Modernising a mature software platform is rarely about starting again. For Integrated Environmental Solutions (IES), the challenge was how to unlock decades of trusted engineering IP and evolve it for a cloud and AI-enabled future. Working together, IES and Ciklum are showing how AI-assisted engineering via Claude Code can help teams understand complex legacy systems, automate previously manual work, and create reusable foundations for faster modernisation.
For more than 25 years, IES has built software used by engineers around the world to model the energy performance of complex buildings.
At the heart of that platform is deep physics and domain expertise. It is also a substantial C++ software estate that has evolved over decades, with important parts of the product originally designed around desktop applications, Windows environments and proprietary binary data formats.
That history is an asset. It represents years of engineering knowledge, validated calculations and product capability that customers depend on. But it also creates a familiar challenge for established software businesses: how do you evolve something this valuable without sacrificing the trust that made it successful in the first place?
IES’s product direction is increasingly centred on creating a more seamless workflow across building simulation, mechanical engineering, compliance and operational performance. Cloud infrastructure and AI are important enablers, but the physics remains at the core.
Measurable impact
By using Claude Code alongside Ciklum's PRODIGY platform, IES has been able to achieve:
- 4+ weeks saved on onboarding and context-building for a highly niche domain
- 100+ hours saved reviewing and understanding legacy code across the developer pool
- 100+ hours of manual UI testing saved, by unlocking automation across testing
Understanding before changing
Ciklum’s work with IES began not with an assumption that the legacy platform should simply be replaced, but by building an understanding of what existed, where value was embedded, and how the architecture could progressively evolve.
That understanding extended beyond the codebase. The relationship evolved from engineering assessment into product vision and strategy, helping connect IES’s longer-term product ambitions with the architecture and engineering changes needed to realise them.
That meant connecting product strategy with engineering reality: identifying which capabilities should move towards APIs and cloud execution, which tightly coupled components could be separated from the desktop, and where new foundations would be required to support future AI-enabled experiences.
For a mature codebase, however, understanding the system can itself be one of the biggest constraints on change.
Knowledge may exist partly in documentation, partly in the architecture and partly in the experience of the engineers who have worked with the platform over many years. Some source files in the IES estate extend to around 50,000 lines of code.
This is where AI started to change the way the engineering problem could be approached.
Ciklum engineers began using Claude Code directly against the IES VE codebase, supported by dedicated engineering skills created for the project. Claude is used alongside Ciklum’s PRODIGY platform, with its Knowledge Engine module providing IES-specific contextual knowledge to the engineering team, making sense of a large codebase within weeks.
The two work together in a deliberate way.
When an IES-specific question is asked, Claude can call the PRODIGY Knowledge Engine through an MCP integration. The Knowledge Engine searches the information made available to it and returns a grounded answer. Where that knowledge does not exist, it is designed to say so rather than invent an answer. Claude can then validate the response before presenting it to the engineer.
The result is not an autonomous system making engineering decisions.
It is an engineering team with a much more effective way of navigating a complex software estate.
Making the invisible visible
One of the clearest examples came from something apparently simple: reading the data produced by IES’s Virtual Environment (VE), its core building performance modelling and simulation platform.
A number of important IES data structures lived inside proprietary binary files. The VE itself could understand them, but they were difficult to inspect or consume independently, creating a barrier both to automated testing and to future cloud services.
Ciklum engineers reverse-engineered the VE’s serialisation mechanism and created a file-access layer capable of converting binary formats into structured JSON.
AI became part of the engineering method used to tackle this problem.
Because some of the original C++/MFC model classes were too large to work with effectively as a single context, the team created dedicated AI-assisted skills to first produce slim representations of the relevant classes.
From those representations, the team could generate cross-platform C++ classes, implement serializer and deserializer capabilities, and ultimately convert the binary structures into human-readable JSON.
Unit and integration testing remained fundamental throughout, including checks designed to maintain binary compatibility with the existing product.
The scale is notable.
During the work, one developer was able to refactor or port more than 300 C++ data-model classes, with the largest source file around 50,000 lines.
This wasn’t AI generating an isolated piece of code.
It was AI being used within an engineering process to help a developer understand, transform and validate a substantial body of mature software.
From manual checking to automated evidence
The value of making that data accessible quickly became visible elsewhere.
Historically, validation of parts of the VE relied heavily on engineers and QA teams inspecting inputs and outputs through the VE and Vista desktop interfaces.
