German startup Invertix has developed a suite of 22 AI “workers” designed to independently manage operations and maintenance (O&M) functions for solar power plant operators.
“Unlike general purpose agents like Open Claw,” CEO and co-founder Joseph Perrotta shared pv magazine“Invertix agents are highly specialized in the field of solar and renewable energy, understand inverter fault patterns, Modbus registers and act autonomously within operational limits. Critical decisions are escalated to people, similar to an experienced technician who works independently but provides advice on important issues.”
Perrotta and co-founder Kaan Durmaz say their experience with large language models and autonomous AI agents has shown that these systems can handle complex operational processes. They conducted more than 500 discovery calls with asset managers in Milan, O&M executives in southern Spain, energy traders in Scandinavia and CEOs of independent power producers (IPPs) in Germany. They also analyzed SCADA systems, maintenance schedules, regulatory reports, and network operations to identify tasks that cannot be scaled efficiently.
Perrotta said Invertix’s customers range from large utilities to specialty O&M providers, asset managers and project developers, all of which face similar operational bottlenecks. The company focused on three segments with acute challenges, namely IPPs and asset managers managing portfolios ranging from hundreds of megawatts to gigawatts of solar, wind and storage assets.
Register now and secure your spot.
Get to know Invertix during the start-up pitches at the Battery Business & Development Forum, co-organized by pv magazine, on March 31 in Frankfurt.
After the first successful and fully booked BBDF last July the event takes place for the second time in Frankfurt on March 31 and April 1, 2026. On the first day, discussions will cover topics such as how to attract equity and debt capitalwho takes on what risks, and how the grid connection situation is expected to develop. The second day will include the virtual BESS laba discussion about it co-location business modelsand beyond quality in storage planning and execution. On the first day, before the networking dinnerincluding start-ups Invertix, re-twin, reLi, Module Energy, Re-Twin and Scale Energy present their pitches and are available for discussion.
You can find more information and the registration link here
“While these companies are growing faster than ever, the underlying processes remain largely unchanged, relying on manual workflows, siled data and fragmented tools,” said Perrotta. “Teams are not too small; processes are simply too slow, which limits growth.”
The company said the discovery calls revealed several recurring challenges.
For example, modern SCADA systems generate thousands of data points per minute per installation. Combined with weather, market, grid, contract and regulatory data on disconnected systems, this creates significant complexity. Existing tools are often unable to process this volume effectively, leaving potential insights underutilized.
The analysis also found that daily workloads are dominated by manual, low-value tasks, including exporting SCADA data to spreadsheets, compiling performance reports, classifying alarms, completing regulatory forms, checking invoices and reviewing power purchase agreement (PPA) clauses. In some cases, staff spend hours each morning manually classifying hundreds of SCADA alarms, delaying corrective action.
Furthermore, the calls showed that PV plant operators also face constraints in expanding their portfolios. As companies strive to deploy new projects, existing workloads limit capacity. The problem is not the size of the workforce, but the processes that scale linearly with asset growth.
In addition, regulatory complexity and national reporting requirements were found to place further pressure on teams, diverting resources to compliance and increasing the risk of errors.
Perrotta said existing tools don’t fully address these issues. For example, SCADA and monitoring platforms mainly visualize data without taking action, while asset management software still requires manual interpretation. Generic AI tools lack domain-specific capabilities, such as handling Modbus registers, OPC-UA configurations, or distinguishing between PPA-related curtailments and grid events. Consulting projects, he added, typically produce reports rather than autonomous solutions.
“We have built a full team of 22 specialist AI staff across seven core departments, allowing portfolio managers to oversee significantly greater capabilities than would otherwise be possible. In technical asset management, for example, three AI staff handle SCADA monitoring, alarm and event analysis and network analysis,” Perrotta points out.
According to the company, the SCADA monitoring agent processes thousands of data points per minute, while the alarm and event analysis agent classifies alarms in real time. A network analysis agent correlates network frequency with installation performance to identify network problems before outages occur. Each AI employee contributes data that continuously improves the models.
According to Perrotta, costs range from $2,000 to $4,000 per month per AI employee, plus department costs. “Invertix is responsible for software functionality, while customers retain final decision-making authority,” he said. “Actions follow a three-phase model – inform, recommend, act – requiring human intervention for high-risk decisions.”
Perrotta added that conventional monitoring systems rely on static thresholds, which often cause false alarms. In contrast, Invertix’s AI staff evaluate alarms using SCADA telemetry, historical patterns, weather data, grid conditions and performance from neighboring plants, prioritizing relevant issues and providing contextual analysis.
He also acknowledged that the proposed approach has limitations. AI workers cannot perform physical tasks such as replacing inverters or conducting on-site inspections, nor can they manage strategic negotiations or investor relations. “The goal is to free people from repetitive work and focus on higher value activities,” he said.
Although Invertix was officially founded in February 2026, the company said it started working with live customers within three weeks. Pilot projects include a major O&M supplier in Italy managing 1.5 GW of solar, and an Italian IPP managing 350 MW. The AI workers are used for SCADA monitoring, alarm classification, portfolio performance analysis, automated reporting and identifying optimization opportunities.
The company plans to invest new capital in product development, expand its portfolio of AI workers and further train models using customer data. It focuses on the DACH region and Southern Europe, citing the increasing demand for automation in renewable energy activities.
Invertix also invites renewable energy companies to submit SCADA data from underperforming assets for free analysis. The company said it aims to identify inefficiencies and gather industry insights to guide future developments.
This content is copyrighted and may not be reused. If you would like to collaborate with us and reuse some of our content, please contact: editors@pv-magazine.com.
Popular content
