Industry knowledge. Custom AI.
Custom AI for energy. Better decisions from site to customer.
Energy decisions connect physical assets, weather, location, operations and customer demand. HolisticRM develops a bespoke model set around that whole picture: the right data, appropriate prediction and optimisation models, and a workflow your specialists can use. Our published experience includes solar siting, EV charging location intelligence and service-station retail modelling.
- Screen investment locations with richer evidence
- Forecast and prioritise within operating constraints
- Connect operational intelligence with customer growth
Published company project · Service-station retail modelling
ESA demonstration · HolisticRM subcontractor to GeoX
Completed ESA feasibility study · HolisticRM subcontractor to GeoX
How to choose AI for solar, charging and roadside retail investment
A high-scoring location is useful only when it supports the investment you are actually considering. A solar developer needs land and grid feasibility; a charging operator needs accessible demand; a service-station retailer needs the right format for the local catchment. HolisticRM’s energy work connects these different decisions to different data and models.
By HolisticRM team.
Three energy decisions from the HolisticRM portfolio
Our MOL Group / Fresh Corner reference describes location and retail-format modelling using GeoX spatial information, competition context and ensemble machine learning. The delivery combined Azure ML, Azure Data Factory and Power BI. Its lesson for an energy buyer: predicting traffic alone does not establish which commercial format fits the location.
For OPT4SOL.AI, ESA identifies HolisticRM as GeoX’s subcontractor. The ongoing demonstration targets solar suitability at 100 m resolution, combining approximately 70 spatial layers with Earth observation. ESA’s October 2026 update records design and pilot preparation, not a completed five-country rollout. E-Spacement is a completed EV-siting feasibility study, also with GeoX as prime and HolisticRM as subcontractor.
Choose by the investment decision, not the AI label
| Decision | Model or workflow | Data and acceptance test |
|---|---|---|
| Solar pre-feasibility | Spatial suitability scoring plus explicit exclusions | Land use, terrain, protected areas, grid proximity and site labels. Separate suitability from grid-connection approval. |
| EV network expansion | Utilisation prediction plus scenario comparison | Charging sessions, catchment, access, competitors and available power. Compare opening cohorts and charger formats. |
| Roadside retail format | Sales/affinity modelling and constrained format comparison | Local demand, competition and observed format performance. Account for cannibalisation and operating economics. |
A practical sequence for an investment team
- Start with one decision gate. Choose the screening step where analysts can reject, investigate or advance a site. Name the development or investment owner and record the current shortlist method.
- Make exclusions visible before scoring. Grid, land and permitting constraints should be explicit fields, not buried in an overall score. Date each external layer and distinguish missing information from a favourable value.
- Test which data earns its cost. Compare business data alone, geospatial enrichment and Earth-observation features on the same held-out regions. HolisticRM’s data science approach develops these inputs and the model together.
- Deliver a reviewable shortlist. Show the main drivers, uncertainty and unanswered due-diligence questions in the map or dashboard analysts already use. An AI agent may assemble a site dossier; it should not invent connection capacity or approve an investment.
The near-term benefits are less repetitive screening and a clearer investment discussion. Validate time per reviewed site and the proportion of shortlisted sites that pass engineering review. Realised utilisation, revenue and return on capital are later measures; they require operating evidence after the investment.
Sample: prioritise a solar-screening pilot before wider automation
Sample — illustrative evaluation plan, not a client result.
An illustrative developer has 100 candidate parcels. Its agreed exclusions remove 25 before modelling; 75 remain for prioritisation. Compare the current rules with a spatial model and ask both for a 15-site shortlist. Engineers review the union of the lists without being told which method proposed each site. Record missed hard exclusions, review hours and sites worth further investigation. These numbers define a worked pilot design, not an OPT4SOL.AI result.
What to prioritise
- First: make exclusion data and review outcomes dependable; otherwise a more complex model simply ranks unreliable inputs.
