Product

Agents do the work. Your team makes the calls.

Secure AI agents for solar O&M and asset management teams. They triage every alarm, bring in the right technician and check the fix.

WO-2291Plant 1INV-04Fix verified

  1. Monitoring agentInsulation fault found on INV-04.

  2. Dispatch agentMarco Rossi booked. Giulia Conti approved.

  3. Marco Rossi, technicianDamaged DC cable replaced on site.

  4. Monitoring agentFix checked in FusionSolar. Work order closed.

Illustrative example on a demo fleet.
50+ GW

managed by the energy companies we work with.

01The agents

One agent for each part of the job.

Start with three: the Monitoring, Dispatch and Asset management agents. Our engineers build more with you, job by job.

Each agent takes on one part of the work and hands the rest to another agent or to your people. Together they triage alarms, find the cause, bring in the right technician, check the fix and prepare the reports that follow.

Demo fleet. Figures are illustrative.See it for your role →
  1. Monitoring agentEvery alarm triaged.

    Clears the noise and raises the few alarms that need a person.

  2. Dispatch agentThe right technician, booked.

    Can propose the nearest free qualified technician and, once approved, book the visit.

  3. Asset management agentEvery asset, accounted for.

    Tracks every asset against its P50 budget and prepares the monthly fleet report.

In detail about the agents

Monitoring agent. Reads your monitoring portals, such as FusionSolar, the way your operators do. When an alarm needs a person, it posts the cause with the evidence: the fault, the reading and the energy at stake. After the repair, it checks the plant in the portal before the work order closes.

Dispatch agent. Can propose the technician in Microsoft Teams and, once your team approves, book the visit and open the work order with the evidence. The technician confirms in the thread.

Asset management agent. Logs each fault against the asset’s warranty, with the evidence attached, and keeps the record behind the monthly fleet report.

One alarm, end to end. In the example at the top of this page, INV-04 trips at 11:45. The Monitoring agent posts it a minute later, a technician is booked by 11:48, and the fix is checked in FusionSolar at 13:52. See it in Teams →

02For your role

Four desks, one set of agents.

O&M, leadership, data teams and asset management each see the work their own way.

Each short film follows one job on a demo fleet, from the first signal to the result that desk sees.

The O&M film, on a demo fleet: 38 alarms come in, the agents find the two that need a person, the nearest free technician is booked in three minutes, and the fix is verified in FusionSolar before the work order closes.

The leadership film, on a demo fleet: asked what lost production cost in Q3 and why, three agents check the portals, work orders and PPAs, and answer in euros, with what is recoverable.

The data teams film, on a demo fleet: a question to the fleet from the terminal, an app built on the same agents that alerts the O&M team in Microsoft Teams, and the same answer from the command line, the Python SDK and the REST API.

The asset management film, on a demo fleet: every asset is tracked against its P50 budget, a warranty claim is drafted with the evidence attached, and the monthly fleet report is prepared.

Demo fleet. Figures are illustrative.

Every alarm triaged. Every fix verified.

In detail about each role

O&M. On the demo fleet, 38 alarms come in and the Monitoring agent finds the two that need a person. The Dispatch agent books the nearest free technician in three minutes, and the work order closes once the fix is checked in FusionSolar.

Leadership. Ask in plain words, such as “What did lost production cost us in Q3, and why?” The agents check your portals, work orders and PPAs, and answer in euros, with what is recoverable.

Data teams. Ask the fleet from your terminal, build apps on the same agents, then ship them to your team or call the API.

Asset management. Every asset is tracked against its P50 budget, and the monthly fleet report is prepared for your review. Faults are logged against warranties, and claims can be drafted with the evidence attached.

In preview with partners. The terminal, API and app builder for data teams, and warranty-claim drafting.

03Your systems

Your systems stay. The agents join them.

No rip-and-replace. The agents work in your monitoring portals, Microsoft Teams and your work-order system.

Invertix connects to each system, such as FusionSolar and Microsoft Teams, and brings their records into one operational model: a single, shared record of your plants, alarms, work orders and people.

Illustrative example on a demo fleet. In Microsoft Teams, the Monitoring agent posts an inverter fault a minute after it trips, with the evidence. The Dispatch agent proposes a technician, the head of O&M approves, the technician confirms, the Asset management agent logs the fault to the inverter’s fault history, and after the repair the Monitoring agent confirms in FusionSolar that the inverter is back online and the work order is closed.
In detail about your systems

Monitoring portals. The agents sign in to your portals, such as FusionSolar, with logins kept in an encrypted secrets vault. The logins are never visible to the models or to Invertix staff.

Microsoft Teams. The agents post in your team’s channels, answer questions in the thread and ask for approval with a button, as in the example here.

Work orders. The Dispatch agent can open work orders in your work-order system with the evidence attached. Each one closes once the fix is checked.

Connecting a system. Our engineers connect each system with your team, as part of the deployment. How we deploy →

04Your team in control

Set the rules for every workflow.

For each workflow, choose: the agent works fully autonomously, or a person verifies each action.

Each workflow also defines the tools an agent can use and when it brings in a person.

Example rules: which workflows an agent runs on its own and which a person verifies
WorkflowFully autonomousA person verifies
Alarm triageYesNo
Technician bookingNoYes
Work ordersYesNo
Monthly fleet reportNoYes
Plant controlsOnly with your approval
Example. Your rules are agreed with your team.
In detail about control

Approvals where you work. When a person is needed, the agent asks in Microsoft Teams, with the evidence attached.

