Construction enterprises operate under stringent operational, security, and governance requirements. Bringing generative AI to the jobsite requires more than a generic consumer chatbot: it demands deep construction domain intelligence, data security, and seamless alignment with enterprise cloud infrastructure.
Today, we are excited to announce that Procore AI on Amazon Bedrock is generally available. This integration combines Amazon Bedrock’s secure, fully managed foundation model infrastructure with Procore’s construction intelligence. Procore AI on Amazon Bedrock sets a new standard in enterprise security and model flexibility for Procore's customers as they deploy AI agents across sensitive drawings, species, RFIs and daily jobsite logs.
By integrating with Amazon Bedrock, we are empowering our customers with unparalleled model choice and enterprise-grade security. Procore AI brings the deep construction context and workflows, while Bedrock gives our customers immediate access to the frontier of generative AI, including Anthropic Claude, OpenAI ChatGPT, SpaceXSI Grok, and Amazon Nova. As AWS continues to release new frontier models like Nova Sonic, Procore will look to integrate them into Procore AI so our customers can stay at the leading edge, all while keeping their jobsite data secure, confidential, and governed.
Three challenges stand between construction and deploying AI at scale
Challenge 1: Unstructured Data Silos and Information Chasing in the Field
Over 80% of project data is locked in unstructured PDFs, CAD drawings, submittal packages, and daily photo logs. Field superintendents and trade leads lose hours every day chasing answers across disconnected documents, leading to execution lag and costly jobsite delays.
Challenge 2: Enterprise IT Governance, Data Leakage, and Shadow AI
When field teams adopt consumer AI tools, enterprise CIOs and legal teams face serious risks, including data leakage from proprietary blueprints training public models, lack of role-based access control, and absence of audit trails. Enterprise IT requires zero-training guarantees, IAM governance, and financial alignment with enterprise cloud commitments before approving AI at scale.
Challenge 3: Scaling AI Across Decades of Institutional Knowledge
Moving from AI pilots to enterprise execution requires scaling across decades of historical project records without compromising accuracy or security.
How It Fits Together
Procore AI on Amazon Bedrock addresses each of these challenges. Here’s how:
Construction Context: The Procore platform supplies the project data, workflows, construction intelligence, and Procore AI agents that understand how work happens on the jobsite.
Governance: Procore’s guardrails, knowledge base, and evaluation tools govern how AI operates, giving enterprise IT and security teams visibility and control as AI moves from pilot to production, regardless of which cloud environment you choose.
Model Infrastructure: Amazon Bedrock provides secure, managed access to multiple foundation models.
Data Privacy: Project drawings, RFIs, and financial data stay within Procore’s secure environment. Customer data is encrypted in transit and at rest, and is never used to train the underlying foundation models.
Architecture Deep Dive: Key AWS Services Powering Procore AI
Procore AI’s architecture has two parts: how data moves internally before it ever reaches a model, and the AWS infrastructure that powers the model calls themselves. Here’s how both work.
The Context-Passing Pipeline
Procore AI doesn’t send raw, unorganized project documents directly to a model’s inference endpoint. Instead, Procore’s Magpie reasoning engine acts as an active middleware and orchestration layer:
Pre-Indexing (Generative Indexing): Project files, specs, drawings, RFIs, submittals, schedules, are ingested, parsed, and vectorized ahead of time in Procore’s multimodal index.
Semantic Retrieval: When a prompt or agent task is initiated, Magpie queries the multimodal index to pull only the relevant chunks and records.
Semantic Compression & Data Fusion: Magpie filters boilerplate, condenses long documents and multimodal records into dense summaries, and merges retrieved content with active UI context, user memory, and conversation history into a token-efficient context payload.
Model Inference Call: This structured, grounded payload is sent alongside the user prompt to AWS Bedrock, via secure API call.
For agent actions (not just retrieval), Procore AI separately uses MCP (Model Context Protocol) to call tools on a configured MCP server; results are fed back into Magpie’s reasoning loop to generate the next step.
