Notifications
Announcements
Currently, at the start of a construction project, we review the project's specifications and extract the prescribed products listed to send the required technical documentation as outlined in our contract. Since we work with multiple suppliers, we often have equivalent alternatives to the products specified in the document.
Next, we submit the technical data sheets for our suppliers' products, filling out the shop drawing identification form provided by the general contractor.
I believe this process takes far too much time and could be easily automated.
I'd like to know if it’s possible to train an AI model that can:
Read an architectural specification document: Extract all the prescribed materials and place them in an Excel file (based on a template).
Allow manual review and selection: Let me review the information in the Excel file. Through dropdown menus, I can choose whether to submit the technical data sheet for the prescribed product or an equivalent product.
Generate complete technical documents: After I submit the finalized Excel file with my choices, the AI would:
Send the documents automatically: Optionally, I could approve the AI to send these technical data sheets directly by email.
2. Change orders (CO)
When I have a contract, I often receive Change Orders (CO), which are instructions that affect the scope of work by adding or removing contractual tasks. To handle these, I need to:
AI Builder isn’t really the right tool for what you’re trying to achieve. It’s best suited to pattern recognition, document scanning, invoice processing, image classification, and similar structured extraction. You could technically force it to work in your scenario, but the result would be expensive to design, train, maintain, and extend over time, and you’d quickly hit limitations when the project grows.
What you’re describing aligns much better with Copilot Agents powered by generative AI. This approach is more flexible, can handle unstructured documents such as architectural specifications, and can manage multi-step logic rather than just extraction. It does require a deeper skill set, particularly in prompt engineering, agent design, and integrating the agent with your existing systems.
The most realistic path is to break your solution down into dedicated agents for each step of the process, rather than trying to make one “super agent” that does everything. For example:
one agent to read and extract prescribed materials into your Excel template
one agent to retrieve and package the technical documentation
one agent to produce shop drawing forms and generate PDFs
one agent to handle email delivery and logging
This keeps your system modular, easier to maintain, and less likely to break when your processes or suppliers change.
From a cost perspective, do look at licensing and ROI. Every automated action and generation run has a cost. If you’re a large enterprise processing thousands of submissions, exploring a dedicated AI foundation or custom model might make sense. For small to medium-sized operations, Copilot Agents are usually the sweet spot.
Under review
Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.
In our never-ending quest to improve we are simplifying the forum hierarchy…
We are honored to recognize Ajay Kumar Gannamaneni as our Community Spotlight for December…
These are the community rock stars!
Stay up to date on forum activity by subscribing.
WarrenBelz 739 Most Valuable Professional
Michael E. Gernaey 343 Super User 2025 Season 2
Power Platform 1919 268