AI and software for distribution utilities, transmission and generation companies and renewable developers, built on meter data you have already paid for.
The expensive part is already installed. The missing part is what happens to the data after it arrives.
In brief
EigenSpark builds AI systems and the software around them for distribution utilities, transmission and generation companies, and renewable developers. We take the meter data you already collect and turn it into loss attribution down to the feeder and the connection, a ranked list of where to send an inspection team, and a billing process that stops leaking. We do the same for network health, asset planning, forecasting and market scheduling.
What we cover
Loss reporting, metering rules, contract performance and market scheduling.
The starting position
Data that goes nowhere, losses measured too far up the network, and three maps that disagree.
Interval reads, tamper events and voltage signals arrive daily and land in a billing table. What is missing turns them into a work order.
A division-level number tells a field team nothing. Attribution has to reach the transformer, and it has to separate physics from theft.
GIS, billing and field reality disagree about which consumer sits on which transformer. Consumer indexing is the first job.
Metering and command success are service levels reported by the party being measured. You need your own view from the raw data.
How an engagement runs
Read the feed, prove it in one division, then make the next division cheap.
Two to three weeks inside your head-end, meter data management and billing systems to establish what is actually arriving, how complete it is, and which use cases the existing feed can carry today. Consumer indexing quality is scored in the same pass.
A single workflow taken to a working interface for one division: the analysis, the ranked worklist, the field app and the outcome recorded. Loss attribution and theft prioritisation is the usual first build, because recovery is measurable.
Ingestion, indexing, the loss model and the rule engine are shared. The second division and the second use case are configuration, and your own engineers do the configuring.
Use cases
Losses and revenue, then network and assets, then planning and markets, then the back office.
Input energy, transformer reads and consumer consumption reconciled down to the segment, with technical loss separated from theft and each segment ranked.
Data Engineering & Platforms →Tamper flags, consumption drops, load shape and neighbour comparison read into a ranked list for the vigilance team, with every visit outcome fed back.
AI & Data Strategy →Zero reads, frozen meters, negative consumption and unmapped connections surfaced as an exception list with the money attached, largest gaps first.
Continuous Compliance Monitor →Loading, unbalance, voltage excursions and failure history read together to rank the transformers most likely to fail this season, with the case prepared.
Predictive Maintenance System →Last-gasp and restore events clustered into an outage with a probable extent and cause, raised before the complaints arrive and tracked to restoration.
Agentic AI →GIS, billing and the field survey reconciled into one map of which consumer sits on which transformer, with every disagreement listed for verification.
Data Engineering & Platforms →Consumption history, weather, calendar and category mix read into a forecast by feeder and by hour, with the error measured openly.
AI & Data Strategy →Plant history and weather read into a generation forecast in the format the load despatch centre expects, with forecast error tracked against deviation cost.
AI & Data Strategy →Data availability, command success and communication failures computed from the raw feed, so you measure the service level and can evidence a deduction.
Continuous Compliance Monitor →An assistant that answers billing, tariff, connection and outage questions from your approved material and the consumer’s own account, and cites what it used.
Generative AI →Applications and complaints classified, documents checked against the requirement, and cases tracked against the timelines your regulator sets.
Document Intelligence Engine →The annual revenue requirement and true-up assembled from source systems, with every figure carrying its record and prior commission directions checked.
Generative AI →The engineering half
Four layers that decide whether a theft ranking ever reaches a person with a van.
We read what your installed base already sends. Meters, RTUs, communication and field instrumentation are a specialist trade and we work alongside whoever supplies them.
The head-end, the meter data platform and the billing system stay. Most of the effort is getting them to agree about which consumer is where.
Built for a vigilance team, a division engineer and a call centre, with the outcome of every visit recorded so the ranking gets better.
Defined once and used everywhere, so the scheme report, the commission filing and the field plan cannot disagree. Your team runs and owns it.
We also run the training that goes with it: SQL and data engineering, cloud, enterprise systems and cybersecurity, alongside the AI tracks. A team that cannot query its own data cannot specify a model either.
What we offer energy and utility teams
Most utilities start with one division and one workflow, and use three of the four.
Where a utility programme starts.
Where a single workflow starts.
AI tracks, and the foundations under them.
The teams that own this work.
Working in a specific function? See how we help Data, Analytics & IT teams.
Security and controls
No system disconnects a consumer. Every figure traces back to a meter reading you can open.
Theft rankings, billing exceptions and notices go to the officer who holds the power to act. The system evidences and ranks; a person visits, verifies and decides.
A loss number, an exception or a ranking links back to the reads and events it was built from, on the days they arrived, so a disputed case is answered from the record.
What the field team found, and what was recovered, is recorded against the case. Without that loop a ranking is a guess that never improves, which is how most of these systems end up ignored.
Loss thresholds, tamper rules and forecast parameters are managed in a console by your own engineers. A new division or a changed tariff category does not need a release.
FAQs
No. Metering, communication networks and field instrumentation are a specialist trade with specialist vendors, and in most utilities they are already contracted. We build the software, the integration and the analytics on top of what those systems send, and we work alongside whoever supplies them.
No. Partial coverage is enough to start, and starting now is what tells you whether the data arriving is fit for anything. Loss attribution works wherever the transformer meter and the consumers under it are both metered, so it can be proved on one division while the rollout continues elsewhere. It also surfaces data quality problems while the contract is still live and they can be fixed.
It is the first piece of work, and we would rather say that in the assessment than discover it in month four. The GIS, the billing system and the field reality usually disagree. We reconcile them, list the disagreements for verification, and improve the map with the evidence the meter data itself provides. Everything downstream inherits the quality of that map.
Yes, and it is one of the most valuable things on this page. Data availability by meter and by day, command success rates and communication failures by area are computed from the raw feed you receive, independently of the provider's own reporting, so a contracted service level is measured by you and a deduction can be evidenced.
No. Nothing on this page disconnects, penalises or issues a notice. The system produces a ranked list with the evidence behind each case, and a person visits, verifies and decides. The outcome of that visit is recorded, which is also what makes the next ranking better.
In your tenancy or on your own hardware. Utilities that are state undertakings usually run this on their own infrastructure with open-weight models, and everything on this page can be built that way. Where a hosted model is used at all, what is sent to it is agreed and recorded before build.
No, and the parts that are not are usually what makes the AI work. Ingestion from the head-end and the meter data platform, consumer indexing, reconciliation and the field application are ordinary engineering, and they are most of the effort. The same is true of the training: alongside the AI tracks we run SQL, data engineering, cloud, enterprise systems and cybersecurity.
Parts of it. Generation and demand forecasting, deviation exposure, asset failure prediction, contract review and regulatory filing support all carry over. The loss, theft and consumer indexing work is distribution-specific and would not apply.
Related
The service this ships as, the products behind it, and the page on governance.
The meter, GIS and billing data under every number.
→ ProductTransformer and asset failure from the signals you get.
→ ProductBilling exceptions and service level verification.
→ IndustryThe governance side, for state utilities and boards.
→ FunctionThe function that owns the meter data platform.
→ PillarThe engineering side, for systems that run a network.
→A month of interval reads and events for one division, with the transformer mapping you hold. We call you within 48 hours and go through what the feed can already carry, where the mapping breaks, and what a first build would take.
Talk to us