- Next: test the additional value of satellite and catchment features on a separate region.
- Later: automate approved dossier preparation or recurring refreshes after the screening workflow proves useful.
See space-enabled AI delivery · Compare our model-development approach · Develop time-correct spatial features · Connect approved site-dossier agents.
Bring a candidate-site list, the screening rules and the decision your analysts struggle to make. We can define the first evidence-bearing pilot together.
Top 10 AI use cases for energy
Ten high-value opportunities to assess with your energy team. This is a practical shortlist, not a universal ranking: priorities depend on your assets, data and business case. Each card describes a proposed implementation and measurement plan; delivered project references appear separately below.
01Solar development site screeningRank candidate PV sites before committing detailed engineering effort, using a consistent view of location suitability.
What we implementCombine terrain, land use, grid proximity and Earth observation features; compare suitability models and expose constraints in a map-based shortlist.
Business benefitConcentrate feasibility work on stronger candidates and reduce repetitive desktop screening.
How value is measuredMeasure analyst hours per screened site and shortlist progression to engineering review.
Risk: Historical installations can reflect past incentives rather than future site viability. Control: Keep engineering exclusions explicit, validate across regions and require grid and permitting due diligence.
02EV charging location and format selectionIdentify promising charging locations and charger configurations for the demand patterns your network needs to serve.
What we implementJoin charging sessions with traffic, catchment, nearby services and grid data; estimate demand ranges for candidate sites and configuration scenarios.
Business benefitSupport more disciplined capital allocation and improve the evidence behind network expansion.
How value is measuredTrack utilisation forecast error and realised energy dispensed by opening cohort.
Risk: New competitors, fleet contracts or connection constraints can change local demand. Control: Test scenarios, reserve unseen locations for validation and refresh assumptions before investment approval.
03Electricity demand forecastingGive planners a clearer view of demand at the time horizon and aggregation level of their decisions.
What we implementCreate time-aligned features from meter histories, weather forecasts, calendars and operating schedules; compare statistical and machine-learning forecasts with uncertainty intervals.
Business benefitImprove planning inputs and reduce manual forecast adjustment where the pilot demonstrates value.
How value is measuredMeasure forecast error by horizon, peak period and customer or asset segment.
Risk: Using revised weather or late-arriving readings can make backtests unrealistically optimistic. Control: Reconstruct data available at each forecast time and test rolling future periods.
04Renewable generation forecastingAnticipate variable solar or wind output so operations teams can plan around likely production and uncertainty.
What we implementCombine plant telemetry, equipment availability and weather forecasts; compare physical baselines with learned corrections and site-specific forecasting models.
Business benefitSupport scheduling and reduce avoidable surprises in short-term production planning.
How value is measuredTrack generation forecast error, interval coverage and downstream planning cost against baseline.
Risk: Curtailment or sensor faults may be mistaken for normal weather-driven output changes. Control: Label operational events separately, monitor input quality and provide a fallback forecast when feeds fail.
05Asset condition and maintenance prioritisationFocus maintenance attention on assets whose operating patterns suggest a developing issue requiring engineering review.
What we implementConnect sensor streams, inspections and work orders; compare condition baselines with anomaly or failure-risk models and route alerts into maintenance workflows.
Business benefitReduce unnecessary inspection effort and create earlier opportunities for planned intervention.
How value is measuredMeasure useful alert rate, warning lead time and unplanned downtime by asset class.
Risk: Rare failures and changing operating loads can produce noisy or misleading alerts. Control: Validate with maintenance engineers, start in advisory mode and tune thresholds to response capacity.
06Grid and infrastructure inspection triageHelp inspection teams prioritise images and locations that deserve closer assessment, without replacing required engineering checks.
What we implementCombine permitted aerial or field imagery with asset records; train defect or vegetation classifiers and attach reviewable evidence to each flagged item.
