:21
SPEAKER_00: I love that greatly.
18:22
SPEAKER_00: Ankita, now let's switch gears a little bit.
18:24
SPEAKER_00: And I want to talk to you more in terms of,
18:26
SPEAKER_00: I would say, personal and professional development for all of us in the field.
18:31
SPEAKER_00: At the conference,
18:32
SPEAKER_00: many people came and talked to me,
18:34
SPEAKER_00: specifically asking me, hey,
18:35
SPEAKER_00: you know,
18:36
SPEAKER_00: AI is coming and we are concerned about whether it'll take away our jobs or not.
18:40
SPEAKER_00: Some people believe it'll be bad for us.
18:42
SPEAKER_00: Some people believe it'll be good for us.
18:44
SPEAKER_00: What is your perspective in terms of what we should be doing as QA, RA,
18:47
SPEAKER_00: risk professionals to make sure that we are prepared for an environment where,
18:52
SPEAKER_00: you know, we have to accept it.
18:53
SPEAKER_00: AI is going to be utilized in some way or another.
18:55
SPEAKER_00: How can we continue to stay ahead of the curve,
18:58
SPEAKER_00: making sure that our So, yes, there will be changes.
19:24
SPEAKER_01: I appreciate that comment from you, Ankita.
19:53
SPEAKER_00: We are going to run out of time very soon,
19:55
SPEAKER_00: but I want to give you an opportunity to ask questions or share your comments
19:59
SPEAKER_00: or thoughts in the chat.
20:00
SPEAKER_00: What do you feel about AI being introduced into all aspects of our work,
20:05
SPEAKER_00: whether it's product development, quality, regulatory, clinical?
20:08
SPEAKER_00: We know we're going to be using this.
20:09
SPEAKER_00: What should we be thinking about in terms of responsible deployment of AI
20:13
SPEAKER_00: and upskilling our own perspective?
20:15
SPEAKER_00: So please share those thoughts and I'll bring them here as much as possible.
20:18
SPEAKER_00: Now, you mentioned lifelong learning.
20:20
SPEAKER_00: Yes, and people will say,
20:21
SPEAKER_00: How do we still inspire people to say,
20:22
SPEAKER_00: hey, it's no longer an issue of going to a classroom.
20:37
SPEAKER_00: Definitely, Naveen.
20:38
SPEAKER_00: One of the things that I have personally started doing is interacting with a lot
20:40
SPEAKER_00: of thought leaders in this space.
20:41
SPEAKER_00: And they may not necessarily all be within your organization.
20:43
SPEAKER_00: There's so many other platforms to do that.
20:44
SPEAKER_00: If you're
21:03
SPEAKER_01: Your company can sponsor you attending conferences.
21:05
SPEAKER_01: That's just like a hub for this thought exchange process.
21:09
SPEAKER_01: And I mean, you mentioned the recent conference we were at.
21:12
SPEAKER_01: So that's one way to look at it.
21:13
SPEAKER_01: There's so much thought leadership that's going on on LinkedIn as well, for example.
21:17
SPEAKER_01: Anybody can access it.
21:19
SPEAKER_01: There are new standards or at least white papers being published on a regular basis.
21:24
SPEAKER_01: I mean, so kind of following the trend and seeing what's happening in this space.
21:27
SPEAKER_01: Start posting.
21:28
SPEAKER_01: That's something I started doing as well.
21:30
SPEAKER_01: Share your thoughts.
21:30
SPEAKER_01: Somebody can challenge it.
21:31
SPEAKER_01: Somebody can add on to it.
21:33
SPEAKER_01: So that's where like this whole very positive exchange of thoughts.
21:37
SPEAKER_01: One thought leads to another,
21:39
SPEAKER_01: brings up a whole new dimension to this whole thinking
21:41
SPEAKER_01: about how can we solve this problem?
21:43
SPEAKER_01: Because it's not just a sector-specific thing that one specific sector is facing.
21:48
SPEAKER_01: Yes, for life sciences, the risks and stakes are much higher.
21:51
SPEAKER_01: So we tend to go much deeper than some other areas.
21:54
SPEAKER_01: But let's lean.
21:55
SPEAKER_01: Let's leverage what's being done.
21:57
SPEAKER_01: Let's see how we can pivot, how we can move.
22:01
SPEAKER_01: And that's
22:06
SPEAKER_00: a new way of learning too, right?
