Empower AI Announces Jeff Bohling as CEO READ MORE
Originally published by Federal News Network.
The federal government has been working toward information technology modernization for decades. While several agencies have made significant progress, many systems today are still unable to support one of the most vital innovations to federal systems transformation: artificial intelligence.
AI can go a long way toward automating tasks and making agencies more effective and efficient, and funds available through legislation such as the 2022 Omnibus Appropriations Bill represent a unique opportunity for agencies to capitalize on AI’s promise. However, a mature AI posture requires more than the technology itself. Preparation, coordination and a clear road map with achievable benchmarks aligned with agency mission goals are all critical elements of a successful AI strategy.
By following four key steps, agencies can ensure that AI is built into their IT modernization plans so the adoption and integration of these powerful new automation and analytics tools can better empower teams, support citizens and help achieve mission success.
1. Quality data, quality AI
The first step is always about the data. Agencies need a strong data governance process to ensure they have a mature set of data to power the AI they want to implement. Government data sources must be validated, cleaned and standardized.
This step is prone to misconceptions as many agencies think they already have enough of the right type of data. Agencies today are largely capable of collecting a lot of information with the rapidly expanding use of sensors and data input technologies. However, the more information agencies collect and the less they use, the less value they get from their overall data universe.
As a result, agencies often have less viable data than they thought, or what they have is not cleaned, tagged and handled appropriately because they haven’t considered how to acquire, store, analyze and disseminate data effectively.
Without mechanisms in place to support the total data life cycle, implementing AI and machine learning becomes nearly impossible.
To avoid pitfalls, agencies need end-to-end observability of their data so engineers, solution architects, business analysts, data analysts and other team members can both use and build upon their data.
2. Align AI with agency objectives
It’s imperative to align agencies’ business models with AI-powered processes to maximize return on investment. For instance, to improve customers’ experiences with the service desk, it is critical to calibrate AI- or ML-specific objectives within the broader business model. While new technologies can be exciting, scaling IT modernization is as much about business rules and enterprise architecture as the tools themselves. In fact, the easiest way for agencies to over-promise and under-deliver is through a misalignment of business operations and the expectation of AI’s effect on those operations. Too often, agencies rush to automate a process without first understanding how it will affect internal or external customers. This approach can lead to the initial perception that the AI has failed, but the root cause is usually a business alignment problem, not an AI problem.
Agencies can avoid this challenge by asking these questions at the outset: How will AI help me? How do I have to change my processes? What’s the impact on the user? What’s my expected ROI?
3. Get infrastructure AI-ready
Determine each agency’s infrastructure readiness for layering in AI. Returning to the service desk example, the system must have the computer resources necessary to train the models, the network capabilities to move data and the security practices in place to protect everything.
It is also important for enterprises not to ignore data access and control of the systems, particularly when the data resides in scattered locations. Successful AI requires effective data access management and ensuring that the underlying network configuration and integrations are suitable to meet AI requirements.
Lastly, it is vitally important to ensure the architecture is optimized. AI requires modern hardware platforms designed for compute-intensive workloads. But throwing in new hardware is not always the solution. A creative architecture with high bandwidth, low latency and proper configuration fixes many issues, while distributed data processing and streaming software optimize the setup to create AI-compliant architecture.
4. Take AI baby steps
Agencies should start small. Rather than replacing the entire service desk, conduct a pilot test in which AI classifies incoming tickets so that human agents can route or resolve them faster. Once that capability is proved — and supported with metrics — it can be expanded and scaled.
Metrics are imperative to proving ROI and AI application effectiveness, especially when agencies are just starting their AI journey. Metrics data provides needed assurances that small pilot programs are working, particularly as employees are onboarded to new tools and programs. Ultimately, the federal workforce represents the front line of IT modernization, so it is critical to demonstrate value in terms grounded in federal employee experience. An early AI practice should be able to clearly show federal workers and IT leaders how much their capacity has increased as a result of leveraging automation to complete rote tasks.
Laying the foundation for future AI success
AI offers the federal government many benefits, including augmenting agency workforce capacity, increasing the speed and effectiveness of service delivery and saving costs. However, without actively laying the foundation for AI’s success within the framework of agencies’ IT modernization plans, government risks falling years behind organizations currently expediting AI implementation to deliver the experience employees and federal customers have come to expect. These four fundamental steps will save time, lay the foundation for scalable progress and jump-start agencies’ AI-enabled IT modernization journey.
Dr. Allen Badeau is the chief technology officer for Empower AI (formerly NCI Information Systems), as well as the director of the Empower AI Center for Rapid Engagement and Agile Technology Exchange (CREATE) Lab.