That approach required considerable manual effort and exposed only a portion of the parameters available inside the underlying files.
Once those structures could be converted into JSON, the team could build automated tests around them.
The resulting pipeline allows the automated quality framework to inspect and compare almost all available parameters, perform more sophisticated tolerance checks and run independently of the desktop UI. The supporting code was also written with cross-platform execution in mind.
That capability became particularly important in another part of the modernisation programme: proving that IES’s APACHE simulation engine could produce trusted results when moved from Windows towards Linux and cloud execution.
Instead of engineers comparing results manually, model by model, the team created an automated validation framework that runs the same models across Windows and Linux, compares the results using agreed tolerance profiles, and produces reports highlighting genuine differences.
For software used in building-performance and compliance workflows, that distinction matters.
Modernisation cannot simply be faster. It has to remain explainable and trustworthy.
Turning individual projects into a repeatable modernisation pattern
The programme is also producing more than the individual deliverables it originally set out to create.
For example, a critical VE module used by engineers to size HVAC systems was tightly coupled to the core desktop product, limiting how easily that capability could be reused elsewhere.
The teams have extracted it into a standalone library that can run headlessly, while validating that its sizing results remain byte-for-byte identical to those produced through the existing UI.
Elsewhere, a BEM API is creating a central cloud location for projects, buildings and models, giving IES Live (IES’s cloud-based building performance platform) direct access to shared model data that previously sat separately within the wider product estate.
Cloud simulation can then operate against centrally managed models rather than depending on the desktop.
Across the delivery streams, six reusable capabilities have emerged:
- A VE file-access layer
- A repeatable pattern for decoupling compute modules
- A headless model tool
- A cross-platform parity framework
- Cloud execution capability
- An AI-assisted legacy-engineering method used daily by the team
That is an important shift.
Instead of solving each legacy problem as an isolated piece of engineering, IES and Ciklum are creating patterns that can be applied repeatedly across the platform.
The first decoupled component teaches the team how to approach the next.
The first binary format becomes the basis for additional converters.
The first cross-platform validation framework becomes a reusable trust mechanism for future services.
And the knowledge generated during that work increasingly becomes available to both engineers and their AI tools.
Accelerating without losing control
AI is often talked about as a way to make software engineering faster. At IES, the more important benefit has been its ability to help engineers make sense of complexity faster.
Claude is being used to interrogate and work through a mature codebase, while PRODIGY’s Knowledge Engine helps bring the relevant IES context into that process. But the acceleration only works because it sits inside a disciplined engineering approach: automated testing, comparison frameworks, tolerance thresholds, structured knowledge and human review remain essential.
The impact is beginning to show in the pace of the modernisation programme. Across several workstreams, the teams have been able to move from discovery into working capability faster than originally anticipated, while also creating reusable tools and patterns that make subsequent work easier.
That distinction matters.
AI-assisted engineering does not remove dependencies, architectural trade-offs or the need for experienced judgement. What it can do is reduce the friction involved in understanding a complex system, shorten the path from investigation to implementation, and allow a relatively small team to tackle a breadth of legacy engineering challenges that would previously have required significantly more time.
For IES, that is where the real acceleration lies: not simply producing code faster, but increasing the speed at which the organisation can understand, change and confidently evolve a highly complex product.
Creating the foundations for an AI-enhanced product
For IES, the significance goes beyond modernising legacy technology.
The longer-term product opportunity is to make its trusted physics and engineering capabilities easier to access across new workflows: through cloud services, APIs, automation, partners and ultimately AI-powered experiences.
That requires more than adding an AI interface to an existing application.
An AI agent that wants to understand a building model needs structured access to the model.
An agent that wants to run a simulation needs headless execution.
An AI-generated recommendation needs a validation framework behind it.
And any output presented to an engineer needs to remain grounded in the underlying physics and evidence.
Those are the foundations IES and Ciklum are now putting in place.
The lesson from the work so far is not that AI eliminates the complexity of legacy modernisation.
It is that, when used alongside experienced engineers, strong product direction and rigorous validation, AI can make that complexity far more tractable.
Twenty-five years of software does not need to become an obstacle to innovation.
Handled carefully, it can become the foundation for the next generation of Innovation.
By Ciklum Editorial Team
Ciklum’s Editorial Board is a collective of experienced writers and industry experts, bringing together perspectives shaped by real-world engineering and delivery experience. Through collaborative insights, the team explores how technology, AI, and digital innovation move from concept to execution across industries.