A legacy payroll & HR platform on the edge of shutdown one budget, two ways to lose, a closing market window

A municipal public tender asks specific things, and each had to be answered to even bid. The platform had to be secure enough to pass procurement on data protection, priced to beat the rival vendors bidding for the same cities, fast enough to demo without the daily complaints, and predictable enough to deliver on a promise. The estate Kurt inherited met none of these. Acquired years earlier and never integrated, it ran on ageing servers carrying critical vulnerabilities, with thousands of custom scripts nobody could still explain.

Rather than a task list, the turnaround became one strategy broken into a few parallel tracks, each answering one of those demands:

The long backlog came first, including the compiler upgrade that had been bleeding delivery capacity and shipping decade-old packages; build automation, a security baseline, proactive monitoring and the reorganisation ran alongside it. Clearing that backlog meant putting roughly five years of nearly-finished work live in a single push, which was always going to be turbulent: five years of problems that had never been spread out arrived together. Holding it one more year would have been worse, adding another year of problems on top and another year of the team losing a large share of its planning capacity to maintaining two code bases at once. The turbulence was the cost of the waiting, surfaced by the release rather than caused by it. Sometimes the only way past accumulated risk is straight through it, and the longer the wait, the worse it gets.

Security and a working demo got the platform to the door of a tender. Price decided whether it could walk through, and the segment was contested: other vendors bid for the same cities, so the figure had to beat what they were offering on value. The licence was billed per registered user, which charged for access for every client, when granting access costs almost nothing, a single row in a table. The real load comes from somewhere else. The application is a heavy desktop client, so what consumes a server is the instances open at any one moment, each reserving memory and processors, and one person may keep several open. Repricing the licence to concurrent instances tied it to the figure a server can actually be sized against. That mattered most for the segment in view: public bodies carry many occasional and seasonal users, so a per-registered-user price would have inflated the bid as if every one of them worked at once, and quietly closed the door on the cities the unit was trying to win. With the IaaS operating cost cut alongside it, the price followed the capacity the software actually used. That is what the bid figure rested on.

The same re-architecture decided a deal already on the table. An existing public-sector customer who had been weighing whether to stay was kept on board by the re-architected, managed infrastructure.

The platform had been delivered by three teams that had grown up in separate silos without clear ownership, so finished work was rarely agreed across the seams. Central IT, meanwhile, had gone unheard for years.

The reorganisation split the unit into a customer-facing side and a product side, and gave the product side a product focus: cross-functional teams that each hold the skills and the dedicated capacity to take a change all the way to done. Before, a change could only be functionally tested by a separate team that had no capacity set aside for it, so it waited, and the wait was the bottleneck. Moving that responsibility and that capacity inside the product teams is what loosened the flow: releases came faster and landed more predictably. Some of that capacity had to be created from scratch. There had been no functional analysts at all, so an analyst was hired to coach the team, ten people were trained in the discipline, and a Scrum Master came in alongside them. The IT budget was grown by design to fund the change, releasing some roles and hiring the capability the turnaround needed.

A separate enablement team carried DevOps, automation and infrastructure, the standing capacity that keeps raising efficiency over time. Without that floor, the product could not be offered at a price that made the business work. Work the third line did was documented and handed down to first and second line, so support could resolve more on its own.

The counterparts in these decisions were the group's central-IT directors of architecture, infrastructure and business applications, the function that supplied the unit. Their security concerns became the unit's own concerns, because a breach of unprotected client data would end the business for everyone. The unit that had been held together by the goodwill of a few key people started to hold together by design.