Theresa Kushner is the Head of the North American Innovation Center for NTT DATA Services and a data.world Advisory Board Member.
The CIO sat quietly through the meeting. His Zoom image showed him in his home office listening intently to his team, looking down at his lap, trying to absorb the presentation on data governance. As the team explained how he should establish a governance committee to manage the data and AI project outcomes that he had commissioned, he shifted uncomfortably.
Here we go, he thought. “I’ve spent tens of millions of dollars over three years getting this data environment in shape to take on new data and AI projects to better understand the operations of our business from how we can do better with pricing to the productivity of our salespeople. But now I’m being told that I need an entirely different team dedicated to overseeing development and use of these assets. It seems to me like scope creep guided by the consultants involved.”
The CIO was frustrated.
Sensing that frustration, one of the consultants stepped forward with a novel solution: “Don’t do data governance the traditional way. Use agile data governance and work it into your organization with people you already have. Create a collaborative environment that helps to govern the collection, management, and use of your data.”
The origin of Agile Data Governance
The suggestion was borne of the consultant’s work with software developers over the course of 20 years. His thoughts were specific to the Manifesto for Agile Software development released in 2001. The manifesto described what developers valued: individuals and their interactions, working software, customer involvement and constant change.
Applying those same values to data, the consultant said, would allow the CIO to move faster, ensure greater usability of data, identify the right investments in data management tools, and build a team that operates well in an ever-changing environment.
After all, the CIO had specific objectives:
He wanted to provide a data environment that could be used by all his data citizens and scientists, a useful dataset free of bias and as clean as possible. He knew if he did that, he would be enabling better decision making, reducing friction between organizations, reducing costs, and increasing effectiveness. He would also be protecting his stakeholders and the needs they had for understanding the data that they needed for decision-making.
As his teams moved more into building AI algorithms, the CIO also wanted to ensure that the application of these algorithms did not adversely affect day-to-day operations. He wanted to ensure that every use was compliant across industry standards, both legally and ethically. He also wanted to ensure consistent understanding across the organization of the data used in the algorithms, the models’ intended uses and the expected effects. In addition, he had the goal of fostering collaboration across the organization to ensure that his teams were identifying appropriate projects for his AI team and that they were managing and monitoring the models’ outputs and usage.
In short, he wanted a governance environment that was built for his team, not one that was provided as a standard template by the hired consultants.
Working with the consultant, the CIO developed a three-step program that used the existing staff and gave them the freedom to invent what would work in their environment. The steps were simple:
Identify a data governance team responsible for ensuring the objectives of the data program. Pick leaders who embrace the values of agile governance.
Instill in the team the values of agile governance and teach them how to collaborate across organizations with different stakeholders. Teach the team how to educate their stakeholders about the value of data and its governance.
Hold the team accountable for meeting the objectives, not with weekly or monthly business reviews, but with daily touchpoints to ensure they are working collaboratively on the objectives as outlined.
The results of their Agile Data Governance initiative
Six months later, the CIO sat at a conference table with the consultants, his governance leads and the head of the company’s audit team. They were reviewing the program, successes and learning opportunities.
The team lead for agile data governance spoke first. “Not everyone has bought into this new way of working. The daily check-ins were a challenge to begin with, but once we got used to applying agile principles to data management, we worked better. We still have a ways to go to ensure that the team has internalized the values of agile data governance, but we’ll get there.”
A lead data analyst piped up by extolling the virtues of the new data environment that included tools for data discovery. “With our new infrastructure, we’re now exploring, analyzing and identifying patterns and trends in our data corporate wide. We’re getting a lot of insights just by doing this and making connections for our data community. Although it’s an input to data governance, it’s become a very important part of our success.”
The company auditor spoke next: “I agree. The data discovery capability, along with the knowledge graphs we’ve built, are becoming invaluable. They are helping us make sense of complex and large datasets and identifying new opportunities to help us improve our decision making. Last month when we thought we had a data breach and we didn’t know where to look, we went to the data analyst community, and it only took 15 minutes to pinpoint the problem and resolve it. I attribute that fast turnaround to the fact that, with this program and the capabilities built around data discovery and knowledge graphs to organize our data, we have created an environment that allows our people to collaborate effectively. These communities are really maturing and everyone is beginning to understand the importance of managing our greatest asset – data.”
The data scientist working on the productivity project for the CIO chimed in with her view of the new governance strategy. “When we showed the finance team our new models for salesperson productivity, there were a lot of questions about how to use the information we had generated. I think we should have some kind of education on these aspects of data. Do we have or are we planning a data literacy program?”
Nods around the conference room suggested that the idea was a good one, but no program had been discussed or designed yet. The action item to get it started went to the data scientist.
Satisfied that the team was on the right track, the CIO was beginning to relax. The overall data quality report card was good. The stakeholders weren’t constantly complaining about the quality of their data. The AI projects that were being implemented were being governed by a team that understood the data and the implications of the AI outputs. It was working.
