Tangible use cases are key to AI adoption
- By Jamie Milne
- Jul 31, 2019
In a very short time, artificial intelligence has emerged as a top IT priority in the federal government, receiving a high level of support from the White House, Congress and federal agencies. The challenge now is to translate that energy into concrete use cases that deliver meaningful and lasting results.
The administration is pushing agencies to begin exploring the possibilities of AI. “Our goal is to get agencies just to start,” Federal CIO Suzette Kent said at the AI World Government conference in June.
But how should agencies get started?
Given the hype around AI, IT leaders may be tempted to look for potential applications of AI in their programs -- to ask, “What can we do with AI?” While this approach might result in some interesting technology demonstrations, it is not likely to deliver tangible value in support of agency missions.
A better approach is to begin with an important mission-valuable opportunity or challenge and to ask, “What can AI do for us?” That means identifying operational pain-points, whether in terms of effectiveness or efficiency, and exploring how AI might address them. This mission-centric discovery process will result in use cases that provide deep insight into how AI creates value and what’s needed to leverage it successfully.
In particular, a strong use case will help agencies overcome one of the most common barriers to AI: Good data governance. Because data is the lifeblood of AI, agencies understand that they need better policies and practices to guide how they gather, manage and use data. That is the gist of the recently released Federal Data Strategy and the draft Year One Action Plan.
Data governance, tackled at the enterprise level, is a daunting challenge. A use-case-focused AI project provides a means of propagating good data governance at a manageable and impactful scale. For each particular use case, an agency can look at what data is required and how it is managed and made accessible to the people or systems that need it.
In fact, this same approach was taken by the team working on the Federal Data Strategy. They sought use cases from the public that solve problems or demonstrate solutions, and they used those projects as a lens through which to look at such issues as data governance; data use, access and augmentation; and decision-making and accountability.
In time, a use case-based approach will help agencies to improve their data maturity, first developing better data policies and practices around individual programs and then streamlining and scaling those policies for broader adoption across the organization.
Organizations go through five stages as they move along what we call the data maturity curve:
- Zero: An organization does not consistently collect or store key data for analysis.
- One: Individual teams or offices begin collecting data, but with no shared definitions or processes.
- Two: The organization begins creating data hubs or data lakes with well-defined management and governance.
- Three: Power users have access to expanded data for exploration, while business users can run queries as needed.
- Four: The organization can rapidly deploy data platforms designed to solve specific problems.
- Five: Data-driven insights are ingrained in processes and accessible across the organization to inform decision-making.
Use cases provide a self-perpetuating approach to driving AI adoption. They help an agency adopt more mature data practices, and those data practices pave the way for developing new, more innovative use cases that deliver even greater value.
In a recent report, the Professional Services Council suggested a similar iterative approach. Given current budget pressures, PSC recommended agencies get started by finding “quick hit” AI projects that offer a clear return on investment, use those projects to gain experience “and then expand further as experience, funding and ROI are realized.”
Without a doubt, the emergence of AI will have a significant impact on how agencies across government deliver on their missions -- in ways that we cannot yet imagine. But we won’t get there all at once. Instead, the task now is simply to work toward that future, one use case at a time.
Jamie Milne is the big data engagement manager at World Wide Technology.