Creating a Modernization Roadmap for a Mission-Critical Energy Platform

A leading energy logistics company depends on a custom software platform to support critical day-to-day operations. Several years after launching a major modernization effort, the platform was working, but the organization around it was struggling to keep pace.

Development teams were still carrying a substantial backlog. Delivery processes had become heavily driven by tickets. Technical debt made changes harder to predict. And with several longtime employees approaching retirement, years of business and platform knowledge were at risk of walking out the door.

The company brought in SPR to assess both sides of the problem: how its teams were delivering software and how the application itself could become easier to maintain and evolve.

The client began modernizing its core platform in 2019 and released the MVP in 2022. At launch, more than 6,000 items remained in the backlog, spanning unfinished functionality, defects, and other needs.

Teams had since reduced that number to roughly 1,300 items, but progress had begun to plateau as new business needs arrived alongside production issues and remaining feature work.

Over time, collaborative software development practices had also given way to complex Jira workflows and a delivery model centered heavily on completing tickets. Production support and longer-running project work often followed the same processes, even though the work itself was very different. The organization also lacked strong data connecting estimated effort with actual effort, making it difficult to understand team velocity or forecast delivery with confidence.

Technical debt added another layer of friction. Business logic was hard-coded throughout portions of the application, logging and tracing were limited, and source-control practices relied heavily on cherry-picking changes between branches.

At the same time, critical business and technical knowledge was concentrated among a relatively small number of longtime employees, some of whom were expected to retire soon.

SPR conducted an eight-week organizational and technical assessment to understand not only what work remained but why delivering it had become difficult.

The team began with workshops covering the platform, business processes and delivery practices. SPR consultants then joined stand-ups, planning sessions and other recurring meetings to see firsthand how work moved through the organization. More than 20 team members, stakeholders, and leaders were interviewed about what was working and where they encountered friction.

In parallel, SPR examined the client’s development methodology, application architecture, and technical processes. Together, those activities allowed the team to compare how software delivery was intended to work with what was happening day to day.

The assessment pointed to several interconnected issues: an operating model that had not scaled with demand, inconsistent delivery practices, concentrated institutional knowledge and architectural friction that made new work harder to predict.

One of SPR’s recommendations was to establish clearer ownership of the platform and give the product organization greater responsibility for requirements, testing, and ongoing stewardship.

SPR also recommended restoring collaborative delivery practices, including cross-functional backlog refinement and regular demonstrations of completed functionality. Rather than managing every type of work through the same process, project delivery and production support could follow approaches better suited to their different needs.

Testing was another area of focus. Business users had taken on a large share of functional testing, adding work outside their primary responsibilities while repeatedly exposing them to defects and regression issues. SPR recommended shifting left with more testing responsibility to product and IT teams, with business stakeholders focused on final validation and sign-off.

SPR also identified targeted changes that could make the application easier to maintain and extend. The team recommended introducing design patterns and greater use of dependency injection to reduce hard-coded business logic, along with structured logging and trace IDs to simplify troubleshooting.

SPR also recommended replacing a branching model dependent on frequent cherry-picking and back-merging with a more standard source-control approach supported by feature flags.

The goal was not to rebuild a platform that was already doing important work. It was to address the areas creating the most friction and make future changes easier to deliver. That same practical philosophy underpins SPR’s approach to legacy modernization without starting over: improving what is holding an organization back while preserving the systems and business knowledge that still provide value.

The client had already begun using GitHub Copilot, but primarily through chat. SPR saw an opportunity to move beyond basic assistance and use more agentic capabilities for targeted development work, including generating tests and supporting coding tasks.

One early recommendation was intentionally focused: enable agent mode and begin with unit-test generation. From there, the organization could expand AI-assisted development as teams gained experience and identified where it offered the most value.

SPR also proposed using AI to address the organization’s knowledge-continuity risk. The roadmap included an AI-powered knowledge repository that could capture application documentation and institutional knowledge before longtime employees departed. Instead of requiring someone to know where a particular answer lived, employees could query that information through an AI agent.

The idea reflects the principles behind context engineering: useful enterprise AI depends on giving systems access to the right organizational knowledge and context, not simply adding a model or chatbot.

SPR did not leave the client with a long list of findings and equal-priority recommendations. The team evaluated initiatives based on both their potential value and the effort required to put them into practice.

That helped separate the work into clear priorities:

• Quick wins that could begin immediately, including restoring regular sprint demos and piloting agentic AI for unit-test generation
• Do-first initiatives focused on stronger platform ownership, cross-functional refinement, better requirements practices, and knowledge capture
• Plan-and-invest initiatives requiring broader organizational change, including restructuring delivery practices and shifting testing responsibilities
• Revisit-later opportunities that could provide value but were less urgent than the issues affecting day-to-day delivery

SPR then translated those priorities into a phased implementation roadmap, giving the client a practical sequence for improvement rather than asking teams to change everything at once.

The assessment also uncovered areas where the company wanted deeper guidance. SPR was asked to assess the potential for Power BI to serve as a reporting platform and evaluate the effort required to migrate a legacy optimization module from a specialized mathematical programming language to Python.

What began as an assessment of software delivery had expanded into a broader modernization effort spanning applications, data, AI and knowledge continuity.

The client had already proven it could reduce a large backlog. The harder question was how to keep new demand from recreating the same problem.

SPR helped identify the forces behind that cycle and separate immediate improvements from changes requiring longer-term investment. The result was a practical roadmap designed around the way the organization works, giving its teams a clearer path to deliver software more predictably, preserve critical knowledge, and put AI to work where it can remove real friction.

Technology and focus areas

GitHub Copilot and agentic AI
Application architecture and modernization
Python
Power BI
Testing and quality engineering
Product ownership and software delivery
Source control and feature management
AI-enabled knowledge management