Agentic AI Framework for Mainframe Modernization at Scale

western_union_/_unum · Finance · 2025 · source Agentic: yes · promotional bias: med · source-backed · extracted with prompt 0.5.2 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Mainframe modernization

Western Union and Unum needed to modernize large COBOL/mainframe estates faster than traditional migration programs.

Constraints:

Legacy code scale
Multiple partner tools
Tool overlap/confusion
Need testable converted applications

Architecture

The mainframe modernization framework composes AWS Transform, partner agents, customer agents, and domain tools through an orchestration layer. AWS migration/modernization agents handle assessment and code transformation; third-party agents from Accenture and Pega contribute industry knowledge and process modeling; customers can register their own agents; orchestration can run agents side-by-side under a supervisor or linearly through transformation, testing, and data migration. Western Union used the framework for COBOL code conversion and documentation; Unum used Pega Blueprint and modernization agents to collapse claims-manager and claims-examiner workflows.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  customer-modernization-program[Customer modernization program] --> agent-orchestration-layer[Agent orchestration layer]
  agent-orchestration-layer --> aws-transform-agents[AWS Transform agents]
  agent-orchestration-layer --> partner-industry-agents[Partner industry agents]
  agent-orchestration-layer --> customer-agents[Customer agents]
  aws-transform-agents --> code-conversion-workflow[Code conversion workflow]
  partner-industry-agents --> business-process-modeling[Business process modeling]
  code-conversion-workflow --> testing-and-validation[Testing and validation]
  business-process-modeling --> testing-and-validation

Drag to pan · scroll or use +/− to zoom · reconstructed from the source's described component topology

Components

Component	Role	Interacts with
`Customer modernization program`	Enterprise migration effort such as Western Union or Unum.	`Agent orchestration layer`
`Agent orchestration layer`	Coordinates AWS, partner, and customer agents in parallel or linear workflows.	`Customer modernization program`, `AWS Transform agents`, `Partner industry agents`, `Customer agents`
`AWS Transform agents`	AWS migration and modernization agents for assessment, transformation, and conversion.	`Agent orchestration layer`, `Code conversion workflow`
`Partner industry agents`	Accenture and Pega agents providing domain knowledge, documentation, and process modeling.	`Agent orchestration layer`, `Business process modeling`
`Customer agents`	Customer-provided modernization agents plugged into the same orchestration layer.	`Agent orchestration layer`
`Code conversion workflow`	COBOL/mainframe code conversion, technical documentation, and testing path.	`AWS Transform agents`, `Testing and validation`
`Business process modeling`	Pega Blueprint and partner analysis for process modernization.	`Partner industry agents`, `Testing and validation`
`Testing and validation`	Checks converted applications, data migration, and like-for-like outcomes.	`Code conversion workflow`, `Business process modeling`

Data flow

A modernization program registers AWS, partner, and customer agents. The orchestration layer decides whether agents run in parallel or sequence, combining discovery, code conversion, documentation, process modeling, testing, and data migration. Outputs are evaluated for like-for-like behavior, performance, documentation, and process experience before customer teams proceed.

3 supporting quotes

“tools together and drive the co-build and co-ell motion. So how we do that AWS migration and modernization agent and then we have the third party agent registration where we talked about with Aenture and PGA and the customer can bring their own”transcript, stated

“how we do that AWS migration and modernization agent and then we have the third party agent registration where we talked about with Aenture and PGA and the customer can bring their own agent also and then we have this agent orchestration layer”transcript, stated

“and PGA and the customer can bring their own agent also and then we have this agent orchestration layer and I will talk more about the agent orchestration that's where we design the overall agentic workflow whether these agent is going”transcript, stated

Why this architecture

The framework exists because modernization programs face tool overlap, stakeholder misalignment, and integration challenges. A composable orchestration layer lets enterprises use the best AWS, partner, and customer agents together instead of choosing disconnected tools that operate in silos.

