Best AI Consultants for Energy-Intensive Industry and Utilities in 2027

If you run an energy-intensive plant or a utility in the Netherlands, 2027 is the year the EED energy management duty lands, and the right AI partner depends on the job you need done. Our pick for building the steering layer of an industrial energy management system is Crux Digits. For certification, flexibility revenue, trading forecasts and grid-operator analytics, the specialists below fit better.

Disclosure: Crux Digits publishes this list and is listed first. We say why, and who the others are better for. The guide is written for the IT manager, informatiemanager or CIO at a site of 250 to 5,000 FTE who has to bring an AI supplier into a landscape that already holds an ICT partner, an OT integrator and an energy supplier.

Why is 2027 a deadline year for energy-intensive sites?

EED Article 11 requires a certified energy management system by 11 October 2027 for any enterprise using more than 85 TJ a year. The source is the Energy Efficiency Directive (EU) 2023/1791. In the Netherlands, RVO sets out what that means: proof of the system, or of an energy performance contract that meets Annex XV, plus a list of cost-effective measures, is due by 11 October 2027. The system must be certified by an accredited body and must cover the whole company, not one site.

Below that line, the audit duty starts at 10 TJ a year. RVO expects the amended Dutch legislation to apply after the first quarter of 2027, and until then the current EED audit duty applies. A company that becomes audit-obliged under the new threshold has until 1 April 2028 for its first audit; a company already under the old rule submits a new report within four years of its previous one. RVO checked that page on 5 October 2026 and notes that everything is subject to the final regulation.

To make the thresholds concrete: 85 TJ is about 23.6 GWh and 10 TJ about 2.8 GWh, by straight unit conversion. Gas, heat and fuels count towards the same total, so a site that buys mostly gas reaches the line on a different electricity figure.

The national energiebesparingsplicht runs alongside it. RVO lists an onderzoeksplicht for certain sectors from 10 million kWh of electricity or 170,000 m3 of natural gas equivalent per location per year, and an informatieplicht reported once every four years.

Grid congestion adds a second calendar. The ACM code decision on alternative transport rights introduced the tijdsduurgebonden transportrecht, which guarantees transport for at least 85% of the hours in a year on TenneT's high-voltage grid and which TenneT must offer from 1 October 2025, and the tijdsblokgebonden transportrecht on regional grids, which grants transport within time blocks agreed with the grid operator. Both carry a lower network tariff in exchange for that adjusted use. Either contract turns your connection into a schedule that something has to keep, which is where software comes in. Our hub on grid congestion for industry covers the contract options in detail.

Which job are you actually hiring an AI firm for?

There are five distinct jobs on an energy-intensive site in 2027, and no firm on this list does all five well.

• Certify: the EED energy management system or energy audit. This needs an accredited certification body, not a software house.
• Steer: the control layer that keeps load inside the connection value and the transport right, using batteries, flexible processes and on-site generation. This needs a builder that can integrate with SCADA, PLCs and meters.
• Earn: revenue from offering flexibility on balancing markets. This needs an aggregator with market access.
• Trade: forecasts of renewable output and short-term power prices. This needs a forecasting product.
• Analyse at network scale: asset risk, capacity and investment questions at a grid operator or utility. This needs a data consultancy with sector depth.

A sixth job sits outside AI altogether: the physical capacity, meaning the transformer, the battery and the metering. The decision rule we use is simple. Write the job down in one sentence before you send a request for information. If the sentence contains "certify", shortlist accredited bodies. If it contains "steer", shortlist firms that can show you how their software writes setpoints through your existing control system, with your OT integrator in the loop.

How did we choose the firms on this list?

We applied four criteria, verified every firm on its own website on 29 October 2026, and list the others alphabetically after our own pick.

• It serves this segment: the firm's own site shows work for energy-intensive industry, energy companies or grid operators, not generic work relabelled.
• Published, checkable specifics: named products, services or cases on its own pages. We quote prices only where a firm publishes them.
• Reachable and current: a live site on the verification date. Two energy software firms we wanted to include were left out because their sites would not load for verification.
• Honest fit: for each firm we name the buyer for whom it is the better choice, including where that is not us.

The list is not a ranking beyond our own pick. There are no paid placements, affiliate links or reciprocal listings. We describe each firm from its own public positioning and do not repeat ratings or review counts.

Why is Crux Digits our pick for 2027?

Crux Digits is our pick because it builds the steering layer energy-intensive sites need, at fixed published prices, and hands you the code.

Disclosure: Crux Digits publishes this list and is listed first. We say why, and who the others are better for.

What it does: an AI layer on top of the SCADA, PLCs, energy meters and ERP you already run. It forecasts demand and price, shifts flexible load, charges and discharges a battery ahead of a forecast peak, and keeps consumption inside the connection value. Your control technology, safety logic and operators stay in place.

• Fixed prices, excluding VAT, published on the pricing page: AI Audit & Strategy €2,500 in 1–2 weeks, Production-ready MVP €20,000 in 4–6 weeks, Production Launch from €50,000 in 8–12 weeks. The audit fee is credited in full if the MVP follows within 60 days.
• Code, model, prompts and test set are transferred to you at handover.
• It works alongside your ICT partner and your OT integrator. Crux Digits does not provide ICT beheer, hosting, helpdesk or 24/7 support, and does not pretend to.

• Bilingual delivery in English and Dutch, from Utrecht.
• A published predictive maintenance case for a sugar producer, a continuous-process plant; the client name is withheld on that page.

