by Brendan Bradley | 3/11/2024
While ambitious goals to utilize generative artificial intelligence (AI) to automate the writing of code have gained popularity with recent advancements in AI, engineering functions elsewhere within the software development life cycle (SDLC) may prove to be better suited for near-term deployment of AI. User competency issues and security vulnerabilities remain considerable barriers to widescale use of generative AI coding tools.
The process of requirements management (RM), which involves steps such as the gathering, analyzing, defining, prioritizing, and validating the needs of development projects, has long been a critical yet time-consuming process for developer organizations. After gathering all input from relevant stakeholders, engineers must manually write associated requirements. The manual nature of this process often leads to errors and ambiguity introduced early on into the development process, leading to late-stage complications that delay project timelines and require intensive rework. By introducing AI into the requirements management process, developers can both speed up the initial requirements definition phases while also introducing intelligence into the review and editing of written requirements. Combined with market demand for increased automation in this repetitive and monotonous manual task, AI-driven RM solutions are well-poised for rapid and widespread adoption within the engineering community. The maturity of AI and machine learning (ML) within natural language processing (NLP) offers strong overlap for use in the generation and analysis of system and user requirements.
Multiple leading RM vendors have already released commercial AI-driven requirements management solutions. While approaches differ, these solutions are largely introduced as complementary capabilities within core RM and application lifecycle management (ALM) offerings:
IBM’s (NYSE: IBM) Engineering Requirements Quality Assistant (RQA) acts as an add-on assistant to its legacy DOORS solution in a similar fashion to that of GitHub’s AI-driven pair programming Copilot tool. IBM RQA leverages the resources and library of IBM Watson’s NLP solutions to help developers refine the definition of requirements and assess the quality of existing requirements.
A longstanding leader in the RM market, Jama Software has also begun to integrate AI-driven requirements management into its portfolio. As the name suggests, Jama Connect Advisor serves a similar function to that of IBM’s RQA, using NLP to guide system engineers and product developers during the authoring of requirements. Jama Connect Advisor complies with industry leading practices including the Easy Approach to Requirements Syntax (EARS), published by the Institute of Electrical and Electronics Engineers (IEEE), and guidance from the International Council on Systems Engineering (INCOSE).
San Francisco-based Visure Solutions has taken perhaps the most trailblazing approach seen thus far to AI-driven requirements management. Rather than the assistant-like, add-on approach of IBM and Jama, Visure has invested considerably in the product-wide integration of AI capabilities within its Requirements Management ALM solution. Embedding AI into all areas of its solution has resulted in capabilities for automated writing of requirements and test cases, automated suggested edits to written requirements, intelligent real-time quality analysis of requirements, automatic standards and regulations compliance checks, and the ability to not only generate requirements, but also generate requirements of requirements.
The outright generation of requirements may prove more of a challenging area for the incorporation of automation in certain development use cases. For those industries facing complex standards and regulations that present developers with a layered and intertwined structure of requirements, relying blindly upon AI to generate requirements introduces considerable risk. Still, developer organizations involved in these areas of engineering may be able to take a pair-programming-style approach, working alongside an AI-driven RM solution assistant to provide useful support in the review and analysis of requirements.
As commercial AI-driven RM solutions continue to be released, offering training and consulting services will prove critical to the continued success and wider acceptance of such tools. While arguably less risky than blindly generating production code via AI coding tools, requirements stakeholders will still need to ensure that requirements generated via AI-driven solutions adhere to quality and accuracy expectations and comply with all relevant standards and regulations.
To learn more about the technologies, practices, and trends impacting modern requirements management, stay tuned for VDC’s upcoming research covering the embedded and enterprise markets for Requirements Management & Definition Tools.