Support Statement: Overcome the Limits of AI for Large Scale Software Modernization
- Mark Juras
Q: Why should we use gmStudio instead of just using AI?
A: We believe the optimal solution is to use both. Allow us to elaborate…
There is no question that Artificial Intelligence platforms are bringing important and exciting changes to many fields: art, music, science, business, and computer programming. Technical users of AI are already routinely and productively using AI as a living technical reference for research and to expedite their development work. Even this WIKI can be accessed and summarized by chat GPT. The application of AI to the field of large-scale software transformation and modernization is particularly important to GM and we are actively working towards integrating AI into our platform.
Comparing Modernization Tooling: gmStudio + AI versus using AI alone
In our article comparing the tool-assisted rewrite to a more traditional manual-only approach, we describe how teams doing manual upgrades can benefit from also using gmStudio. Likewise, gmStudio can also benefit upgrade teams using AI. Note that gmStudio does not intend to compete with manual development, or AI-assisted development: on the contrary, gmStudio intends complement these activities allowing teams to deliver much more value with less risk and less effort. Some key points on this topic are presented below.
Point | gmStudio | manual-only |
|---|---|---|
Requires a properly trained LLM based on the detailed information from the legacy code and the required reengineering rules for optimal results | no | yes |
Teams may leverage gmStudio to prepare inputs to the AI and integrate outputs generated by the AI | yes | no |
Teams may leverage AI to assist with analysis, redesign, and verification | yes | yes |
Teams should have development skills in both the source and the target platforms | yes | yes |
Provides a flexible, deterministic, precise, and repeatable means of software transformation | yes | unlikely |
Typically require powerful hardware or SaaS | no | yes |
Can transform a collection of interrelated components as a coherent system | yes | unlikely |
May be driven by user-defined transformation rules | yes | unlikely |
Offers an extensible library of pre-built transformational rules | yes | unlikely |
Can process most large systems in less than 1 minute [1] | yes | unlikely |
Automates identifying and removing dead code | yes | unlikely |
Automates consolidating redundant code | yes | unlikely |
Automates customized COM API replacements | yes | unlikely |
Automates customized language transformations | yes | unlikely |
Automates integrating hand-written code | yes | unlikely |