Establish localized Large Language Model (LLM)

Overview

Development stage

PilotSource: b) Pilot – The use case has been deployed in a limited test or pilot capacity.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

The tool, once deployed, is only for productivity and does not impact any of the defined high-impact areas

Start date

2025-09-10 (day precision)Source: 9/10/2025

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Information Technology

Operational functions

Search and knowledge retrieval, Cyber threat detection, FOIA and records management Derived by keyword rules; see methodology

Problem the AI is intended to solve

AI Enablers (Enterprise solutions that enable AI to integrate into IT services infrastructures)

Expected benefits

Consolidation of multiple research initiatives into an aligned approach to research an enterprise LLM that will be trained on approved internal documentation to provide accurate, consistent, and rapid responses to employee inquiries. This will improve service delivery, reduce case escalations, and free personnel to focus on complex, mission-critical work

System outputs

AI-Powered Service Performance Metrics. Efficiency and Responsiveness: Track reductions in response times versus the pre-LLM baseline, first-contact resolution without escalation, and inquiry deflection handled entirely by the LLM. Accuracy and Consistency: Evaluate response accuracy through audits against official documentation, measure alignment with current Fiscal Service policies, and track reductions in incorrect or inconsistent responses compared to manual baselines. Security and Compliance: Ensure zero unauthorized access through effective role-based controls, conduct quarterly reviews for compliance with federal privacy, cybersecurity, and records management standards, and pass all required internal and external security audits. User Adoption and Satisfaction: Measure internal adoption rates, collect user satisfaction feedback through surveys or ratings, and assess training effectiveness using feedback and usage metrics. Operational Impact: Quantify staff hours reallocated from repetitive inquiries to higher-value work, track operational cost savings, and measure the breadth of internal systems and processes covered by the LLM’s knowledge base.

Technology

AI classification

Generative AISource: Generative AI: AI that generates new or synthetic content (e.g., images, videos, audio, text, code).

System name(s)

Not reported

Custom-developed code

Yes

Public source code

Not reported

Sourcing

How it was built

Developed in-houseSource: b) Developed in-house

Vendor (as reported)

Not reported

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

No

Privacy Impact Assessment

Not reported

Authorization to Operate

No

Demographic features

k) None of the above

Training and evaluation data

N/A. Evaluating tools only at this stage

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported