Two European AI companies are taking sharply different approaches to automating utility-scale solar operations, diverging on where the boundary between human decision-making and machine execution should sit.
As solar portfolios grow through acquisition, the manual inspection, reporting, and fault-response model that worked at 200 MW becomes increasingly difficult to sustain at scale.
That pressure is now producing two distinct models – stopping at the analytical layer in one case and extending into field operations in the other.
Two models
Invertix, a Germany-based startup that closed a pre-seed funding round in May 2026, has deployed specialised AI agents across utility-scale solar portfolios in Italy and is now expanding into Iberia. All agents operate with human sign-off before any action is taken. Asked where it currently defines the operational boundary, CEO and co-founder Joseph Perrotta was direct: “Detection, analysis, and drafting = agent. Decision and execution = human. This boundary is explicit in our architecture and our contracts.”
The platform’s agents handle monitoring, reporting, document review, and maintenance prioritization. Perrotta said reporting and root cause analysis work that previously took weeks is now delivered as draft outputs for same-day review. “The agents don’t replace people,” Perrotta said. “They make each person dramatically more capable.”
On accountability, Perrotta said every agent output carries full data lineage – sources read, logic applied, confidence level – and that human sign-off before any action is contractually embedded. “Our contracts explicitly state that agents are support tools and decision responsibility stays with the operator. This isn’t a limitation – it’s what makes enterprise adoption possible,” he said.
Areg.AI, which secured pre-seed funding in October 2025 and currently manages about 100 MW of third-party solar assets in Armenia, is building toward a different architecture.
The platform is designed to close the loop from anomaly detection through work-order creation to operator-approved delegation to field execution, including a robotics layer of unmanned ground and aerial vehicles. An operator can configure the platform to run from recommendation-only up to operator-authorised execution, with the level of automation set by the customer and the site, the company said.
It draws a specific distinction from competitors.
“In competing platforms, tasks route to human technicians and stop there. In Areg.AI, the receiving entity can be a human team or a robotics-enabled field workflow, coordinated across a fleet of purpose-built UGV and UAV systems,” the company told pv magazine.
Areg.AI’s AI agent, ARPI, is partially live, deployed across monitoring and maintenance modules, with rollout across the remaining sections ongoing. The company said more than half of on-site staff have adopted ARPI as their primary diagnostic reference.
Performance claims
Areg.AI’s most concrete performance data comes from the New Energy Group portfolio in Armenia, managed by Energy Service LLC, which Areg.AI later acquired after validating the platform’s results on its assets. The baseline average generation across 2023 and 2024 was 113.19 GWh.
The company said it recorded a 9.72% weather-normalised generation increase in 2025, of which 7.26 percentage points were attributable to soiling detection and targeted cleaning alone, with a further 3.70% projected for 2026. Energy Service LLC can verify the figures independently, the company said. Results vary by site, asset condition, and operating period.
Before deployment, cleaning operations on the portfolio ran on a fixed calendar-based schedule applied uniformly across the plant, with no account taken of localised soiling conditions or actual energy loss impact. Preventive maintenance was similarly time-interval driven. Areg.AI replaced both with drone-based aerial assessment at the zone level, quantifying energy loss per area and routing cleaning tasks accordingly, with maintenance now condition-based and triggered by AI-detected degradation signals.
Areg.AI is preparing a significant expansion. An acquisition agreement for a Spanish portfolio of approximately 100 MW is scheduled to be signed at Intersolar in June, with an additional 200 MW of O&M contracts to follow. The commercial model is fully managed O&M – platform integration, robotics-enabled ground operations, and aerial thermal inspection delivered as a unified service rather than a standalone software license.
Perrotta framed the sector’s trajectory in terms of productivity rather than autonomy.
“Over the next two to five years, we expect this to shift toward delegated-authority workflows where agents act within defined bounds. But we think the industry needs to build trust first – and that happens by running agents in production, proving accuracy, and progressively expanding scope,” he said. “The companies that will win the next decade of renewable operations are the ones that figure out how to run 10 GW with the team they built for 200 MW. That’s what we build.”
Other platforms active in adjacent segments of the market include SenseHawk and Envision Digital, which focus on monitoring, analytics, and workflow rather than extending automation into field operations.
From pv magazine Global
Comments
Please login to comment
Wednesday, June 10, 2026
3:00 pm – 4:00 pm CEST, Berlin, Paris, Madrid
Tuesday, June 9, 2026
11:00 am – 12:00 pm CEST, Berlin, Paris, Madrid
Thursday, June 11, 2026
5:00 pm – 6:00 pm CEST, Berlin, Paris, Madrid
Monday, June 1, 2026
5:30 pm – 6:30 pm CEST, Berlin, Madrid, Paris
Tuesday, June 16, 2026
6 am – 7:00 am CEST, Berlin
Friday, June 12, 2026
2:00 pm – 3:00 pm CEST, Berlin, Paris, Madrid
The new pv magazine Global May issue is now available!
Mountains to climb
Available in print and digital formats.
Entries open in seven categories: Modules, Inverters, BoS, BESS, Manufacturing, Sustainability, Projects.
April 01 – August 31, 2026
Energy-hungry data centers open new doors for solar and storage.
Available in print and digital formats.
You have no items in your basket.
