AI, Software Development
Angular, Angular Material, ASP.NET Core, SQL Server, Azure Government, Microsoft Entra ID, Azure AI Foundry, SAP, DocuSign, Ivanti ITSM, Azure DevOps
A large manufacturer in a tightly regulated industry faced a hard deadline. An upcoming K2 licensing renewal meant five critical business workflow applications had to move off the platform. Those applications supported purchasing, vendor management, employee requests, regulatory disclosures, and device billing, and each came with years of business requests the old platform couldn’t deliver. Green Leaf approached the engagement as a rebuild rather than a like-for-like migration. A senior-led team of three worked in two phases on a modern stack the client owns, inside the client’s own Azure Government environment. The first phase replaced the legacy applications. The second phase added governed approval authority, process reporting, and AI-assisted quote parsing.
A large manufacturer in a tightly regulated industry ran five critical business workflows on K2, a legacy platform with expensive licensing, a small pool of specialists who could support it, and a long list of business requests it couldn’t deliver.
Green Leaf rebuilt all five applications on a modern stack the client owns and delivered them ahead of the licensing renewal. A second phase added governed approval authority, process reporting, and AI-assisted quote parsing inside the client’s own Azure Government environment.
The client runs five business workflow applications that keep critical operations moving: purchase requisitions, vendor master maintenance, employee information requests, a regulatory disclosure workflow, and mobile device billing. All five were built on K2 workflows and SmartForms hosted in SharePoint.
The platform had become a constraint. Licensing was expensive, and an upcoming renewal set a hard deadline. When something broke, the business waited on a handful of people who knew K2. Even modest changes turned into projects, so years of requests piled up with the same answer: not on this platform. Approval authority lived in spreadsheets someone manually reviewed. Requisitioners retyped vendor quotes line by line. There was no meaningful reporting on any of these processes.
In this industry, those gaps carry real weight. Approval authority is a compliance record. Who approved what, at what level, and when it was approved, must remain verifiable. Any replacement also had to run entirely inside the client’s own infrastructure, in Azure Government, under their own identity controls.
Green Leaf treated the engagement as a rebuild rather than a like-for-like migration. A three-person team, led by a Green Leaf architect with two senior full-stack developers, worked backward from the licensing renewal. The team gathered requirements from the live applications rather than outdated documentation. They also asked one question about every off behavior: is this a business rule, or a platform limitation? That question surfaced improvements users had wanted for years. Examples include a request from that forced multi-select checkboxes for a single request type, and requisitions that were locked to an absent buyer.
Speed came from building the foundation once. The team set up a single architecture, reusable workflow components, and a templated starter with authentication, logging, and CI/CD already in place. Then they ran every application through it. The client’s developers already support the chosen stack – Angular, ASP .NET Core, and SQL Server – secured through Microsoft Entra ID. The team integrated SAP, DocuSign, and Ivanti ITSM. Two-week sprints ended in live demos where the client’s architect, designer, and analysts settled decisions in the room.
The work was phased to protect the deadline. Phase one, about six months, moved all five applications off K2 on time and carried the natural improvements with it. The requisitions data model was designed from the start for what came next. Phase two, about two months, extended it without a single breaking change. It added:
The AI work succeeded because of how it was scoped and governed. Green Leaf doesn’t build AI for its own sake. Of four AI candidates considered, three were better served by deterministic rules that are easier to test, explain to an auditor, and keep outside the security review. Quote parsing was the one place AI removed real manual friction that rules could not, so that is where it went. Information security approval of the model, the hosting boundary, and the permitted data was written into the statement of work as a gating dependency. The feature runs inside the client’s boundary with full call logging and admin controls, and the team built an evaluation harness on real quote files before any turning began.
The client is off K2 ahead of its licensing renewal, with the recurring license cost retired. They own the source code, and it runs on a technology stack their own team supports. The specialist dependency that slowed every change is gone.
The cost of change has dropped sharply. All five applications share one architecture and one interface standard. A change that K2 would have turned into a project is now routine sprint work. That is why long-standing business requests finally became affordable.
The client also has capability that never existed before. Approval authority now resolved from governed, effective-dated data. Vendor quotes get read automatically instead of retyped. For the first time, the business can measure its own processes, including approval cycle time, spend by vendor and cost center, and approvals that sit too long.
The strongest signal came from the client. They signed phase two before phase one finished testing. The application owner then demoed the new application to another business unit on his own initiative. That unit asked Green Leaf to bring them onto it, and the team priced the work in days.
Discover how Green Leaf Consulting Group’s strategic IT solutions can bring you growth through technology. Contact Neil McCole at neilm@greenleafgrp.com or (484) 535-4405.