3 supporting quotes

“your uh workload but it there were slowdown and this slowdown happening because of the tool confusion So these these different tool proposal coming to them there is overlapping tool how to take the decision and move forward. Second is stakeholder ”transcript, stated

“or the code conversion or testing or or the integration these all working in silos how to bring them together so that was the another challenges that's where we introduced composibility What composibility do it bring the best out of the all”transcript, stated

“then uh partner like Asenture and PAGA has the differentiator we bring the best of the all the tools together and drive the co-build and co-ell motion. So how we do that AWS migration and modernization agent and then we have the third party”transcript, stated

Models used: AWS Transform agents · Pega Blueprint agents · Accenture agents

Scale

Western Union converted about two and a half million lines of code in the pilot context.
“read uh the details here, right? Uh just over uh two just about two and a half million lines of code. Um I want to talk about speed here. Uh so in throughout the pilot uh which obviously we're still running through but we've got some good”transcript, stated
Western Union converted 53,000 COBOL lines in about an hour and a half.
“Um so not only that but uh the 53,000 lines of uh cobalt uh were converted in just about ”transcript, stated
Unum eliminated a 7,000-hour yearly claims-manager workload and moved from a seven-year, $25M estimate to a three-month testable application.
“does this have on lives? Um, remember our claims our claims manager experience? 7,000 hours a year for our claims manager. It's gone. It's eliminated. Right? Figure out what to do with your time now because your team's got 7,000”transcript, stated

Alternatives considered (roads not taken)

Choose one modernization tool or partner stack

Rejected because: The source frames overlapping tools and integration silos as the reason for composition.

“the tool confusion So these these different tool proposal coming to them there is overlapping tool how to take the decision and move forward. Second is stakeholder misalignment. If someone is proposing different tool versus different pattern”transcript, stated

Run tools independently in silos

Rejected because: The orchestration layer coordinates discovery, transformation, testing, and migration across agents.

“orchestration that's where we design the overall agentic workflow whether these agent is going to work side by side like if I have the uh the FSI agent uh FSI knowledge base agent by centure and then I have this knowledge base um uh this code I'm creating”transcript, stated

Patterns

`multi-agent-orchestration` (modernization)

Multiple specialized agents and partner tools are composed into a modernization workflow.

“co-inovation with our partner as uh with Western Union built This composible agentic AI solution ”transcript, stated

`tool-use-orchestration` (modernization)

The architecture reduces confusion from overlapping modernization tools by orchestrating them.

“uh partner and AWS to avoid that market confusion this tool versus that tool because ”transcript, stated

Tradeoffs

How to combine modernization tooling.

Chose composable agent orchestration across AWS, partner, and customer agents over isolated tool selection.

Rationale: Composition addresses tool confusion and lets each agent contribute its differentiated capability.

“What composibility do it bring”transcript, stated

Gotchas (what bit them)

Enterprises face tool confusion during modernization.

Root cause: Many partner/ISV/GSI tools overlap and slow decisions.

Resolution: The composable agentic framework coordinates the tools.

“So these these different tool proposal coming to them there is overlapping tool how to take the ”transcript, stated

Evaluation

The case reports code-conversion volume and project-timeline reductions.

Metrics: lines of COBOL · conversion time · timeline reduction

“nd Pega Blueprint agents) worked together through orchestration layers. Results included converting 2.5 million lines of COBOL code in approximately 1.5 hours, reducing project timelines from 3+ months to 6 weeks for Western Union, and achieving a complete COBOL-to-cloud migration with testable appl”summary, stated

Outcomes

The summary reports converting 2.5M lines of COBOL in about 1.5 hours and reducing timelines.

The results are reported across Western Union and Unum examples.

“5 hours, reducing project timelines from 3+ months to 6 weeks for Western Union, and achieving a complete COBOL-to-cloud migration with testable applications in 3 months for Unum (compared to previous 7-year, $25 million estimates), while eliminating 7,000 annual manual hours in claims management.”summary, stated