A First Look at Invertix: Betting on AI Workers in Energy

AI workers for operators, or another AI tool that writes reports?

Invertix prices its agents like a hire, not a software seat. That only holds if they actually do an analyst’s job; if they mostly help analysts work faster, Invertix is selling into a much smaller software budget.

Business logic

Invertix sells software agents, which it calls “AI workers”, to solar, wind and storage operators. The agents handle monitoring, alarm triage, reporting and maintenance coordination, trained on each operator’s own SCADA and work-order data. The buyers are asset owners, IPPs and O&M providers.

What changes for the operator:

	Today	With Invertix
Who does the work	Analysts manually triage alarms and write reports	AI workers run monitoring, alarm triage, reporting and maintenance coordination. Public sources don't show which tasks run without human review.
Monthly cost	€5,000–6,500 per fully loaded analyst	€2,000–4,000 per AI worker, plus department fees
Limits	Portfolios grow faster than operators can hire and train staff to run them	Models are trained on each operator's own SCADA and work-order data, so they improve with each deployment
Benefit	—	Faults caught earlier should reduce lost output; not quantified in public sources

Evidence scorecard

[V] = verified by a third party, [C] = company claim, [A] = my assumption.

Dimension	Rating	Evidence
Pain & budget	★★★½☆	[C] Operators cannot hire enough people for this work: the industry grows about 30% a year, but teams don't grow with it. [C] AI worker at €2,000–4,000 a month vs. €5,000–6,500 for a fully loaded analyst. Note: the cost comparison assumes one AI worker replaces one analyst, which is not verified.
Demand evidence	★★★★☆	[V] TotalEnergies On named as a customer, plus four European pilots (Trending Topics). [V] Pilots with a 1.5 GW Italian O&M provider and a 350 MW IPP (pv magazine). [C] About €200K ARR three months after founding; about 2 GW under management by September. Which pilots are paid is not disclosed.
Team–problem fit	★★★☆☆	Strength – founder-led sales. [C] Cold-contacted thousands of people on LinkedIn and WhatsApp (2,000+ per the founders, about 5,000 per an investor) and won the first customers themselves. Strength – technical depth. [V] Kaan Durmaz studied electrical engineering and applied math/AI safety, worked on autonomous driving at BMW and received the Rel-ai grant (investor write-up). The two met at TUM.ai, the Technical University of Munich's student AI group. Gap – no energy background. Neither founder has worked in the energy industry, and public sources show no US market experience, though the Munich-based team of eight is now seeking US customers. Joseph Perrotta (CEO) built consumer apps before Invertix.
Market size	★★★½☆	[A] Rough ceiling at today's pricing: about €100K of revenue per GW managed (€200K ARR on about 2 GW, both company figures) × 652 GW of EU solar and wind ≈ €65M a year; × 3,682 GW globally ≈ €370M. Excludes storage, and pilot-stage pricing likely understates revenue per GW. [A] Solar and wind are likely the entry point rather than the whole market: the company describes itself as "AI that operates critical infrastructure" (YC), so storage, grids and other assets could extend it.

Capital efficiency: [V] €1.7M pre-seed led by Vireo Ventures (Munich Startup) and a team of eight, against about 2 GW under management.

Where this business could fail

Failure mode	Why it could happen
Stuck as a services business	Each deployment needs engineers inside the client, and eight people serve customers across asset types and countries. Public sources don't disclose how long a deployment takes.
Bundled away by incumbents	Monitoring vendors and O&M providers already hold operators' SCADA access and contracts, and could add similar agents to what customers already pay for.
Budget doesn't move	If operators treat AI workers as an add-on rather than a replacement for analyst hires, Invertix competes for a much smaller software budget. There is no public evidence yet of a customer cutting or not filling an analyst role.

Three questions I'd ask the founders

Each question tests the opening question or a failure mode above.

Budget: Which tasks do your AI workers complete today without a human reviewing them, and has any customer cut, or decided not to fill, an analyst role because of Invertix?
Pricing: Why do customers pay per AI worker, rather than for the work delivered or per site or MW covered, and how does a customer decide how many workers it needs?
Services: How many engineer-weeks did your last three deployments take, and how much of each was custom work?

Sources

All information comes from public sources as of October 2026.

About this series

This series is how I look at early-stage companies: identify the question behind the business and keep testing it as new evidence comes in.

I screened all 142 companies in YC's Fall 2026 batch, keeping only those that replace spending a buyer already has, show demand someone outside the team can verify, and have founders with earned insight into the problem. Invertix is one of them.

If you'd like the full spreadsheet, or want to trade notes on any of these companies, reach me on LinkedIn or at jz3938@columbia.edu.