Plant controls. Whatever a workflow’s setting, agents write to plant controls only with your approval.

Action log. Every agent action is logged, so your team can see what each agent did and when.

Access. Your people sign in with single sign-on (SSO). Roles and permissions decide who sees and does what.

05How we deploy

Built with your team, one job at a time.

You don’t get software and a login. Our forward-deployed engineers work inside your operations team.

Your people bring the operating knowledge; we bring the data and AI engineering.

  1. 01Embed

    Your team

    Shows us how the work is really done.

    Our engineers

    Join your operations team.

  2. 02Map

    Your team

    Chooses one recurring job to start with.

    Our engineers

    Map it end to end: the alarm, the systems, the people and the decisions.

  3. 03Build

    Your team

    Sets the approval rules, workflow by workflow.

    Our engineers

    Connect your systems and build the agents for that job.

  4. 04Scale

    Your team

    Sees results on one portfolio, then picks the next plants and jobs.

    Our engineers

    Run and improve the agents alongside you.

Who does what, phase by phase.
In detail about deployment

Forward-deployed engineers. The engineers who build your agents sit with your operations team, so the agents follow how your people really work.

Where it starts. One recurring job, mapped end to end before anything is built. More plants and more jobs follow at your pace.

Your rules first. The approval rules are part of the build, set workflow by workflow before an agent acts on its own. Your team in control →

06Research & security

Frontier models, behind our harness.

Your agents use frontier models from OpenAI, Anthropic and Mistral. Every request first passes through our harness: the layer that checks it before it reaches a lab.

The harness keeps your data apart from every other company’s, replaces site names and people’s names with placeholders, and sends only the data you agreed to the models you agreed. Your data is encrypted in transit and at rest.

Research · Transactions on Machine Learning Research, 2026

Amplified Patch-Level Differential Privacy for Free via Random Cropping

Kaan Durmaz, Jan Schuchardt, Sebastian Schmidt, Stephan Günnemann

Peer-reviewed. Our co-founder and CTO, Kaan Durmaz, is first author. Work carried out at the Technical University of Munich.

In detail about research and security

Differential privacy. It runs in our product, alongside the harness. Our engineers walk your security team through where it applies in your deployment.

Sandboxes. Agents run in virtual-machine sandboxes.

No training on your data. The labs’ terms forbid training on your data.

Hosting and deletion. Hosted on EU cloud by default. Invertix is GDPR compliant, and your data is deleted on request.

Questions.

What do Invertix agents do?

Invertix agents take on the operational work of solar O&M and asset management teams: they triage alarms, find the cause, bring in the right technician, check the fix and prepare the reports that follow. You start with three: the Monitoring, Dispatch and Asset management agents.

Which systems do the agents work in?

Your monitoring portals, such as FusionSolar, Microsoft Teams and your work-order system. Our engineers connect each system with your team as part of the deployment.

Do we need to replace our monitoring platform or SCADA?

No. Invertix connects to the systems you already use and brings their records into one operational model. Your monitoring platform, SCADA and work-order system stay as they are.

Can an agent act without a person?

You decide, workflow by workflow: an agent works fully autonomously, or a person verifies each action. Plant controls change only with your approval, and every agent action is logged.

Who deploys the agents?

Invertix’s own forward-deployed engineers. They work inside your operations team, start with one recurring job and add plants and jobs at your pace.

Invertix at a glance
Invertix at a glance
What Invertix isA company that builds and deploys AI agents for energy operations
What the agents doTriage alarms, bring in the right technician, check the fix, track assets and prepare fleet reports
AgentsMonitoring agent, Dispatch agent and Asset management agent; our engineers build more with you, job by job
Where the agents workYour monitoring portals (such as FusionSolar), Microsoft Teams and your work-order system
Who uses themSolar O&M, asset management, leadership and data teams at energy companies
Works withAltenia, Terna and TotalEnergies On
Scale50+ GW managed by the energy companies we work with: their combined portfolio, not capacity operated by Invertix
DeploymentBy Invertix’s own forward-deployed engineers, working inside your operations team
Human controlSet per workflow: fully autonomous, or a person verifies each action. Plant controls change only with your approval.
AI modelsOpenAI, Anthropic and Mistral, always through the Invertix harness
Data protectionPer-company isolation in VM sandboxes; site names and people’s names replaced by placeholders; encrypted in transit and at rest; EU cloud by default; GDPR compliant; deleted on request
AccessSingle sign-on, roles and permissions, an encrypted secrets vault for portal logins, every agent action logged
ResearchPeer-reviewed paper on differential privacy, TMLR 2026; first author Kaan Durmaz, co-founder and CTO
FoundersJoseph Perrotta (CEO) and Kaan Durmaz (CTO)
HeadquartersMunich, Germany
Legal entitiesInvertix GmbH (Munich, HRB 311292); Invertix S.r.l. (Viterbo, Italy)
Backed byY Combinator, Vireo Ventures, Italian Founders Fund

Bring us one recent alarm.

Our engineers will walk you through how an agent would handle it, step by step, with your team in control.

Europe +34 606 62 83 84 · US +1 (415) 718-7732