Ask an expert: How to build a robust data foundation

There is huge hype around AI and ML (machine learning) tools right now. But without robust data foundations tied to business outcomes, even the most innovative tools will fail to scale AI sustainably and profitably.

We sat down with Valtech Director of Strategy & Operations Jared Johnson to explore the current state of data foundations in financial services and learn how leaders can innovate while keeping customer data secure.

What does a strong data foundation look like in financial services?

Jared: What I've seen successful leaders do in planning their data foundation is start at the business outcomes that they're trying to drive. They then base the data domains that get pulled into their data foundation on those business outcomes.

I’ve seen data foundations fail when they're built completely horizontally across the business. Trying to boil the ocean of data and build a foundation that works for everyone all at once usually doesn't work.

The most successful leaders usually focus on a vertical slice of their data foundations, linking it to those outcomes that matter most. From there, once you've got a good solid example, then you can start to build out more horizontally. This means ‌that a foundation is going to look different in each financial services firm based on what those outcomes are.

What are the unique considerations for financial services leaders when it comes to building their data strategy?

Jared: Financial services is an interesting space because organizations are making financial recommendations, and to do that they need to have extremely high confidence that they are making the correct recommendations.

So, the need for human validation is greater than in other industries and, what’s more, it is a highly regulated space. Financial services firms need to offer a lot of explainability around how, for example, this person gets a higher credit limit than this one. They need to be able to justify these types of decisions to regulators.

These companies are also dealing with privacy concerns. They have to deal with data in a very secure and robust way to make sure they are respecting the privacy of their users. Because there is so much risk around offering data and AI-driven recommendations, we often see financial services firms building out AI or ML models for internal use or for human-in-the-loop use.

These internal use cases become the primary focus rather than having insights or recommendations generated by AI and ML directly exposed to customers.

Do you think we will start to see financial services firms extending these AI and ML tools for customer use?

Jared: It’s a journey. I think as we get more financial services firms building out their foundations and they’re able to generate more confident, validated, explainable insights, there will be opportunities to expose more and more of that to their customers.

I think the priority right now is to use AI and ML to scale employees, to make them more productive, and to increase their bandwidth, giving them knowledge to service customers better and faster. That’s the more common business case that we see.

For example, using data to power dashboards that a wealth manager or customer services rep can use to see recommendations, filter them and advise the customer better.

How important is the quality and structure of your data foundation when it comes to implementing AI and ML tools?

Jared: It’s critical. The more we’re talking about making recommendations — especially numbers-based recommendations — the more important it is that you have a data foundation composed of trustworthy, harmonized data. There are some sectors and areas where, if your data isn’t quite right, there is more flexibility and less risk.

In the space of advising someone on what to do with their life savings, you need a very justifiable, explainable recommendation. Your data foundation has to be incredibly robust.

And again, the more you’re building something that goes directly to the customer, the more risk that you have. The more that you're building something ‌that goes to an employee for review, the less risk that you have. And we'll slowly move along that spectrum over time with more direct exposure to the customer.

How can leaders maintain a strong level of data security, privacy and regulatory insight while innovating?

Jared: There are a few ways. One is creating a sandbox for innovation that’s not necessarily connected to the internet or for outside use. Just for internal use and understanding of the model. Another approach is starting with internal use cases —keeping data from being exposed to customers reduces risk.

Banks and financial services have been dealing with security and encryption for decades, so they know how to do that. The novel thing here is building models on top of that data and thinking about how to incorporate third-party data to enrich their own.

So, maybe by understanding things their customers have bought or looked at on the web, they can learn to service their customers better. But by moving through that process of sandboxing proof of concept before internal and then finally external use, they can reduce the risk of things working incorrectly or of data being surfaced where it shouldn’t be.

What’s the one piece of advice you would give to financial services leaders on creating a successful data strategy?

Jared: Too often we find there is no comprehensive data strategy for the business. Financial services leaders should be thinking, what is our firm-wide data strategy? What is our corporate data strategy? How are we thinking about data beyond a product or business unit?

This holistic view can help them to really unlock the potential of their data.

Learn more

Building strong data foundations is a key to leading impactful innovation and navigating constant change. Read more in our latest guide, Building Future-Ready Financial Services.