Procore is available on the AWS Marketplace. Procore leverages the trusted infrastructure of AWS, with seamless integration across:
Amazon Bedrock: A fully managed service that offers unified API access to leading foundation models (Anthropic Claude, Amazon Nova, OpenAI ChatGPT, SpaceXSI Grok), featuring built-in guardrails, model evaluation tools, and zero base-model data training guarantees.
Amazon Bedrock AgentCore: Developer services that power Procore AI agents to execute multi-step workflows and orchestrate agentic tools.
Amazon S3: An object storage service that provides scalable, highly durable storage for raw project artifacts, including high-resolution BIM models, CAD drawings, site photos, and document backups.
AWS Identity and Access Management (AWS IAM): An identity management service enforcing fine-grained role-based access control (RBAC), tenant data isolation, and least-privilege security across model endpoints.
AWS Marketplace: A digital catalog facilitating streamlined procurement and unified cloud billing. With annual SKUs for the embedded Procore AI experience now live, enterprise customers can purchase their digital co-worker directly through the AWS Marketplace and directly draw down against their existing AWS PPA commitments.
More Choice, Without Giving Up Control
For most enterprise IT and security teams, the question isn’t whether AI is useful, it’s whether they can run it securely at scale. Running Procore AI on Amazon Bedrock adds infrastructure-level controls your security team already recognizes, without asking you to compromise on the construction intelligence, governance, or data security already built into Procore AI.
“By combining Bedrock’s fully managed, secure foundation model infrastructure with Procore’s construction intelligence, we are helping customers scale generative AI safely and innovate with confidence. This partnership ensures that organizations have the built-in guardrails and data security guarantees that enterprise IT demands, all while accelerating jobsite productivity.” — Jason Wages, Principal Solution Architect, Amazon Web Services
Meet You Where You Already Are
Most enterprises have spent years vetting their cloud provider, negotiating commitments, and building security reviews around it. With this release, you choose where Procore AI runs:
Embedded Procore AI Experience: Work with Procore to be provisioned on your preferred environment from day one.
Datagrid Enterprise Experience: Select your preferred cloud environment directly in the Datagrid Enterprise interface.
Either way, you get the same Procore AI, with the same security, governance, and construction intelligence. What changes is that your AI workloads can now align with the cloud strategy your IT team already supports.
Building the Future of Construction Intelligence Together with Performance & Latency at Scale
For Procore, building real-time, voice-first experiences for the jobsite required overcoming significant technical hurdles. By leveraging Bedrock’s scalable infrastructure, we optimized latency to support near-instantaneous voice interactions in Procore AI. Furthermore, the ability to switch between model sizes (e.g., using smaller, faster models for simple OCR tasks and larger models for complex multimodal reasoning) has been critical in managing computational costs while maximizing performance.
Get Started
By partnering with AWS, Procore is ensuring that the construction industry doesn’t have to choose between rapid field innovation and uncompromising enterprise governance. As the generative AI frontier continues to accelerate with new models and multimodal capabilities, Procore AI on Amazon Bedrock allows your jobsite technology to keep pace with the cutting edge, all while keeping your data securely within the enterprise boundaries you trust.
If your organization runs on AWS or already uses Amazon Bedrock, you can bring Procore AI into that environment today. Purchase annual SKUs directly through the AWS Marketplace to retire your AWS PPA, talk to your Procore account team about provisioning, or visit procore.com/ai to learn more.
Resources:
Explore on AWS: Visit Procore on the AWS Marketplace to learn how you can leverage your AWS Private Pricing Agreement (PPA).
Join us at AWS re:Invent: Attending AWS re:Invent this year? Stop by and take a Procore AI demo to experience Procore AI and our AI-powered jobsite workflows live.
Learn More: Dive deeper into how industry leaders are leveraging AI by reading the full Haskell case study.
Written by