Best for: an energy-intensive site of 250 to 5,000 FTE that wants a working steering layer on its own meter and process data within weeks, owned by the site. Not for: certifying your energy management system, supplying a battery or trading your flexibility. The firms below do those jobs.

Which other firms belong on a 2027 shortlist?

Seven firms do specific jobs better than we do; they are listed alphabetically, each verified on its own website.

Dexter Energy

Dexter, based in Amsterdam, sells forecasts and trading signals for short-term power markets, built for automated asset-backed energy traders, with weather-based forecasting, solar nowcasting and an API for renewable portfolios. Best for: a utility or energy company trading a solar, wind or battery portfolio that needs forecasts and bidding input, not a factory control layer.

DNV

DNV describes itself as an accredited certification body for ISO 50001 and offers gap assessments against the standard, implementation support and training, from a one-day awareness course to a one-week lead auditor course. Best for: a company above 85 TJ that needs its energy management system certified before 11 October 2027. Crux Digits cannot certify anything, and a steering layer does not replace that certification.

Kenter Groendus

Kenter Groendus offers business customers everything under one roof, from groundwork and hardware to software for managing energy and price: energy management systems, batteries, transformer stations, metering as a recognised metering company, and service and maintenance. It states more than 36,000 customers, more than 850 staff and eight locations in the Netherlands. Best for: a site that needs physical capacity, such as a battery or a transformer, delivered and serviced by one supplier.

Sia Partners

Sia Partners opened its Amsterdam office in 2009. Its Energy & Industry team works for grid operators, suppliers and traders, government and industry, on questions such as battery deployment studies for grid operators and congestion management, and the Netherlands hosts one of the firm's four global Data Science Centres of Excellence. Best for: a grid operator or utility with a strategy or market question that needs management consultants who also do data science.

Spectral

Spectral, based in Amsterdam, builds energy software: STELLAR, a virtual power plant platform for monitoring, controlling and trading solar, wind and battery assets, and BRIGHTER for energy use in commercial buildings. It lists energy traders, utilities, independent power producers, asset managers, property managers and installers as customers, and states ISO/IEC 27001:2022 certification. Best for: a utility or producer that wants a VPP platform off the shelf rather than a custom build.

Sympower

Sympower calls itself an independent flexibility services provider. For industrial companies it identifies, connects and monetises flexibility from equipment such as furnaces, compressors, dryers and pumps, and customers are paid for the capacity they provide or for the time their equipment is activated. Its customers include chemical, cement, metallurgy, pulp and paper and data centre businesses. Best for: a plant that wants flexibility revenue from balancing markets without building or owning software.

Xomnia

Xomnia is an Amsterdam data and AI consultancy offering data and AI strategy, cloud data platforms and data products. Its published case for grid operator Enexis built a risk-scoring model for excavation work near underground cables, with a data pipeline, MLflow and Airflow, and weekly retraining with the prevention team. Best for: a grid operator or utility building a data platform and an analytics capability for asset risk.

What should you ask each firm before you sign?

Ask six questions, and treat a firm that cannot answer the first two in writing as a pilot supplier, not a production supplier.

1. Which systems does your software read from and write to: SCADA, the historian, PLCs, the quarter-hour data from your metering company, the ERP? Who changes PLC logic, you or our OT integrator?
2. What happens on a communication loss? The answer should be that the plant falls back to its own control without operator action.
3. Who owns the code, the model and the training data after handover, and in which repository do they live?
4. How does the output feed our EED evidence, such as the list of cost-effective measures or the energy audit report?
5. Which transport right do you plan against: a firm connection, a tijdsduurgebonden or a tijdsblokgebonden right?
6. Is any component a safety component under the AI Act, and where is that reasoning documented?

The fourth question separates the jobs most clearly. A certification body answers it with an audit plan, an aggregator with a market contract, and a builder with a data field list. None of those answers is wrong, but only one of them matches the job you wrote down.

Does AI in an energy system count as high-risk under the EU AI Act?

Only when it is a safety component in managing or operating the supply of water, gas, heating or electricity (AI Act Annex III, point 2).

Article 3(14) of the AI Act defines a safety component as one that fulfils a safety function, or whose failure or malfunction endangers the health and safety of persons or property. The Digital Omnibus on AI, Regulation (EU) 2026/1744, moved the obligations for Annex III systems to 2 December 2027, so the classification question and the EED deadline now fall in the same quarter.

For a factory load-shifting optimiser the practical design rule is to keep protection and interlocks in the PLC safety logic, let the AI layer propose and write setpoints within limits that logic enforces, and record that reasoning before go-live. A grid operator whose system steers the network itself faces a harder classification question and should take legal advice. In both cases the reasoning belongs in the documentation, not in a sales deck.

Where does Crux Digits fit, and where does it not?

Crux Digits fits the steering and software job: an AI layer on your meters and SCADA, built in fixed steps, owned by you.

It does not certify energy management systems, sell batteries or transformers, aggregate flexibility, or run your IT. On a typical 2027 project we would expect to sit next to a certification body for the EED evidence, an OT integrator for the control system, and your ICT partner for the infrastructure the software runs on. If your job is one of the others on this page, hire that specialist first and talk to us when you need the steering layer.

For the wider national field, our buyer's guide to the best AI consultants in the Netherlands covers firms beyond energy.

For the build itself, see our page on the energy management system for industry.

Last updated: 29 October 2026.

Frequently asked questions