Responsible AI use in support of federal agency missions 

The power of artificial intelligence to enhance federal missions is undeniable. Although federal agencies have been using AI for years with the advancements in generative AI and Large Language Model applications, responsible AI usage has moved to the forefront of the federal technology conversation.  

How can agencies design and use AI to improve their mission outcomes and deliver better citizen services while best protecting the public from potential negative consequences? It starts with understanding that AI is not a panacea that can or should solve every challenge but a tool whose use should be considered very deliberately.

Determining the desired outcome 

Maximizing the benefits of AI while mitigating potential risks first requires determining the desired mission outcome. By identifying where and how AI will be used to help achieve specific results, leaders can discern the best use of AI, based on their exact needs and leverage the right AI tool for that specific goal.  

For example, agencies looking to improve the employee experience by automating routine tasks may invest in less complex AI robotic process automation tools, while federal leaders looking to improve customer experience can invest in chatbots or Intelligent Virtual Assistants with higher levels of technical complexity. Investing in AI solely for innovation's sake risks not getting the benefits an objective-based AI solution would yield. 

Understanding the data 

Once agencies have selected their solution, the next step is to create a comprehensive data strategy or evolve an existing one if necessary. Data sets must be carefully selected and vetted for the given use case to ensure accuracy and appropriateness, recognizing there will always be a measure of partiality, minimal bias that needs to be mitigated or, at minimum, known in context of the output.  

Considerations to enhance data quality include accuracy, completion, consistency, timeliness and originality. A mature data governance framework and integration approach is also necessary to efficiently support AI. Holistic data governance considers all roles, responsibilities and regulations that impact an organization's data usage.   

Addressing aspects like risk management, security and data/model operations allows leaders to establish and maintain the necessary controls to act as scaffolding for secure and scalable data architectures. This approach also accounts for an organization's long-term objectives and technical strategy.  

CX considerations for AI initiatives 

Decisions made leveraging AI, based on accurate and timely data, can advance agency missions and both employee and citizen experiences. Using AI algorithms to analyze large volumes of citizen data empowers agencies to develop more personalized services to meet the discrete needs of the citizens. Intelligent chatbot systems can manage citizen inquiries and deliver rapid responses for more routine needs. This practice dramatically reduces citizen wait time, improves citizen satisfaction and results in stronger employee performance, as they are freed to focus on tasks that require more critical and creative thinking.  

Ensuring the use of AI is intuitive and addresses ongoing pain points is crucial to a positive citizen experience. Before implementing AI solutions at scale, testing and validation by a diverse group of stakeholders, including domain subject matter experts, is necessary to ensure the AI solution results in the appropriate mission outcome and will not result in negative CX impact.   

Testing and validation should check that the solution functions correctly, discover the tool's effects at a micro and macro level, determine if it produces the desired results and ensure it has sufficient guardrails in place to guide the output and alert the user to any shifts that may have occurred. 

In most cases, the training of AI is iterative and requires additional tuning to get to the desired outcome. If the intended use of AI fails to achieve the specific goal or additional or alternative datasets are needed, those alterations should be implemented before deployment. It is critical to have mechanisms to monitor AI after deployment as new and unique scenarios or data may arise that require additional tuning of the model and awareness of the parameters used to generate the output. 

Human-centered design, or the philosophy of keeping people at the center of everything an agency does, is already an ongoing government priority, but the concept becomes even more important when implementing new and innovative technologies like AI.  

Utilizing the principles of HCD from the start of any AI development and implementation process can enhance the overall experience by understanding how humans will be interacting with the technology and the needed outcomes, effectively turning pain points into opportunities. To do this, end-user research is essential. AI developers and integrators must understand the user's objectives, how users interact with the solution and how the solution can help users achieve those objectives.  

Industry and government working together 

With AI evolving at an extremely rapid pace, collaboration, education and transparency are paramount. Coordinating with industry partners that have a demonstrated history of success working with and understanding the government's specific needs at the operational and strategic levels will assist with integrating AI into effective program development.  

Looking to the future, with the right safeguards and responsible development and use practices, AI stands out as a revolutionary force capable of unifying people, processes and technology within all levels of government. By integrating these elements, AI lays the groundwork for a future in which innovation thrives, citizens are better served and governance achieves unprecedented effectiveness and efficiency.