How Information Governance Generates Value with AI

The majority (86%) of financial services executives who have adopted AI believe that it's going to be very or critically important to their business, says Deloitte¹. While AI is powerful, it only responds to the training data it has available. If that data is bad, AI’s results can have a detrimental impact on both financial institutions and consumers.

In this environment, Information Governance (IG) has perhaps never been so important. IG provides the policies, requirements, and capabilities that allow AI to generate value and mitigate risk.

In a recent Leaders Forum hosted by American Banker, David Smythe, Principal Advisor, Strategic Deals and Alliances, Iron Mountain, and Galina Datskovsky, Member of the Board of Directors, OpenAxes, discussed how AI will transform banking, promising financial institution use cases, areas of concern, and the role of information governance in AI implementation.

The term AI and generative AI are sometimes used interchangeably. Are there are fundamental differences between them?

Galina Datskovsky, Member of the Board of Directors, OpenAxes: AI algorithms based on traditional machine learning have been used for decades to ingest data, recognize patterns, and predict outcomes. Like traditional AI, generative AI also learns from a large collection of data sets, but what’s different is that generative AI creates something new.

David Smythe, Principal Advisor, Strategic Deals & Alliances, Iron Mountain: Yes, it’s the creating that distinguishes generative AI from AI. For example, AI tools embedded in a contract management system can alert you that a contract was not signed. Generative AI will write the contracts.

What’s exciting is that generative AI adoption has been more rapid than other transformative technologies such as the internet and cloud computing.

How can financial institutions use generative AI?

Datskovsky: There are lots of use cases. Generative AI could analyze banking products and services and create new and compelling offerings. It could totally replace human assistance. It’s already integrated into office productivity tools that read emails and schedule meetings or draft responses.

Financial institutions will be able to spend less time and resources on support and more on product-focused initiatives.

Are there dangers of relying on AI? What should bankers consider?

Datskovsky: There are dangers. For example, in the area of cybersecurity, generative AI can be used to detect computer viruses and write antiviral code much faster than humans. But generative AI could also create destructive viruses.

Smythe: Bankers need to perform due diligence to mitigate risks. Does the person using these generative AI tools understand the tools and their level of security? Are the tools public or private? What data is underlying the tools? Hallucinations—generative AI models assuming something exists that actually doesn’t--are a concern. Information governance to ensure data quality is incredibly important.

Can bankers use AI with both internal and external data?

Datskovsky: Once your data house is in order and you have reliable, clean, assured data, you can combine that data with other demographic analysis to develop AI-powered financial analysis and reporting systems personalized to the specific needs of individuals or organizations. You could create new tailored investment portfolios of products based on risk appetite and past investment behavior.

This is feasible today with generative AI, but I would feel more comfortable if generative AI acted as an assistant and made the recommendation and a private wealth advisor vetted and reviewed the recommendation.

However, at some point, it's not unthinkable that most of that work will be done by generative AI. The result will be democratized access to information.

What are the benefits of having a strong information governance practice for AI? What best practices can you share?

Smythe: Information governance provides an awareness of the AI tool you are using and an understanding of the data that's feeding it. Think about investments. If you are looking at returns and options for a portfolio, is the data that you're looking at market data or does the data include geopolitical data that might impact the results? What about weather data? There's all kinds of data that you can pull into these tools but you need to understand what that information is, where it comes, and the quality of that data.

The role of information governance and data governance personnel is to make sure that policies, procedures, controls, and checks and balances are in place to ensure that the tool is being used properly and that the outcomes are validated.

Datskovsky: Financial institutions have an amazing treasure trove of data, but it’s in disparate systems. Especially in the larger financial institutions, there's no semantic alignment in business terms. Financial institutions may think that they manage and govern their data well, but most are not managing it as well as they should.

As long as this wealth of information can be normalized, assured, and governed properly, it could be a tremendous training ground for generative AI tools. These tools could really dig into that data and provide insights that you might not even be thinking about.

Once you put all those safeguards in place, the sky’s the limit for using generative AI.

What's the importance of human expertise and judgment in conjunction with generative AI?

Datskovsky: Generative AIs is not a reference system. It’s not a Google search. Instead of finding information and presenting it, generative AI learns. But it can also embellish and hallucinate. There still needs to be a human element to vet the outcomes.

For bankers, it’s their fiduciary responsibility to fact check what generative AI presents.

It’s extremely important to remember that the regulatory landscape applies to institutions and people, not to algorithms. Don’t just throw results of generative AI over the fence. Yes, generative AI could generate your contract, but you still need read it.

Smythe: Bottom line is that you can't transfer the responsibility to the AI tool. AI is not going to pay penalties, fines, and settlements if you get things wrong.

If you are interested in learning more about how Information Governance generates value with AI, view the full webinar here.

¹ https://www2.deloitte.com/content/dam/Deloitte/us/Documents/process-and-operations/us-ai-transforming-future-of-banking.pdf