3 Ways to Improve Data Readiness for AI

Executive Blog
Written by Vishal Patel, Chief Data & Analytics Officer, Webster Bank

Vishal Patel

Chief Data & Analytics Officer

Webster Bank

JUNE 18, 2024

Generative and Traditional AI are a top priority for Chief Data and Analytics Officers across Evanta communities this year. In fact, AI has vaulted to the top of their functional priorities as data leaders for the first time. However, there are challenges to effectively implementing AI initiatives, and 35% of data and analytics leaders report that “data quality and availability” is a roadblock for their organization. 

In this Executive Blog, Vishal Patel, Chief Data & Analytics Officer at Webster Bank and Co-Chair of the New York CDAO Community, shares insights into how data leaders can make incremental gains to improve their data readiness for AI. 

Vishal is an experienced Chief Data and Analytics Officer and data modernization director with more than two decades of success leading all phases of diverse data modernization programs and technology projects. He is also a business strategist, aligning business goals with technology solutions. Vishal is experienced in enterprise data and analytics management and cross product data strategy and development with the goal of implementing "Data as a Service" (DaaS). Vishal is an Advisory Board member for MIT’s Center of Information System Research. He was recognized as DataIQ’s Top 100 “Most Influential People in Data” in the USA.

What are the challenges for CDAOs when it comes to implementing AI initiatives? 

Traditional AI and Generative AI are picking up in activities across companies, especially in the finance industry. This is most likely either for an efficiency play or for driving an improved customer experience. 

As we look at the banking industry, in general, and how it is evolving, there is a renewed sense of expectation from clients and customers in terms of understanding their needs even before the client or customer expresses their needs to a bank. This necessitates improved customer service and satisfaction, using data and information derived out of data at scale as the key factor for the success. This is where AI plays a big role. 

There are quite a few key challenges when it comes to using AI to drive the outcome above, such as data not being consolidated in a centralized place, inconsistent data used by the entire organization, and the need to ensure data privacy and protection. 

One of the very critical factors is the health of the data -- aka “data quality” -- used for an AI operating model. I call these the critical factors as they are the foundation for a successful and directionally right AI strategy to support improved client and customer experiences to drive business growth.

Are there 2-3 things that CDAOs and their teams can do to get their data in a better position for AI? 

Personally, I am a big believer in driving incremental value in using data for a business-aligned strategy, either for a business growth opportunity or for risk mitigation. Having said that, these are three things that CDAOs can consider focusing on, depending upon their business strategy and priorities:

  1. Catalyst to a desired outcome. Focus on business outcomes, and then align key data programs that would be of value for investment in a data strategy. In a simple example, if your company is focused on increasing sales, box your scope of data strategy around only those critical data elements that would support marketing, sales, campaigns, etc.
  2. Consistency of data. Ensure that data is consistent for all the business operations. This will avoid any differences in interpretation in data that is ultimately used for critical, business-driven decision making.
  3. Data Quality, Data Quality, Data Quality. The cleaner the data, the better the output will be from AI driven models. Having a focused approach on data quality for those critical business processes that would be driven using AI will be a key factor for a successful outcome.

Are there any quick wins that CDAOs can achieve during this process? 

There are potential quick wins using AI purely from an efficiency play perspective. Using a no risk or low risk approach can certainly drive incremental benefit for the organization, and at the same time, build the desired trust in AI as part of the mainstream operations in the future. Focusing on a process that would improve efficiency gains could certainly benefit an organization and will support the expanded use of AI for other business outcomes. 

How do you build trust in AI as you are getting started with AI initiatives?

One effective way to drive awareness and trust in the use of AI is based on a proof-of-concept approach. In simple words, using a low risk or no risk use case, which has a defined business outcome, applying AI (Traditional or Generative) in a non-production environment, and displaying the results can show what is possible using AI. 

Depending upon a firm’s risk appetite, one can deploy such AI-driven solutions to a very targeted scope of operations to ensure it performs as expected in a production environment. This is the same approach that one uses typically for product development. I call this phase a “pressure test” for your product. 

If the AI-driven solution meets defined success criteria -- aka passes the test -- then incremental expansion of that solution is a possibility, and that drives trust in the user community. The idea here would be to start small, acknowledge the success and then start the expansion of AI based solutions. 

Also, it is worth noting that continuous monitoring of these AI solutions is also recommended to ensure this approach does not drift from the desired outcome. 

To summarize, I would say to consider these four points:

  1. Identify a low risk or no risk use case for an AI based solution.
  2. Define and have agreement on success criteria for the AI based solution upfront.
  3. Realize the benefit from a narrowed scope of an AI solution implementation and gather feedback from the user community for continuous improvements and/or training.
  4. Deploy a continuous monitoring operating model to ensure the relevancy of the AI based solution.

Are there any lessons learned you can share when it comes to data readiness and AI? 

There are 2 lessons learned that I have heard based on my engagement with a broad scope of data leaders across the industry.

  1. While AI can be a great differentiator in driving desired business outcomes, it is still important to have a human-driven operating model in the loop to ensure AI is doing what it needs to do within its boundary.
  2. Consistent, complete and accurate data is the foundation for getting your desired results using AI. It will depend on a company’s appetite to decide how important these factors are for AI-driven outcomes versus accepting directionally right AI-driven results to support time-to-market needs.

To learn more from your peers in data and analytics, find your local Evanta CDAO Community and join today. If you are already a member, check out MyEvanta to view upcoming opportunities to collaborate in-person and virtually with your CDAO peers.

You can find a previous blog written by Vishal here: 5 Tips for a Successful Data Mesh Implementation.

by CDAOs, for CDAOs

Find your local community and explore the benefits of becoming a member.