AI Governance Requires a Full Team Effort – Techstrong AI Podcast EP16

Hello and welcome to this week's edition of the Techstrong AI Podcast. I'm Amanda Ani, and I have a guest with me today. I am so excited to introduce Heather Gentilly.

She is the executive director of Product Management for Watson X Governance. How are you doing today? Great, Amanda.

It's fantastic to be here today. Wonderful. Well, um, to start out, can you share a little bit about your experience and background and, um, exactly what do you do in IBM?

I'm happy to, um, so as global head of product for what's Next governance as part of our What's Next platform I'm responsible on for the product strategy delivery, um, and collaboration with clients and partners on how to accelerate responsible and accountable adoption of ai. Um, and there's been, you know, so much focus in this area given an opportunity with generative ai. Um, in my role I work with IBM research, um, which is really an innovation group here, uh, to look at different data science methodologies that we can productize into guardrails to help mitigate different types of risks associated with AI adoption.

And I also serve as a software focal on IBM's AI ethics board. Um, and that helps us to drive, you know, our own responsible AI adoption standards into all of the product work that we do. And even how IBM adopts AI use cases in support of our own business use.

Wonderful. Well, that's a great segue into our discussion, which is about AI governance and how it starts at the top with the c-suite, you say. So let's talk about that for a minute.

Um, can you explain what you mean by that and, and how does that trickle down to the rest of the organization? Absolutely. So what we've been seeing is AI governance becoming more strategic to organizations of all sizes.

Um, you know, organizations had adopted AI predictive ML for years, really in support of individual business units. Those projects were often run in silos, um, collaboration between the business and data science and it, and we saw all of that start to transform a little upwards of a year ago with the opportunity presented by generation generative ai, um, because there is so much innovation that the organization could gain, but there's also a lot of risk, and this is a newer technology. So what we've been seeing is even organizations who have been successfully adopting predictive models for years are taking a more strategic view of how they want AI to support their business strategy.

At IBM, we have a huge focus on AI for business. Um, but with that, organizations wanna make sure that their goals and aspirations for AI adoption align with their own culture, their ethics, their standards. And that's where, um, you know, IBM's approach was that we had implemented an AI ethics board, um, over eight years ago, and over time have included, you know, more and more layers of employee interaction as, as part of that ethics group, and really matured our processes, what our clients were seeing.

The first step being to bring together that diverse group of stakeholders that can inform the overall AI adoption strategy. So we're not just talking about a collaboration between the business and, um, you know, data science and IT anymore, you've got perspective coming in from hr, certainly from risk and compliance, the chief data office. The chief privacy office remains very involved, but you have stakeholders like marketing, thinking about reputational risk, um, and just at that top level creating that informed view, but then supported by additional layers of management going down to the employee level.

So that, similar to what we saw with data privacy, um, responsible AI adoption really does become part of everyone's job. So breaking down all the silos and having a full company collaboration. Yeah.

And, and that really helps to not only accelerate adoption of AI if you've got a good standard, but align on a best practice for adoption. Because ultimately you don't wanna have to stop throughout the process with these different stakeholders and say, wait, I need to think about governance or, wait, I need to think about compliance. It's actually a great opportunity for technology to automate your best practices and a workflow that gives you confidence.

Um, you've got a full audit trail of your AI adoption and it's explainable. So in the company, I'm sure there are those employees or staff members who are still uncomfortable with AI or maybe lack some skills there as it is evolving so rapidly. So what are some tips that you have for business leaders when it comes to those issues?

Mm-Hmm. So it's a tremendous opportunity for learning for employees in, in all positions really. Um, and that's where, you know, leadership can help with employee training.

Um, but we saw a lot of focus on employee policies and procedures, um, really being strengthened and written out in more detail as a result of chat GPT going mainstream a year ago. And suddenly, you know, organizations who are very focused on data governance, they had processes in place for predictive ML adoption now had to think about easy access for employees to use bottles that were very easy to interact with. Um, so I think that, um, uh, being able to empower your employee through training so they understand not only the opportunities associated with ai, but the business risks as well is the first step.

And then also considering what controls you need to have in place as an organization to make sure that these employees are understanding and following the training and the policies. So when it comes to AI governance and AI implementation, what are some of the biggest roadblocks you've seen business leaders experience, and where would you say the problem lies and what advice do you have? Mm-Hmm.

