A Slovenian business does not pay for the electricity network mainly by what it consumes. It pays for the capacity it books — a figure agreed per metering point, separately for each of five time blocks, and different between the high and the low season. Book too little and the excess is penalised. Book too much and the company pays every month for capacity it never uses. The figure can be changed monthly, which means it can also be got right monthly.
opten.energy was built on exactly that. They take a client's own 15-minute metering data, work out what the client will really need, and book it. For a single metering point that is an afternoon of careful work. The Opten.energy optimization platform today runs for clients that together account for around 15% of the Slovenian electricity market, and the deadline comes round again every month.
So the business only exists if the work is automated end to end. That is what we were asked to build.
"We optimise energy costs in the B2B segment."
Nejc Frumen, Co-Founder, opten.energy, describing the company in one sentence
The challenge: the same careful job, every month, for every meter
The saving on any one metering point in any one month might not be dramatic. It adds up only if the whole chain — data in, forecast, optimisation, submission to the distributor — runs without a person in the middle of it, and runs the same way every month for every client.
-
Forecasting Forecasting
Demand predicted per metering point from its own 15-minute history, with the seasonality, the working calendar and the shift patterns a plant actually runs.
-
Optimisation Optimisation
The booked capacity for five blocks chosen against the tariff's own penalty rules — the cheapest set, not the most cautious one.
-
Integration Integration
Historical consumption ingested automatically, optimised capacities filed back to the distributor, monthly, per metering point, with nobody retyping numbers.
-
Auditability Auditability
Every figure a client is billed a saving on has to be traceable back to the data it came from.
The solution: a deterministic platform, and a Claude-based contract interpretation layer
Where Claude is used
- In production: Claude Opus reads anonymised supply contracts and writes each contract's valuation. In daily use since June 2026.
- In engineering: we built the contract product with Claude Code. The 2025 network-charge platform was built before Claude and is now maintained with it.
The platform, 2025
An AI time-series forecasting engine on the client's own consumption history. An optimiser that turns each forecast into the capacity to book across the five blocks. And secure API connectors in both directions: historical data ingested automatically, optimised capacities submitted automatically to the distributor each month with only flagged metering points manually checked. Ten of our engineers worked on it between March and November 2025. It has run the business since, and is now maintained in our Claude Code-supported engineering environment.
What the automation removed was not only time. A monthly submission executed by hand, across thousands of metering points, is a steady trickle of small errors that each cost a client money and OPTEN credibility. There is no typing or manual calculation.
Reading the supply contracts, 2026
With the network side automated, the other half of a client's bill became the bottleneck: the supply contract itself. These are not standard documents. Fixed prices and indexed prices, block products, quantity tolerances, pass-through items, price-review triggers, notice periods — the terms that decide what a client pays are rarely in the same place twice and might have different meanings, and reading one properly took an experienced analyst at least three hours.
We upgraded the Opten.energy optimization platform with Claude. Scanned contracts are OCR'd and anonymised on local models, so client data never leaves OPTEN's infrastructure; only the anonymised text is sent to Claude Opus. Claude extracts every commercial term that drives the cost, flags whatever is unusual about that particular contract, and writes a Python valuation tailored to it rather than picking one from a library. The platform stores the script with the contract, runs it every month against the client's own consumption profile and market prices, and reconciles the result against the actual invoice. OPTEN's experts review any mismatch. In daily production since June 2026, used for multiple OPTEN clients.
Claude reads the document and writes the valuation for it; the platform supplies the data, runs the calculation and keeps the audit trail. Every result is validated against the client's monthly invoice. We built the product itself with Claude Code, and it stays deterministic: once a contract's valuation is written, it runs as ordinary code, the same way every month.
The result
The technological backbone of the business. The software generated over €1M in revenue within 18 months of operations.
- Supply contract review: from at least 3 hours to at most 30 minutes, analyst review included, with the client-meeting summary produced in the same run. In daily production since June 2026. Signed contracts with 3 new important customers in September 2026.
- About 30% average savings in network charges for the end client, with no physical change to their installation — no CAPEX.
- Around 15% of the Slovenian electricity market. That is the share the platform optimises for, every month, without the headcount that doing it by hand would need.
- Manual errors eliminated. Automating the monthly submission removed the retyping and the class of mistakes that came with it, and cut the operational overhead of running the whole portfolio.
- Rated 5.0 out of 5.0 on quality, schedule, cost and willingness to refer, in a verified public review.
"Reading a supply contract properly used to take one of our experts at least three hours. Since June we run every contract through Claude, and it takes at most 30 minutes, review included. Claude reads an anonymized version, distils the terms and writes the valuation for that specific contract. Every month the platform checks it against the client's invoice, and our experts review anything that doesn't match. This system works great, enables us to scale and it's in daily use."
Nejc Frumen, Co-Founder, opten.energy
"The developed software serves as the technological backbone of the business model, which has successfully generated over €1M+ in revenue within 18 months of operations.
Since it was an innovative project, the specs weren't 100% defined; the team at Abelium handled this professionally and pivoted when needed. They are experts in the R&D field and extremely flexible to the needs of their customers while providing quality results with real-world impact."
Nejc Frumen, Co-Founder, opten.energy