Business benefitShorten image review queues and concentrate specialist effort on higher-priority findings.
How value is measuredTrack defect recall, false alerts per inspection and time to reviewed disposition.
Risk: Lighting, season and camera changes can hide defects or create spurious detections. Control: Test across capture conditions, retain human sign-off and collect reviewed examples for controlled retraining.
07Storage and flexible-load scheduling supportRecommend operating schedules that balance energy cost, service needs and equipment constraints under uncertain demand and generation.
What we implementFeed validated forecasts into a constrained optimiser; simulate tariff, availability and degradation assumptions before showing recommended schedules to authorised operators.
Business benefitIdentify potential operating savings while protecting service levels and asset life.
How value is measuredCompare net operating cost, constraint violations and battery throughput against existing scheduling.
Risk: Forecast errors or incomplete constraints can make an apparently optimal schedule impractical. Control: Stress-test adverse scenarios, enforce hard limits and use operator approval with a safe fallback.
08Retail energy retention and next best actionHelp customer teams select timely, relevant service actions for households, small businesses and charging-network customers.
What we implementCombine permitted contract, service and interaction data; estimate churn or response propensity and test eligible actions through controlled customer cohorts.
Business benefitImprove retention effort and relevance while reducing indiscriminate contact volume.
How value is measuredMeasure incremental retention, contact cost and complaints against a comparable control group.
Risk: A predictive score alone does not establish which intervention will help a customer. Control: Test action effects, apply eligibility and contact rules, and review outcomes across customer segments.
09Energy knowledge and service assistantsHelp staff answer equipment, tariff or process questions from approved documentation with a clear route to experts.
What we implementBuild permission-aware retrieval over versioned manuals and service content; add language-model responses with citations, refusal rules and approved workflow tools.
Business benefitReduce document search time and support more consistent service explanations.
How value is measuredMeasure verified answer accuracy, resolution time, escalation quality and unsupported-answer rate.
Risk: Outdated documents or fabricated answers can mislead staff or customers about important terms. Control: Evaluate realistic questions, date authoritative content and escalate safety, contractual or uncertain answers.
10AI search visibility for consumer energy servicesMake relevant energy tariffs, EV charging services and household propositions easier to understand in AI-assisted customer research.
What we implementAudit product facts and answer coverage; improve useful comparison content, eligibility explanations and location pages, then monitor a fixed set of consumer questions.
Business benefitSupport qualified discovery and reduce ambiguity before a customer visits or enquires.
How value is measuredTrack cited-page coverage, answer accuracy, qualified enquiries and attributable referral conversions.
Risk: AI answers vary and may repeat expired prices, availability or unsupported savings claims. Control: Maintain dated source content, repeat measurements consistently and prioritise corrections over ranking promises.
These are opportunities to evaluate for your organisation. Priorities, feasibility and targets are agreed against your data, constraints and baseline.
Tangible business benefits
More efficient investment screening
Reduce repetitive analysis and make shortlist decisions easier to compare before expensive specialist work begins.
Measure: Baseline analyst time, cost per assessed site and progression to accepted feasibility review.
Benefits beyond immediate financial returns
Shared confidence in decisions
Engineers, commercial teams and management work from the same inputs, assumptions and recorded reasons for recommendations.
-
Define the decision and value
Choose one operational or commercial decision with its owner. Agree the current baseline, operating limits and the evidence needed to proceed. Our approach builds on CRISP-DM and draws on PMI CPMAI principles.
Output: Business case, decision boundaries and acceptance measures.
-
Map energy and location data
Review telemetry, asset registers, weather, spatial layers and customer data as relevant. Record timeliness, licences, missingness and which inputs actually exist when a decision is made.
Output: Data readiness assessment and prioritised improvements.
-
Improve data and models together
Our holistic method develops meaningful inputs alongside suitable model families. Engineers help define constraints while we compare practical baselines, forecasts, classifiers and optimisation components.
