AWS Transform mainframe modernization tools are accelerating legacy code migration, but human COBOL expertise remains essential throughout the process, according to a senior AWS executive. Asa Kalavade, VP of AWS Transform, told ITPro that artificial intelligence holds significant potential for workload modernization, yet cannot replace domain specialists.
What AWS Transform Does for Legacy Code
AWS Transform is the company’s agentic AI modernization platform, designed to speed up translation and updates for legacy code, infrastructure, and applications. First unveiled in May 2025, the platform was expanded at AWS re:Invent. AWS Transform for Mainframe specifically handles AI translation of mainframe codebases in languages such as COBOL, JCL, and BMS, and systems including CICS, DB2, and VSAM, converting them into modern languages such as Java.
AWS Transform Mainframe and Code Integrity
To preserve code logic during translation, AWS Transform extracts system management facility (SMF) records from IBM z/OS mainframes. These records contain details on correct implementation steps, execution routines, and workload procedures. Organizations can also compare migrated workload efficiency against existing P90 and P95 performance records to confirm no loss in performance.
The platform’s AI can additionally generate test cases to stress test migrated mainframe code. However, Kalavade stressed that human experts remain necessary at every stage of the modernization process, particularly for pressure testing cloud systems after migration.
Why COBOL Experts Remain Indispensable
In February 2026, shares in firms such as IBM dropped after Anthropic stated its latest Claude model could translate COBOL into modern programming languages. Despite this, Kalavade said COBOL specialists are still critical for validating translated code.
“You do need those COBOL experts, because how do you validate this code? The COBOL expertise is declining. So in fact, we find in our team some of our unicorns are those that have come from years of doing COBOL, but now they know how to translate and they can verify — that whole human reinforcement that is needed to verify that the output is matching the input that is required.”
Asa Kalavade, VP of AWS Transform
Kalavade further noted that COBOL is not a single uniform language. Mainframe applications often include variants such as PL/I and Easytrieve, requiring expert knowledge to define the scope of each transformation.
Blu Age Acquisition and Customer Adoption
AWS has drawn on its 2021 acquisition of Blu Age, a French company with over 20 years of experience in like-for-like code translation. Kalavade explained that the acquisition provided the ability to build control flow graphs of legacy applications, enabling accurate comparison between old and new code.
Customers including BMW and Brazil-based Itaú bank are already using AWS mainframe modernization tools. Kalavade noted that mainframe projects typically take two to five years, making early completions on partial workloads a notable development. IBM z/OS mainframes remain widely used across financial services, government, and healthcare sectors.
VMware Licence Changes Drive Cloud Migration
VMware licence changes have prompted 86% of customers to reduce their VMware footprint, according to Kalavade. This shift is motivating organizations to move workloads to the cloud and adopt containers. AWS is currently beta-testing a dedicated agent for containerization, while AWS Transform handles code within migrating virtual machines.
Kalavade described how administrators can use a chat interface within AWS Transform to ask questions about modernization plans and connect with a coding agent to generate code from natural language inputs. As coding agents become more autonomous, she said, they will increasingly handle long-running tasks such as generating code, creating test cases, and validating outputs — areas where large language models are improving rapidly.