1. Deep Search Agent

Pain Point Addressed: Finding specific information across thousands of drawings, specifications, RFIs, and shop drawings is time-consuming and often incomplete. Critical details like dimensions, materials, or installation requirements may be scattered across multiple document types, and specifications can be overridden by RFIs that aren't always updated in the specs.

Solution: The Deep Search agent uses a specialized model that analyzes drawings with precision, understanding dimensions, locations, materials, and installation details. It can compare documents against each other and cross-reference RFIs that override specifications.

Demo Example: When asked via SMS about elevator speed, the agent found that the original specification was overridden by an accepted RFI. It searched shop drawings, identified the relevant RFI, and provided the accurate answer with links back to source documents in Procore. The agent correctly prioritized the RFI over the original spec—something human reviewers sometimes miss.

Field Accessibility: Accessible via SMS (text-based and voice), requiring no app downloads or training. Superintendents can ask questions from the field and receive instant, accurate answers with source links.

2. Super Assist (Superintendent) Agent

Pain Point Addressed: Field teams record hours of video and audio during walkthroughs, but converting this raw footage into actionable observations or RFIs requires significant manual effort. Quality is often poor, with shaky cameras, background noise, and multiple issues captured in a single recording.

Solution: The Super Assist agent performs multimodal analysis on field videos and audio, extracting observations, issues, and next steps automatically. It processes audio, visual content, and even text visible in videos (like drawings shown on iPads) simultaneously.

Demo Example: A video showed a walkthrough where someone spray-painted a column that was supposed to remain smooth. The agent analyzed the video through multiple stages: extracted and transcribed audio, analyzed visual content frame-by-frame, identified the specific location shown on an iPad screen within the video, and generated a detailed observation including the design documentation error, the improper painting, required corrective action (strip and repaint the column), and the exact location reference. The agent then created this observation directly in Procore, complete with the video attachment.

Workflow: Users simply drop videos from their desktop or send via SMS. The agent can be prompted to create observations, RFIs, or other documentation based on the content. The system shows its multi-step analysis process transparently.

3. PM (Project Management) Agent

Pain Point Addressed: Project managers and project engineers spend enormous amounts of time on submittal reviews, RFI responses, and scope of work documentation. A typical submittal review requires cross-referencing specifications, drawings, and standards—a process that can take days per submittal. Creating detailed scopes of work requires synthesizing information from multiple sources.

Solution: The PM agent executes project management workflows including submittal reviews, RFI management, and scope of work creation. It can analyze complex submittals against all project documentation and company standards, identifying compliance issues and generating detailed review comments.

Demo Example: The agent reviewed a 500-page concrete submittal against specifications, creating a comprehensive compliance table that highlighted which sections were compliant, which were not compliant, and which required further clarification. It generated specific review comments explaining each non-compliance with references to the relevant specification sections. The agent also created a detailed scope of work for concrete operations, pulling information from drawings, specs, and previous RFIs to create a comprehensive document that would typically take hours to compile manually.

Performance: Customers report completing 20 submittal reviews in 2 days compared to 20 days for manual reviews—a 10x improvement in turnaround time.

4. Estimation Agent

Pain Point Addressed: Creating accurate estimates requires comparing current project requirements against historical data, but this information is often siloed or inaccessible.

Solution: The Estimation agent combines historical project data with current estimates to evaluate accuracy and identify potential issues.

Capability: Analyzes current estimates against historical performance to flag items that may be over or under-estimated based on past projects.