Output: Versioned experiments and a justified model shortlist.
-
Validate outside the familiar sample
Test future periods, unseen assets or separate regions as appropriate. Examine difficult weather and demand conditions, operating cost, uncertainty and failure modes before selecting a model combination.
Output: Validation report and go, refine or stop decision.
-
Pilot in the real workflow
Start with advisory recommendations and a defined user group. Integrate into the existing map, planning or maintenance workflow; train users and record which recommendations are actionable.
Output: Pilot workflow, operator guidance and measured acceptance.
-
Deploy with accountable monitoring
Release approved versions with access control, input checks and fallback behaviour. Monitor the business measure as well as model quality; feed confirmed outcomes into the next controlled improvement cycle.
Output: Operational handover, monitoring and retraining plan.
Better data and better models, developed together.
Our Holistic Agile ML approach builds on CRISP-DM and draws on the business-led, iterative principles of PMI’s CPMAI methodology. Domain experts shape the decision, usable data and acceptance criteria. We improve relevant inputs while comparing suitable model families, architectures and model combinations against a practical baseline.
Protected test sets, versioned experiments and agreed go/no-go gates help catch leakage, weak assumptions and unnecessary complexity before wider rollout. A controlled pilot, human escalation, monitoring and rollback reduce operational exposure. The result is a bespoke model set selected for your business case and operating constraints.
Make the risks visible. Build the controls in.
Risk controls are tailored to the decision, data and operating environment, with named owners and agreed acceptance evidence.
- Operational and safety consequences
-
An incorrect forecast or recommendation can affect physical operations, service continuity or asset life.
How we reduce the risk: Agree decision authority, keep engineering constraints outside the model and progress from shadow evaluation to supervised use with rollback.
- Misleading historical evidence
-
A backtest may benefit from future information or overlook changes in weather, tariffs, assets and customer behaviour.
How we reduce the risk: Use decision-time data, future-period tests and scenario analysis. Evaluate the data and model together at every release gate.
- Uncertain site economics
-
A high suitability score cannot establish grid capacity, land access, permits or commercial viability.
How we reduce the risk: Expose assumptions and uncertainty, separate exclusions from learned scores and retain qualified technical and commercial due diligence.
- Sensitive systems and customer data
-
Unnecessary access or poorly controlled integrations can expose operational information and personal data.
How we reduce the risk: Minimise inputs, define access by role, separate environments and test integrations before granting production permissions.
For consumer tariffs, home energy and EV charging, AEO and GEO complement your product, local-search and customer-experience work. We improve the content AI systems can find and substantiate, then measure visibility alongside real customer actions.
What the work includes
- Build clear pages for tariff suitability, eligibility, charging access and common customer comparisons.
- Keep prices, coverage, availability and conditions consistent across owned pages and relevant listings.
- Monitor a repeatable question set by audience and location, checking both mentions and factual accuracy.
- Connect prioritised content changes to qualified enquiries, booking journeys and attributable conversions.
Can AI select a solar or EV charging site on its own?
AI can help prioritise candidates and show the factors behind a score. Investment approval still needs engineering, grid, land, permitting and commercial checks. We design the workflow so the model accelerates screening and keeps these decision responsibilities explicit.
Do we need every data source before starting?
No. We begin with the decision and a useful baseline, then identify which missing inputs could change the outcome. A small, well-defined pilot can reveal whether weather, telemetry or external location data adds value before a wider integration programme.
How will we know the project has paid off?
Agree the baseline, measurement period and business owner before development. Forecast quality, operating savings and customer outcomes need different tests. We separate technical improvements from realised business benefit and account for operating effort, intervention cost and relevant external changes.
Which energy decision should AI improve first?
Bring one site, operating or customer challenge. We will help identify useful data, a credible pilot and the measures that would justify scaling.
Bring one business challenge, the decision you want to improve and an outline of your available data. Please leave out confidential customer or patient information.