Beam’s AI Assistant learns from how you estimate. The best way to improve accuracy is to give it real examples and clear guidance. Below are the three most effective ways to train the model.
1. Upload and approve past estimates.
Start with recent projects that best represent how you price work today. Create a new project with the correct address, select a sample client, upload the past estimate via the AI Assistant, and mark it as approved.
Here's a suggested prompt when uploading a historical estimate: “Please draft a bid from this previous proposal. Match the structure, formatting, verbiage, and pricing as closely as possible.”
For your first few projects on Beam, it’s normal to make many hands-on pricing adjustments. Beam’s AI learns from estimates that are sent or approved, so over time it becomes more consistent with your pricing style and expectations, reducing the amount of manual editing needed.
2. Set AI Assistant Rules.
You can give the AI direct guidance on how you typically price and structure bids using AI Assistant rules. These rules help the model align with your standards, from your typical price range to standard exclusions/inclusions to formatting and more. You can find this within Settings → Business → AI Assistant rules.
Examples: list your team’s or subcontractor’s typical labor rates, list your typical price per square foot range, list the tier or finishes for the type of work you typically do, or list how you like to breakout/format your bids.
3. Create a Cost Library.
The AI assistant reads your Cost Library when building estimates, using it for both pricing and line item descriptions. You can find the cost library within Settings → Cost Library. If you have a large cost catalog to import, contact Beam Support and we can help set it up.
Note that the cost library should be updated regularly in order to reflect any pricing changes for specific items over time.