Federal agencies can be particularly susceptible to AI failure, for several reasons.
A recent report by Bloomberg Government estimates that IT investment will reach $93 billion in fiscal year 2020, with a large chunk of that money allocated to AI projects—many of which may fall short of expected outcomes.
Federal agencies can be particularly susceptible to AI failure, for several reasons. First, agencies are short on an AI-centric workforce and are just ramping up data scientist hires (the Pentagon didn't hire its first chief data scientist until August 2018). Next, agencies are stewards of information that often requires classification, cleansing, and labeling before it can be used to create AI insights.
Despite these challenges, agencies are starting to use data and AI to enhance citizen services, improve health outcomes, and drive cyber intelligence. These successes are born from having important prerequisites in place and fall into three distinct categories: operational, technological, and organizational readiness. Together, they enhance an agency’s ability to turn data into a strategic asset while harnessing the power of data to innovate faster.
Operational readiness is about developing the management and governance mechanisms to develop and sustain an AI solution. There are three common organizational readiness scenarios.
First, agencies with large datasets can benefit by focusing on enhancing mission effectiveness. These agencies can staff up a data-centric workforce, develop a data strategy, and define data storage access and governance capabilities.
Then, there are agencies that run workloads in traditional environments and are interested in applying AI or machine learning to explore automation or optimization. These agencies want to expand into algorithmic development for pattern matching, enhanced associations, and anomaly detection.
Finally, some agencies are actively investigating the potential of AI in driving innovation across their missions. These agencies are asking “what if?” questions, as the problems AI can solve may not be immediately obvious. They are developing AI ecosystems that accelerate their abilities to harness their agencies’ data.
In each of these scenarios, the agency’s chief data officer is strategically important. CDOs have a vital job: to understand how external and internal data can be used to support agency missions, and to create policies that drive selection of AI tools and methods. CDOs can provide you with a clear vision of what can be accomplished through AI and help answer important questions, such as which priorities will yield the biggest impact.
Filling this role and rounding out a skilled AI team may require bringing in outside expertise, particularly solution architects versed in data science. Proceeding beyond proof of concept without mission-focused context adds unnecessary risk, particularly in the review and evaluation stages.
Technological readiness ensures the agency has the proper architecture, infrastructure, data integration and interoperability capabilities. Leverage your agency’s best practices, but also be willing to test and modify your infrastructure before rolling out a new AI solution. You will want to understand whether existing data center facilities can manage your AI workload. Often the answer is “yes” for a simple proof of concept but may be “no” for a longer or production solution.
For example, if you are using an on-premises data center, you may consider cloud-based services. This can be particularly useful for processes that require a large amount of storage or the ability to scale, such as image and natural language processing. As your agency looks to scale up the use of AI, you will need to be sure your cloud-based resources can handle the job.
High-quality data sources must be available, trusted and accessible to everyone involved in the project. By focusing on hybrid cloud infrastructure, you can strike a proper balance between data classification, performance, and cost, possibly on a workload-by-workload basis.
Organizational readiness addresses the personnel and cultural structures that agencies need in order to leverage AI-driven opportunities. With an AI and data-centric culture, agencies will be able to use data to maximize their AI investments and accelerate innovation.
A precursor to any business change is to have a clear picture of the benefits that change will bring. Acceptance may not always be straightforward, particularly if job roles and responsibilities shift as a result of implementing AI. Agencies that are just starting their AI journeys focus more on total cost of ownership and expected results. More developed projects look at increasing AI performance, and the most advanced projects are looking to see ROI.
Data literacy is starting to be recognized and embraced by everyone—not just data scientists. Every individual generates lots of data, and they should be thinking about how it can be tagged and used.
We’re well beyond the days when AI was the stuff of science fiction. It’s now a viable and valuable tool that agencies can use to drive actionable insights that can help them achieve their mission objectives. But it’s important to lay the groundwork, or else risk having your AI projects fail. If you’re not thinking of operational, technology and organizational readiness, then you may not be an AI ready agency. If you are, you can avoid risks and significantly increase your chances of success.
Melvin Greer is chief data scientist, Americas, Intel Corporation.