Um, yeah, so challenges of many. Um, what we're seeing is mo most organizations are trying to evolve their AI governance policies and procedures in parallel with understanding what are the opportunities for technology to make sure that these, that, that the AI governance program is successful. And that's where I talked a little bit about being able to push out the best practice in workflows so that regardless of where the employee sits in the organization, everyone's following the same practice for AI adoption.

Um, you don't go to the next step until all the required pieces of information are collected, and that way you're able to build an audit trail starting at the very time that the use cases requested by the business recording the rationale for why it was approved. And if it is approved, do that risk assessment. Um, so that's where the automatic governance data capture starts and then goes to engineering, collecting information about models being prepared compared in which one's the right ones to support the use case, um, capturing performance benchmarks as a fact sheet so that you can do that real time monitoring when the model goes to production and be alerted if there's anything statistically significant that deviated from when the model was accepted.

And it's really these things that help to ensure explainability and transparency throughout that model's lifecycle. So we've seen over the past year or so, just how quickly artificial intelligence, especially the generative AI can evolve and just all the capabilities that it has and the use cases for it. So what do you see happening as far as the future and how will that affect business leaders as they try to stay ahead of this technology?

That's a great question. Um, so in the past, you know, 14 months or so, we saw a lot of organizations beginning to experiment with generative ai. Um, and there's definitely a, a human in the loop, um, in the use cases that we see people focusing on related to internal productivity of employees, um, customer service and leveraging LLMs to be able to have cu the best customer information possible available to help desks on demand, um, and things like app modern modernization and being able to improve developer productivity by converting code.

Um, so those have been some of the top areas of focus. And as they're coming out of an experimentation and, and going into production, we're now seeing more focus on, um, governance at runtime, at an increasing focus around cyber risks and how you present, you protect, you know, LLMs and other Gen AI models once they, you know, are live and in production and constantly gathering information. Um, so I, I think we will see a strawberry security element in the next 12 months.

Um, I think also the different regulatory focus that we've seen led, um, in the EU with the EU AI Act and, um, providing regulatory guidance that's more prescriptive that organizations can then break down into risk and controls, um, to ensure compliance will also be a big area of focus. And even in, within the United States, we're seeing many organizations beginning to take that regulatory focus now, um, because there has been a lot of activity in Washington DC and there's also a lot of activity at the state level. So I know, uh, everyone is also concerned.

You mentioned the regulations and, and part of that too, um, is a concern about, uh, the risk that comes with ai. So what are some of the key areas of concern business leaders are focused on when it comes to AI risk? And what do you suggest as far as protecting against those risks?

Mm-Hmm. Well, so I think one of the biggest risks is what model do you choose, um, confidence in models that are being provided by third parties or open source models. Um, we've taken approach with our technology to govern AI anywhere because if you think about the state that people are coming from, they've been working with preferred vendors for years.

Um, oftentimes the organization is, is working with mul multiple vendors so it wouldn't make sense to rip and replace those trusted relationships or the technology investment. So I think having a AI governance framework that's flexible enough to support the goals and needs of the organization, and that means being able to govern any type of AI model or AI application regardless of where it resides. So any cloud on-prem, um, and technology can be very open and do that very well, and that's what makes all of this so exciting.

Absolutely. So if there was one key takeaway that you could leave our audience with today, what would that be? Um, so my advice would be to, you know, make the time to learn about the opportunity that these advancements in AI are creating, um, for your business and for yourself personally from a career standpoint.

Uh, there's so many, uh, different resources out there that organizations can get free information around this. We do a lot of work in this area, uh, through IBM's Institute of Business Value. And so right on our website there's a library of white papers about the most common AI use cases that we're seeing across different industries.

Um, or there's a good new publication about the CEO's guide to, excuse me, adopting generative ai, which I think is a great opportunity to understand some of the strategy behind AI adoption and where some of the leaders in industry are really focusing those investments. Wonderful. Well, I know you have your hands full, probably.

There's the IBM think event coming up this Monday and it it's gonna last for a few days, correct? That's right. Yeah.

So we are kicking off Think in Boston, which is right in my home. Um, we begin with Partner day on Monday, and then we've got three full days following up. So looking forward to seeing many at this year's Think conference.

Wonderful. Well, I'm sure it'll go well. I enjoyed going last year and thanks for coming on our podcast and sharing your insights with us.

Thank you for having me, Amanda. It's been great talking with you this morning. And to our audience, please stay tuned Next week we'll have more information in the world of ai.

Have a great week.