Generative AI for US federal agencies

Generative AI and large language models (LLMs) are poised to transform federal enterprises and missions by enabling natural, user-focused interactions with agency-specific data.

Now is a tipping point. As it entered mainstream usage, AI was primarily used to answer questions. Today, AI can be your partner in creation. Generative AI offers powerful new ways to converse, synthesize, and generate realistic text and multimedia content.

According to Accenture research, up to 40% of all working hours will soon be supported or augmented by language-based AI. For public sector organizations, 39% of working hours have a higher potential for automation or augmentation.

Generative AI will empower federal agencies to increase efficiency, improve stakeholder experiences, and better address mission challenges. What can it do for your agency?

Accenture Federal Services Applied Intelligence brings the latest advances in generative AI to the federal space. Balancing innovation with safety, security, and risk management, we support agencies from strategic assessments through the development and deployment of scalable generative AI applications.

We have experience operationalizing generative AI for mission impact. Our 750+ data scientists, engineers, and AI/ML SMEs apply emerging technologies to federal use cases in complex environments.

We deliver:

• Enterprise strategies to unlock value and insights in legacy data
• People-first user experiences
• Cloud and on-prem solutions to fit mission needs
• Federal-specific generative AI offerings, including FedGPT, our solution for generative AI models tailored to federal requirements
  

We are delivering generative AI capabilities to federal clients in our Next-Gen Mission Analytics (NMA) platform.

NMA is an open-source intelligence platform which delivers near real-time pattern and behavioral analysis of various data, allowing users to curate more than 100,000 global media sources through a common operating picture.

In NMA, analysts can now interact with a generative chat interface using natural language to automatically create complex search strings structured in NMA’s querying language.

These queries create custom, on-demand visualizations from a repository of millions of open-source documents. Additionally, analysts are now enabled to interact with data in their native language. The hours analysts spend researching, finding topics, translating, and structuring queries are now reduced to minutes.

We applied our responsible AI framework to this use case, implementing guardrails to protect against erroneous or biased outcomes.

Gemini AI empowers federal leaders to make informed, data-driven decisions on multifaceted issues such as conflicts, crises and disaster response by combining human judgment with models and simulations.

Working from a human-generated prompt (e.g., how can the U.S. strengthen the semiconductor supply chain?), Gemini AI automatically generates the following:

• Potential scenarios
• Factors driving the scenarios
• Stakeholders in each scenario
• Stakeholders' estimated positions on the issue

These outputs feed into models and simulations which estimate likely outcomes, providing a starting point for further analysis. Before Gemini AI, generating potential outcomes required hundreds of specialized person-hours.

At every step, a human remains involved to check the model’s accuracy and account for bias. Gemini AI provides a baseline analysis in minutes; the result is accelerated, data-driven research to predict collective group behavior.