AI in Requirements Engineering Processes

AI in Requirements Engineering Processes

Requirements engineering remains a major bottleneck in many industrial quotation processes. The coordination of internal experts, collection of technical and commercial data, alignment with customer specifications, and the creation of customized offers are often time-consuming and resource-intensive tasks. Generative AI now offers the opportunity to streamline and partially automate these tasks without compromising quality or traceability.

We are developing AI solutions to support these processes with a focus on real-world application. Our approach combines natural language processing, document structuring, and intelligent task assignment to reduce process friction and accelerate time-to-quote. The basis is a deep understanding of typical pain points in RFQ handling and their distribution across different process stages - from initial feasibility checks to final documentation.

Common Challenges in Requirements Engineering

Based on industry feedback and practical experience, seven major challenges were identified:

1. Assessing RFQ relevance and scope alignment
2. Feasibility checks involving multiple departments
3. Technical clarifications with customers (e.g., spec conflicts)
4. Assembling and updating relevant information
5. Creating a consistent and complete quotation
6. Managing customer changes and follow-up questions
7. Documenting the process and deriving lessons learned

Each of these steps requires expert involvement, detailed analysis, and coordination between departments—often with limited reuse of historical data or prior knowledge.

AI Support Across the Process

To address these pain points, we have developed a modular AI assistant concept based on the capabilities of large language models (LLMs).
These include:

• Information extraction: Identifying and tagging relevant content across different formats (e.g., PDFs, emails, spreadsheets)
• Task assignment: Automatically assigning tasks to departments based on content categories
• Search and retrieval: Locating similar historical offers or internal documents
• Semantic comparison: Matching current RFQs with past offers for reuse potential
• Draft generation: Pre-filling offer templates with structured information
• Highlighting critical data: Flagging inconsistencies, red flags, or key decision points

These AI functions enable substantial time savings while ensuring that domain experts remain in control of critical evaluations.

Use Case Example: Copper-based Electronics Component

An incoming RFQ requests a copper-based assembly for an electronic application. The AI assistant:

• Identifies technical specs in the RFQ
• Matches components from previous offers
• Flags new compliance risks (e.g., export control to a restricted region)
• Drafts a response document including pricing ranges and lead time estimates
• Supports coordination with engineering and procurement
• Documents the process and updates knowledge for future requests

This approach does not aim to replace expert input - but to structure and accelerate it.

Accuracy and Usability as Key Design Principles

Two critical success factors for AI integration are speed and precision. The system must return relevant suggestions in near real-time to ensure user acceptance. At the same time, it must minimize so-called Type I and Type II errors:

• Type I (false positive): Irrelevant information is marked as relevant
• Type II (false negative): Critical information is missed entirely

While both errors are disruptive, false negatives pose a higher risk to process integrity and must be minimized through training, prompt optimization, and user feedback loops.

Strategic Perspective

AI-supported requirements engineering is more than a productivity tool. It enables scalable quality assurance, consistent knowledge reuse, and greater responsiveness in competitive bidding environments. Especially in industries with frequent customer-specific requests and long quotation cycles, this approach can lead to substantial time and cost advantages.

At INC Innovation Center, we are currently building pilot systems and are seeking companies interested in early-stage collaboration. The goal is to co-develop practical solutions that align with real workflows and deliver measurable business value.