Skip to main content Skip to footer

Generative AI for US federal agencies

Apply emerging AI to solve mission challenges.

A new era of AI

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

At the outset of mainstream adoption, AI was primarily used to answer questions. Today, AI can be a 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?

Federal use cases

Intelligence synthesis & discovery

Quickly derive information from large amounts of text data, presenting pathways for analyst insights.

Assistive coding & vulnerability research

Enable more efficient programming by providing code scaffolding and generating comments, documentation, and unit tests.

Conversational services & customer communications

Create real-time customer response communications for help desks and citizen services.

Knowledge management & enterprise data transformation

Enable conversational interaction with enterprise data.

Content generation & design inspiration

Generate content in a specific style, or mock concept and prototype design examples.

Our services

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

Federal generative AI, enabling faster mission outcomes

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.

A screenshot of NMA’s UI with the generative AI chatbot on the right formatting a user’s input in NMA’s query language.

VISION

VISION (Versatile Intelligence System for Information and Operation Needs) 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?), VISION 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 VISION, 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. VISION provides a baseline analysis in minutes; the result is accelerated, data-driven research to predict collective group behavior.

Trust at the core

Learn more from Generative AI Lead Mike Thieme on the top five focus areas for federal generative AI.

Generative AI brings new challenges to government agencies, including risks of data leakage, adversarial use of generative tools, and potentially incorrect outputs from trained models.

Our approach to mitigating risks in generative AI for federal agencies is rooted in five core principles:

  • Trustworthy: Safe, honest, and diverse in perspectives
  • Reliable: Designed to enhance judgement and decision-making
  • Understandable: Interpretable and transparent in decision-making
  • Secure: Prioritizing privacy and the security of agency data
  • Teachable: Human-centric design supports human + machine co-creation

Our team continually assesses the generative AI landscape, bridging commercial innovation and federal priorities to build strategies and solutions that meet mission needs.

READ OUR PERSPECTIVE ON THE TOP FEDERAL IMPACT AREAS

FedGPT: generative AI tailored for federal use

FedGPT delivers generative AI capabilities to fit federal requirements and use cases with explainable, accurate and traceable outputs. The platform complements and strengthens existing enterprise search capabilities and is deployable in the cloud and on-prem.

Wide applicability

FedGPT can function as a standalone, air-gapped capability – or it can be integrated into third party applications and APIs. It can also serve as a custom content generator for the federal space.

Cost-effective

The combination of models and skills can be run in CPU or GPU environments, bringing cutting-edge AI/ML capability with reasonable resource investment.

LLMs, federalized

FedGPT federalized LLMs for security, safety and reliability, monitoring model prompts and outputs. We apply methods to mitigate the risk of hallucinations that are inherent to generative AI models.

Full-scope generative AI skills

FedGPT’s modular skills offer a fully comprehensive skill set to support cross-federal agency requirements. These include Chat, Search, Draft, Translate, Query, API, Summarize and CodeQA.

Modular and customizable

FedGPT's modular skills can be combined and customized to fit specific tasks and data sources. The skills are flexible and model-agnostic to meet enterprise and mission needs.

Reliable documentation

Citations are provided for each query, linking model outputs back to ground-truth documents. Confidence scores show alignment of model outputs and source content.

How we deliver

Federal generative AI center of excellence

Our Federal Generative AI Center of Excellence brings together generative AI practitioners to design prototypes, scale pre-built solutions and accelerators, and apply models to the real-world mission challenges that federal agencies face every day.

Accenture Federal Services collaborates closely with Accenture's Generative AI and LLM Center of Excellence, adapting Accenture's commercial generative AI capabilities to accelerate transformation for federal missions.

Related capabilities

Federal cloud computing

Powering secure and highly scalable cloud delivery for the U.S. government.

Federal Data & AI

Delivering AI innovation to help agencies unlock value from data at enterprise scale.

The Forge®

Where tech, design and talent converge to solve problems and create more value.

Meet the team

Denise Zheng

Chief AI Officer and Data & AI Practice Lead
LinkedIn

David Lindenbaum

Associate Director – Accenture Federal Services, Generative AI Center of Excellence Lead
LinkedIn