Infosys Modernization Methodology Based on Google Generative AI – A New Approach to Modernizing Legacy Code

Abstract:  Legacy code is typically slow and difficult to maintain.  Enterprises are always looking to move to modern architecture so that it helps in maintainability, scalability, deployment to cloud, etc.

However, modernization is difficult due to the challenges in extracting business logic from legacy code like Cobol or VB6.  This is extremely time-consuming and error-prone.

It is here, that Infosys has experimented with using Google’s Generative AI to expedite the modernization process

Infosys Modernization Methodology based on Google Generative AI has been able to extract business rules, and entities from legacy code, and then implement these in modern languages.

The development effort can be reduced by 50% – 70% by using the Google Generative AI approach.

 

Challenges with legacy Code and the need for Modernization
Enterprises have lots of legacy code running critical systems.  And legacy code comes with its challenges.

1.       Monolithic architecture – This makes maintenance extremely difficult.  Fixing bugs, and enhancements is highly time-consuming and error-prone.

2.       Scalability challenges – Legacy code cannot benefit from the advantages of horizontal scaling provided by cloud providers.  As there is a limit to vertical scaling, the legacy code has a limit to the number of users/load that it can support

3.       Non-availability of SMEs – SMEs who understand the legacy code are in short supply

4.       Non-availability of documentation – This makes the transition to new teams impossible

The aforesaid problems are the prime reasons for Modernization

 

The traditional way of modernization

There have been two approaches to modernization.

1.       AS-IS Migration – The legacy code is converted syntactically to the modern language.  There are tools available to do this automatically to a large extent.  But this is only a quick and dirty approach to modernization.  Even though the code is written in modern language, the architecture is still old.  It does not give one the benefit of OOPS, microservices, micro front end, etc.  Therefore scalability is still an issue with this approach

2.       Smart Migration – To alleviate the problems of As-Is Migration, Smart Migration is desired.

 

The classic approach to modernization is below.

a)       Legacy code SMEs to analyze the legacy code, and manually extract the list of entities, business rules, and interdependencies

b)      Validate with the business users.

c)       Modern code SMEs to take the list of entities and business rules extracted in step#1, and create new generation code in modern architecture using OOPS, Microservices, Microfront end screens, NoSQL DBs and the like

Step #2a is extremely time-consuming and error-prone.  This has a dependency on the skill level of the Legacy Code SMEs.  Thus, the process is slow and buggy.

Similarly, Step #2c is once again time-consuming as code must be written from scratch.

 

It is here that Infosys has created a modernization methodology based on Google Generative AI to cut down on step #2a and step #2c.

The Infosys Modernization methodology based on Google Generative AI

 

Migration Workflow
Migration Workflow

The Infosys Modernization methodology uses Google Generative AI for the below steps:

1.       To automatically extract Entities, Rules, and User stories from Legacy Code

2.       Extracted information is vetted by an SME.  SME reviews, and updates the items extracted in #1

3.       To automatically generate modern code based on the entities, rules, and user stories reviewed in #2

4.       Modern code SMEs, can review, and edit the newly generated code and baseline the same

 

Conclusion
The SME effort involved in extracting entities and business rules from legacy code has been drastically reduced.  Similarly, the effort involved in creating modernized code is reduced considerably.  Overall, this methodology has slashed the manual effort in the range of 50%-70%.

Author Details

Vittal Kurpet Panduranga Setty

Generative AI, Quantum Computing

Leave a Comment

Your email address will not be published.