Key outcomes
Leveraged AI to unlock years of undocumented business logic at scale.
Successfully delivered a reusable AI extraction framework with built-in guardrails to ensure accuracy and quality.
Client saw productivity gains exceeding industry average:
New sport analysis for the new platform takes less than a day, ~ 98% reduction in time.
Per-module analysis down from two to three weeks to two to three hours.
One to two years’ engineering analysis effort reduced to three to four weeks.
Powering outstanding fan experiences worldwide
A global leader in sports data and technology was modernizing a critical platform to improve data services for its customers across the sports ecosystem, including leagues, broadcasters and gaming companies. However, decades of legacy code and complex manual processes meant that modernization was moving at a slow pace.
The company needed a way to accelerate the move to the new platform, but without allowing rapid modernization to introduce additional risks. To solve this, Thoughtworks, AWS Professional Services (ProServe) and the company’s internal team collaborated to build an AI-powered extraction framework to decode the legacy system at pace.
Challenge: Modernizing faster without increasing risk
Before the company could retire the legacy system, more than 80 sports had to be onboarded to the new platform, a lengthy, manual process that could take eight to ten weeks for each sport analysis.
The primary challenge was that onboarding relied on business logic buried in millions of lines of undocumented Java code. The manual approach to extracting this business logic would take years, and any missed logic would surface as bugs in production, leading to time-consuming rework.
To industrialize the onboarding process and unlock years of undocumented logic at scale, the company needed a reusable AI extraction framework with built-in guardrails to ensure accuracy and quality, and prevent the risk of plausible but subtly incorrect specifications.
Solution: Extracting business logic at speed with trusted AI
As an existing partner, Thoughtworks had already established the architectural foundations and deep domain knowledge necessary for the company’s modernization and AI readiness. Building on this, Thoughtworks collaborated with the client’s engineering team and AWS to build an AI harness with strong guardrails - ensuring logic could be surfaced quickly, systematically and accurately.
Key features of the extraction framework include:
Industrialized logic extraction: Rather than relying on documentation or individual knowledge, the AI framework treats the codebase as the source of truth. It systematically extracts relevant logic into structured data then generates human-readable specifications.
Rapid, reusable pipelines: The framework creates a reusable pipeline from extraction to code generation. This allows rapid parallel execution across sports and modules, with each new sport inheriting the complete framework and shared context, enabling onboarding in less than a day.
Built-in quality and security: By applying rigorous engineering guardrails, the framework maintains robust adherence to enterprise security standards.
Strategic human oversight: The extraction process also includes a human review gate as a critical safety net, ensuring that only SME-validated logic reaches production.
Shared context: Every AI session is governed by a centralized repository of golden rules, terminology, code patterns and quality standards, ensuring total consistency across different modules and teams.
Bridging the reliability gap
Thoughtworks built the AI harness specifically to overcome reliability challenges by establishing a shared context layer and guardrails to ensure deterministic outputs — all supported by a robust architectural foundation.
This approach provides a great example of how to inject AI-powered acceleration into a complex project while maintaining a steadfast commitment to quality. Delivering business value with AI isn’t simply a matter of implementing the technology and watching workflows speed up. It’s essential to have human judgement and decision-making at every stage, so quality is built in from the start and maintained throughout.
I spent over a month searching for some information and the AI-generated specification had it on one page. This was eye-opening analysis, giving me a different perspective and knowledge that was completely inaccessible before.
Outcomes: Rapid onboarding to a futureproof platform
The framework shifted the client’s legacy transformation from a risky, multi-year intervention into a controlled, automated evolution. What’s more, it has changed the economics of the project, producing results far beyond the typical 5% to 15% productivity gains seen across the industry.
With the AI extraction framework in place:
Analysis time per module has shrunk from two to three weeks to two to three hours.
New sport analysis for the new platform takes less than a day, ~ 98% reduction in time.
For a 10-sport program, analysis effort dropped from nearly two years to just three to four weeks or as little as one to two days with parallel execution.
Alongside the AI extraction engine, Thoughtworks, AWS ProServe, and the internal team have been running a second workstream using AI to accelerate a RabbitMQ message-streaming migration to Amazon MSK.
AI-assisted discovery provided a complete picture of the messaging landscape and produced a full migration inventory and a plan with prioritized exchange sequence. The AI also generates a backlog with Jira epics and stories directly from the discovery output, further reducing manual effort. The incremental, data-driven migration is now underway and on track to meet key deadlines.
This engagement has helped the client rebuild their fragile, legacy core into a continuously improving digital foundation. By pairing solid engineering with strategic human oversight, it demonstrated how organizations can safely accelerate modernization and unlock rapid innovation, all while maintaining absolute data integrity.
This partnership shows what can be achieved with the right combination of innovative AI tools, world-class engineering expertise, and deep domain understanding. Together, we’re overcoming challenges that once seemed insurmountable, applying AI to produce fast, consistent, and accurate results at scale.
