Building community solar gardens at the pace Novel Energy Solutions LLC is keeping leaves little room for missteps.

The Saint Paul, Minnesota, builder of community solar gardens has completed more than 75 projects since its 2012 founding, including a massive 10-megawatt facility for Minnesota’s largest electric cooperative (pictured) completed last June. More than 100 projects are in development or construction.

Each requires managing a mountain of design, permitting and scheduling documents amid policy shifts, specialized equipment shortages and multistate permitting demands. To keep up, Novel is employing artificial intelligence agents to automate the engineering tasks most vulnerable to human error.

The company has begun adopting Procore Technologies Inc.’s new prebuilt AI agents and its recently announced Agent Builder custom agent creator to reduce manual workloads, improve design accuracy and accelerate turnaround times on engineering and field coordination tasks. The initiative is still in its early stages, but Novel Project Engineering Manager Kiara Carriere said the impact is already tangible.

Little margin for error

Carriere leads a team of technical project managers who are responsible for ensuring that engineering artifacts accurately reflect what happens in the field. Historically, those processes were labor-intensive, involving manual compilation and spreadsheets for root-cause analysis, change tracking and cost impacts. “It’s very slow, and it’s very easy to miss things if you’re doing it manually,” she said.

With more than 50 construction sites active at a time, manual documentation grew increasingly impractical as the company grew. There was also the risk of communication gaps between remote engineering teams and on-site personnel. Details specified early in the field coordination process, such as changes to trench routes or battery storage specifications, could be forgotten or overlooked when the final documentation that guides construction was created.

Novel’s first use cases have centered on an engineering-focused agent it built with Procore Agent Builder. Carriere configured the agent to scan requests for information, submittals and other project documents for changes that impact design. The agent uncovers issues such as module changes, trench route modifications or battery system adjustments so engineers can incorporate the correct information into their designs.

The agent also identifies patterns that may raise red flags, such as recurring permitting objections from local jurisdictions that can significantly delay projects. “For a project with multiple design changes, we may have to permit over 40 times,” Carriere said.

In one case the agent revealed that a specific town repeatedly flagged questions about the battery energy storage system. “That gave us the chance to fine-tune how we start designing battery projects,” Carriere said. That review and planning cycle could have taken up to three months, but the agent made a one-month review possible.

Reduced documentation errors

One early benefit involved detecting discrepancies in as-built drawings, which are blueprints that document the final condition of a construction project. As-built accuracy is essential, especially for transformer ratings used by utilities to validate safe interconnection. Errors can lead to permitting delays, change orders and potential damages.

The agent identified as-built redlines that the engineering team had discussed in earlier correspondence but overlooked in the final documentation package. Carriere said Novel has used agents to significantly reduce the risk of producing inaccurate as-builts and the downstream schedule impacts that result.

While Novel’s deployments currently operate in a question-and-answer mode, the company plans to move toward full agentic behavior. Director of Business Technology Andrew Smith said the next phase involves enabling agents to take actions such as assigning tasks or generating reports that once required a human. Procore’s approach preserves human oversight by documenting plans and requiring approval.

This human-in-the-loop model gives Novel confidence that it can incrementally enable automation across departments, from engineering to safety and cost management. The centralized nature of Procore’s platform, Smith noted, also means agents operate with full contextual awareness: “It’s looking at every single thing that we’ve done within our project,” he said. “It can reference all of the details.”

Although AI has the potential to displace some skills, Carriere said she sees the technology as extending rather than replacing skill sets. “Either my task needs to change or my role needs to change,” she said. “I don’t think it will make my skill set obsolete, but I may have to start applying it in a new way, which could be exciting,” she said. The company’s leadership, she added, has reinforced confidence that people remain central to operations even as automation increases.

With some documented early successes in the hopper, Novel plans to expand its use of agents to cost estimation, safety analysis and cross-project data aggregation. “AI is useful across all departments,” Carriere said.

Photo: Novel Energy Solutions