22:08
SPEAKER_00: You know,
22:08
SPEAKER_00: we might get the impression from what we hear about LLMs and things like ChatGPT
22:12
SPEAKER_00: or Cloud.
22:12
SPEAKER_00: Hey, you type something, you get the answer.
22:14
SPEAKER_00: But to me, that is Based on our prior knowledge.
22:18
SPEAKER_00: And when we talk to each other,
22:19
SPEAKER_00: when we discuss live,
22:21
SPEAKER_00: like we're doing right now, you and I are talking and people are participating.
22:24
SPEAKER_00: We create new knowledge, do we not?
22:26
SPEAKER_00: And that is probably not covered by an LLM yet.
22:29
SPEAKER_00: It'll catch up very quickly.
22:30
SPEAKER_00: So what I'm hearing you say is that pay attention to different types of learning as well.
22:34
SPEAKER_01: Absolutely.
22:34
SPEAKER_01: I mean, the tribal knowledge, it comes with humans.
22:37
SPEAKER_01: LLMs, they still face hallucinations right now.
22:40
SPEAKER_01: So that's why you have to use your own Yeah, yeah, fantastic.
22:57
SPEAKER_00: Ankita, we are coming close to our end time now.
23:00
SPEAKER_00: So I want to give you a couple of minutes to think about some key points you want
23:04
SPEAKER_00: to share with us as we close.
23:05
SPEAKER_00: In the meantime, I will share just one very brief announcement with everybody.
23:09
SPEAKER_00: I'm personally excited about this and I have talked about this in the past sessions.
23:13
SPEAKER_00: I have launched an AI-based coach and I call it LTR Risk Coach.
23:19
SPEAKER_00: You know that Let's Talk Risk is about helping us become better
23:22
SPEAKER_00: at applying risk management to real world situations.
23:25
SPEAKER_00: Not about citing a regulation or a clause in 1497, but how to actually apply it.
23:29
SPEAKER_00: And that requires judgment,
23:31
SPEAKER_00: especially when we don't have data, we don't have enough information.
23:34
SPEAKER_00: So how do we learn to become better?
23:37
SPEAKER_00: At Critical Thinking and Judgment, that is where the LTR Risk Coach can help you.
23:41
SPEAKER_00: Think of it as a very smart chatbot,
23:43
SPEAKER_00: which is customized based on my personal knowledge and experience
23:47
SPEAKER_00: of applying risk-based principles.
23:49
SPEAKER_00: It is in a box.
23:50
SPEAKER_00: It doesn't go out to the internet and find an answer or a response for you.
23:53
SPEAKER_00: And it doesn't give you just no answers.
23:55
SPEAKER_00: It helps you to understand your assumptions, refine your judgment, consider options.
23:59
SPEAKER_00: It doesn't replace your expertise.
24:01
SPEAKER_00: I have had it tested by many people and it is really pretty exciting.
24:05
SPEAKER_00: So I would love for you guys to try it out.
24:06
SPEAKER_00: It is available.
24:08
SPEAKER_00: What
24:12
SPEAKER_00: I would
24:18
SPEAKER_01: re-emphasize is let's treat AI governance by design as
24:23
SPEAKER_01: the connected tissue of a product lifecycle ecosystem.
24:26
SPEAKER_01: And to your point,
24:27
SPEAKER_01: let's embrace this new technology or anything new as a means
24:30
SPEAKER_01: to do something differently and use critical thinking to navigate the ambiguity
24:35
SPEAKER_01: and unknowns that are still Wonderful.
24:38
SPEAKER_01: What a way to close.
24:40
SPEAKER_01: Key points, AI governance by design.
24:42
SPEAKER_01: An enormous opportunity, and we have to really up our game to apply it.
24:59
SPEAKER_00: Thank you so much, Ankita, for coming to the show.
25:01
SPEAKER_00: I appreciate the time you took and sharing your insights with us.
25:04
SPEAKER_00: Guys, thank you again for joining.
25:06
SPEAKER_00: You know, every time we do this, you are showing up.
25:08
SPEAKER_00: I appreciate that.
25:09
SPEAKER_00: And those of you who cannot join,
25:11
SPEAKER_00: our recording will be available soon on my Let's Talk Risk newsletter on Substack.
25:16
SPEAKER_00: With that, have a good weekend.
25:18
SPEAKER_00: I wish you all the best.
25:19
SPEAKER_00: And Ankita, thank you once again.
25:20
SPEAKER_01: Thank you again so much for the opportunity.