LEGACY IT
MODERNIZATION
WITH AI
Several U.S. Federal Government agencies are modernizing
their legacy, mission-critical systems. MITRE’s Independent
Research & Development Program is investigating the use of
Large Language Models (LLMs) to accelerate IT modernization.
Legacy IT Modernization: An Ongoing Challenge
Government agencies need to transition their legacy IT into less expensive,
more agile systems that provide more efficient service delivery. Legacy
government systems may have decades of regulations and laws that must
be retained in modernized systems for continuity of services and consistent
processing that will hold up in a court of law in case of dispute. Because
of these complexities, agencies are spending almost 80 percent of their
IT budgets on operations and maintenance, doubling the ratio seen in the
private sector (Egan, 2022). Modernizing complex government systems
requires dedicated planning and sustained effort. However, eight of ten
critical federal legacy systems in need of modernization lacked or had
incomplete modernization plans (GAO, 2023).
The Federal Government has been working to modernize systems for
decades, some of which are more than 60 years old (GAO, 2023; Powner
et al., 2024). These agencies are exploring how new AI technology,
such as LLMs, can accelerate and reduce the cost of modernization.
LLM performance for the unique combination of legacy IT languages
(e.g., Assembly Language Code, COBOL, VB6, MUMPS) and complex
government systems remains unproven. To understand the best practices
and mitigate risks with AI-assisted modernization, MITRE’s IT Modernization
(ITMOD) team is exploring LLM logic extraction and code generation from
legacy systems to aid in modernization.
Legacy IT
Modernization with AI
Federal government systems
are large, complex, and must
be assured to succeed due to
the criticality in government
operations. These systems
have complexity and scale
(e.g., millions of lines of code)
unrivaled in training materials
for today’s commercially
available LLMs. As such,
legacy system complexity
and logic extracted from
legacy languages creates
uncertainty and requires new
methodologies for using LLMs
that can only be tested on the
legacy system and language.
