In a recent LinkedIn post, Teresa Torres discusses the significant challenges and innovative solutions surrounding the automation of bid requests within the construction industry. Torres highlights how companies like Tendos AI are leveraging advanced agentic systems to streamline a process traditionally bogged down by manual, error-prone tasks.
The post delves into the complexities of handling bid requests, which often involve parsing lengthy PDF documents and determining product relevance for pricing. Torres explains the core problem: “When a construction company receives a bid request, someone has to open that email, parse the attached PDF (sometimes 1,800 pages describing an entire building), figure out which products are relevant, look up pricing, and draft a quote—all before the deadline. It’s tedious, error-prone, and surprisingly manual.” This detailed description sets the stage for the technological advancements Tendos AI has implemented.
The Power of Agentic Systems in Construction Tendering
Teresa Torres elaborates on Tendos AI’s journey, which began with a focused prototype for matching radiator requests to product catalogs. As Torres points out, this initial narrow scope allowed the company to “Start narrow to prove value,” a key takeaway from their experience. The system has since evolved into a comprehensive agentic workflow capable of handling everything from email categorization to offer generation. Torres emphasizes the collaborative nature of these systems, noting the development of a multi-agent architecture where specialized agents work together.
Validation and Iterative Development
A significant portion of Torres’s post focuses on how Tendos AI validated and developed their solution. According to Torres, the company engaged with a design partner and spent valuable time on-site observing users. This hands-on approach was crucial for understanding the intricacies of the workflow. Torres highlights the importance of owning the user interface, stating, “Own the interface: building a web application (vs. integrating into legacy systems) gave them control over UX and the ability to iterate toward full automation.” This control allowed for rapid iteration and refinement of the automated process.
Ensuring Accuracy and Progress Towards Self-Learning
Torres further details the technical underpinnings of Tendos AI’s approach, particularly the emphasis on evaluating individual agents. “Evaluate each agent, not just the chain: per-agent evals make debugging tractable and show exactly where performance changed,” Torres explains, underscoring the benefit for identifying and resolving issues efficiently. The post also introduces the concept of a “review agent,” a dedicated component designed to check the work of other agents before human intervention is required, akin to a code review process.
“Use review agents: a separate agent that checks work (like code review) catches errors before they reach humans.”
Furthermore, Torres discusses the development of custom observability tools when off-the-shelf solutions proved inadequate for the complexity of the agentic system. The insights shared by Torres point towards a future where human-in-the-loop feedback is instrumental in pushing the system towards becoming self-learning. The strong market signals, with customers requesting Tendos AI to replace their existing CPQ software, indicate a clear product-market fit, as noted by Torres.
📝 About This Content
This article is based on insights shared by Teresa Torres on LinkedIn.
📅 Originally posted on January 15, 2026 | View original post on LinkedIn →