4 Ways AI Is Transforming Requirements Management

AI in engineering has seen plenty of use in software development, with LLMs being excellent tools for coding. We even have some examples of this on the main Altium blog, and we think they are pretty impressive. However, on the other side of engineering, where engineers are building physical products, AI has played a lesser role. But there is one area of hardware development where AI can provide huge benefits: engineering requirements management.

If you think about how engineering requirements work, they tend to live in large documents filled with text and tables. They may also include diagrams, but most of the data is in text form. Because this is typically how engineering requirements documents are formulated, LLMs have a real opportunity to shine. They can be used to analyze, summarize, and define processes based on engineering requirements documents. In this article, I'll show you four ways AI is being used to make requirements management less of a headache for engineering managers.

What an Engineering Requirements Document Looks Like

Engineering requirements documents tend to be very long (and boring) documents that describe all of the functional and performance specifications for a product. When that product is a circuit board, this involves a mix of electrical, mechanical, reliability, manufacturing, and usability requirements. Often, these requirements reference specific industry standards or testing methods, as well as conformance requirements that define compliance with the relevant standard or test.

Requirements documents are generally written by a human engineer based on customer research, meetings with product stakeholders, and past knowledge of similar products. However, for engineering management, requirements documents are not very useful for guiding a project. These documents need to be summarized, broken into tasks and milestones, entered into a project management system, and then assigned to team members. This is where AI, specifically LLMs, can help streamline some of these essential tasks.

1. Summarizing Requirements

Because requirements documents can be so large and time-consuming to read, an obvious use of an LLM is to summarize these documents into clear and concise bullet points. The output from the model needs to be actionable in terms of functional specifications, electrical specifications, standards conformance, etc. For electronic systems design or PCB design, requirements that are summarized with an LLM will usually reference a specific standard, part number, part type, or electrical value as part of the specification.

• Good Example: PCB power connector (J4) shall withstand a maximum of 4 A of current across 2 circuits (total of 4 pins) and shall operate at a maximum temperature of 55 °C at maximum current.
• Bad Example: PCB power connector should accept full current load from power supply and not overheat.

The difference here is in the specificity and the language: use of the word "shall," listing a specific reference designator, and stating specific numerical values are all characteristics of a well-written engineering requirement. LLMs are excellent at extracting this data from a long requirements document. As front-end electrical design and schematic capture proceed, the summarized requirements can always be updated to mention additional reference designators, circuits, or schematic sheets.

2. Filling in the Blanks

Another critical task is transferring a specifications list into real engineering requirements as part of a development workflow. AI can take an incomplete requirements document and expand each requirement during the summarization and segmentation process. Elaborating on a short list of functional requirements to form more actionable engineering goals reflects one of the major use cases of LLMs: text generation. This makes your requirements list more complete and offers standards to which requirements can be linked.

3. Parent-Child Sorting

Requirements often form a hierarchy with parent-child relationships. Parent requirements are like a 30,000-foot view that implies multiple child requirements. AI can help sort a requirements document into these parent-child relationships, particularly in cases where those relationships have not been fully specified.

Parent-child relationships between functional requirements could arise for several reasons. In an electronics systems development workflow, a few of these could include:

• Modular design or multi-board design
• Multiple interactions between circuits or groups of circuits
• Relationships driven by embedded firmware for embedded software applications

In many cases, parent-child requirements lists could exist in multiple documents. AI can be applied to each of these to form a requirements hierarchy, helping manage the development workflow.

4. Impact Analysis

Once a requirements list has been extracted from the documentation and refined into a clear hierarchy, the engineering team can begin building the product. But what happens when the team realizes a requirement needs to change mid-development?

The impact of that change is rarely felt in isolation. Most likely, the change affects other requirements, influencing how they will be evaluated, tested, or accounted for in the design. For numerical requirements, this is usually easy, as they are often related by simple equations. However, for text-based requirements, this is more difficult. AI can play a role in evaluating that impact. AI-based analysis of text can help propagate that requirement upwards in the requirements hierarchy, ultimately reaching the key parent requirements for the design.

The Requirements Management Workflow

Effective requirements management software for electronics development teams should take your generated system requirements and organize them into a clear hierarchy. Electronics designers need visibility into these requirements inside their design tool and PDM system, with the best environment giving a direct link between PCB design software, PDM instance, and requirements data.

This is exactly what design teams will find in the Requirements and Systems Portal inside Altium 365. Engineering managers and team leaders can leverage AI to create and organize their requirements, and each requirement can be tagged to a design object inside your PCB project files. To learn more, watch our recent podcast episode with the Requirements and Systems Portal product management team.

Don’t let yourself drown in requirements management tasks, let Altium 365 Requirements & Systems Portal streamline your workflow and keep your design team productive. Experience a seamless transition to a new workflow that expands your reach and capabilities.