Generative AI offers significant potential to accelerate code development. Despite the growing legal and compliance framework for using AI in software development, most assignments prohibit the use of large language models (LLMs) is limited – often due to concerns around intellectual property.

Quality Assurance, may be a domain where LLMs can be effectively applied while remaining within the constraints of confidentiality. To mitigate the risk of leakage of client confidential information, all LLM processes can be confined to a secure local environment.

To measure the gains of this approach, we selected a set of communication protocols for which we already have well – documented test frameworks, including accepted outputs and historical effort – to – build metrics. We conducted an internal pilot project where we used an LLM to generate test scripts for two such protocols. The objective was to assess the performance of this method and to document both best practices and limitations.

We built a structured, reusable test automation framework comprising code (configuration, core logic, test APIs, and modules) and documentation (design specs, setup guide, test cases, user guide, and reports).

GenAI accelerated overall delivery by up to 50%, with the greatest gains in design and documentation (up to 71%), and study (up to 60%).