TC Energy increases office productivity with automation on Amazon Bedrock

TC Energy is responsible for the integrity and safety of over 93,000 km of natural gas pipelines. To meet regulatory standards, pipeline operators must maintain comprehensive records that show component compliance with pressure and material requirements. These records support safety measures across critical infrastructure.

Traditionally, engineers verified data manually, reviewing technical drawings and cross-referencing the information against large packages of corresponding documentation. But this approach was time consuming and prone to error.

TC Energy saw an opportunity to harness AI to accelerate and standardize the verification of critical pipeline records while maintaining engineer oversight for final review and regulatory compliance. So the company developed an agentic AI platform that can validate the pipelines’ material test reports against weld maps and pressure-testing documentation. Designed to scale beyond a single use case, the platform supports TC Energy’s Enterprise AI Strategy and Smart Office Productivity objectives, advancing the company’s transformation into a smart energy infrastructure organization where AI enhances productivity, strengthens regulatory confidence, and unlocks new business value.

TC Energy had developed a proof of concept that used technology from another provider and could extract data from a single page. But it was unreliable for larger documents, and typical packages could exceed 1,000 pages. TC Energy worked alongside AWS Professional Services, using the team’s deep industry expertise to build Pipeline IQ.

The first agent to be hosted on Pipeline IQ is PIPER, the company’s digital compliance assistant. For faster document processing and more accurate verification, TC Energy built the solution on Amazon Bedrock, which gives access to hundreds of foundation models from leading AI companies. The assistant streamlines the review of complex pipeline records by combining workflow automation with AI-powered document processing.

PIPER uses AWS Step Functions, which orchestrates large-scale parallel workloads, and Amazon Textract, which can extract data in minutes instead of hours or days. This helps transform large collections of reports into insights that the TC Energy team can act on so that engineers can spend less time locating information and more time on validation and decision-making. The result is a faster, more consistent process and a scalable platform that supports future AI innovation across the business.

For document analysis, the team chose Claude by Anthropic in Amazon Bedrock. At this stage, the solution identifies heat numbers, which are traceability codes for steel materials, matches them with corresponding material test reports, and confirms that records are complete.

The company built the solution by using Strands Agents, an open source SDK that AWS released for building agents. By deploying the solution on Amazon Bedrock AgentCore, TC Energy can build, connect, and optimize agents. The team deploys its agents and tools on Bedrock AgentCore Runtime, and Bedrock AgentCore Memory supports intelligent, context-aware, and personalized responses. “The magical thing about Amazon Bedrock AgentCore is that we can add as many agents as we want and scale it up as needed,” says Mitchell Browning, lead business solutions specialist for the natural gas unit at TC Energy.

To help engineers conduct a final review, PIPER retrieves source data quickly and presents results in a simple-to-read table that shows where the information appears in the documents. Provided with pipeline-specific information and some pretested custom prompts, the agent can produce more consistent responses. “We’re not trying to replace engineers,” says Stephen White, pipeline project engineering manager at TC Energy. “We’re trying to enhance what an engineer can do through technology.”

With over 100 active users, PIPER has cut document review time by 85 percent, minimizing manual effort and delays between welding completion and pressure testing. The solution achieved 95 to 100 percent accuracy in early testing on midsize documents of up to 120 pages, reducing operational and compliance risk.

“Pipeline IQ demonstrates how we can use the power of AI on AWS to unlock value from our immense data sets, improving productivity and supporting our employees in focusing on higher-value work. There are huge opportunities for us to continue to use additional agents in the way we work—we are just getting started,” says Rachelle Gordon, director of pipeline integrity program management at TC Energy.

Saving thousands of engineering hours a year, PIPER frees engineers to focus on technical reviews, problem-solving, and design governance. Building on this success, TC Energy is creating more agents to scale Pipeline IQ across the entire business unit of natural gas pipelines. For example, the team designed SCOUT, an AI-powered agent that can discover records among millions of stored pages. Together PIPER and SCOUT are expected to deliver more than $18.7 million in avoided maintenance capital costs over the next 5-year project planning horizon through efficiency gains, accelerated decision-making, and reduced manual effort. “We are entering an era of AI-driven change,” says Paul Ngo, data analytics and reporting manager for pipeline integrity at TC Energy. “Success won’t come from redefining our work but from reimagining how we do it, and we can realize that by working alongside AWS.”