Tarun Bulchandani
AI-led products

AI-led purchasing products

Four AI-led products covering the energy purchasing pillar: demand forecasting, market execution against OMIE and MIBEL, real-time guarantee monitoring, and invoice reconciliation. Built against a single reference architecture, deployed in your tenancy, designed for the regulatory posture an Iberian retailer actually needs.

The advisory pillar covers what good looks like in energy purchasing. The products on this page are the AI-led implementations that deliver it. Four products, scoped tightly to the sub-topics in the pillar.

Product 1: Forecast

AI-led hourly demand forecasting calibrated to the Iberian retail context.

Most retailers under 500 GWh forecast with a mix of seasonal averages and last-week shape, with manual overrides for known events. Forecast replaces this with an hourly model that learns from the retailer's customer mix, weather, calendar, tariff shape and metered data, and that updates against the day-ahead and intraday auction schedule.

What it does:

  • Hourly demand forecast across the day-ahead and intraday horizons, calibrated to each DSO region in the retailer's footprint
  • Customer-mix awareness (residential PVPC, free-market residential, SME bilateral, corporate structured) so the forecast is a weighted aggregate rather than a single shape applied to the whole book
  • Weather feed at customer-postcode granularity, not city level
  • Recalibration on metered data as it arrives from the DSOs, with automated drift detection
  • Outputs in the shape the downstream purchasing workflow consumes: hourly granularity for day-ahead and intraday, daily for hedging

What it integrates with:

  • The retailer's CRM for customer-mix data
  • The DSO metered-data feeds (with their 30 to 60 day lag accounted for)
  • Met Office and AEMET weather feeds
  • The retailer's bid generation workflow downstream

Engagement: 6 to 10 weeks to build into your tenancy. Optional 90-day support. Managed-service retainer available afterwards.

Best for: Retailers with a residential PV-rooftop share above 15%, or with material SME exposure to heat-sensitive demand, or with a customer book spread across multiple DSO regions.

Related reading: Demand forecasting in an Iberian retail context.

Product 2: Execute

Agent-mediated execution against OMIE day-ahead, MIBEL intraday, MEFF futures and the GdO secondary market.

The execution workflow today, in most retailers, is a small team running spreadsheets against the SIOM platform, with limited validation and one or two named people carrying the operational risk. Execute is the working agent layer that runs the bid generation, validation, submission and post-trade reconciliation, with the trader in the loop for material decisions.

What it does:

  • Bid generation from the Forecast output, the hedging position and the published price curves
  • Pre-submission validation against the OMIE schema (catches the misformatted bids that get silently rejected today)
  • Submission with hard cutoff discipline, secondary named operator on the workflow, alerting on system clock drift
  • Intraday position management across the six MIBEL intraday auctions and the continuous platform
  • MEFF futures execution on the hedging cadence the retailer defines
  • GdO secondary market participation for green-tariff retailers

What it integrates with:

  • OMIE SIOM platform via the documented connector
  • MEFF / BME Clearing via the published interfaces
  • MIBEL intraday platform
  • The Forecast product (or any equivalent forecast)
  • The retailer's position-tracking layer

Engagement: 10 to 16 weeks. Material customisation to the retailer's specific risk policy. Managed-service retainer covering ongoing operations strongly recommended.

Best for: Retailers running 500 GWh or larger, or smaller retailers running structured products that benefit from active intraday participation.

Related reading: Operating against OMIE, Purchasing in MIBEL intraday markets.

Product 3: Margin

Real-time trading guarantee monitoring across OMIE, MEFF, REE and any cross-border CCP positions.

The guarantee position is one of the largest working-capital items on a retailer's balance sheet and one of the most operationally fragile. Margin replaces the spreadsheet that runs the calculation today with a live engine that recomputes against each position change, reconciles against the counterparty's published statement daily, and escalates margin events inside the counterparty SLA.

What it does:

  • Live calculation of the guarantee requirement against OMIE, MEFF, REE and the SIDC infrastructure
  • Daily reconciliation against the published counterparty statements
  • Working-capital optimisation across cash, bank guarantee and insurance-policy instruments
  • Margin call escalation with named accountabilities and explicit SLAs
  • Pre-agreed credit-line check for the default scenario, with monthly review at the treasury level

What it integrates with:

  • OMIE position feeds
  • MEFF / BME Clearing statements
  • REE balancing position
  • The retailer's treasury system

Engagement: 8 to 12 weeks. Strongly recommended ongoing managed-service support given the working-capital weight.

Best for: Every retailer running active positions in OMIE, MEFF or REE. Single most common starting point for the suite.

Related reading: Calculating and monitoring trading guarantees.

Product 4: Reconcile

AI-led invoice reconciliation against DSO ATR settlements, REE balancing, OMIE clearing and MEFF mark-to-market.

The monthly reconciliation cycle against the system operators, DSOs and market operators is where most retailers find their largest month-on-month variances. Reconcile is the working engine that ingests each invoice in its native format, normalises against the retailer's canonical book of record, and surfaces the variances against defined materiality thresholds.

What it does:

  • Invoice ingestion across the 10 to 12 monthly invoices a typical national retailer receives, each in its native format
  • Normalisation against the canonical book of record
  • Variance detection against materiality thresholds, by line item and by counterparty
  • Dispute workflow with documented resolution paths and target SLAs
  • Monthly close discipline with named accountabilities
  • Recurring-pattern detection so root causes are surfaced rather than recurring variances repeatedly

What it integrates with:

  • The retailer's billing engine (canonical book of record)
  • DSO ATR settlement feeds
  • REE balancing settlement
  • OMIE settlement reports
  • MEFF clearing daily and weekly reports

Engagement: 8 to 12 weeks to build. Ongoing managed support strongly recommended (the monthly close is operationally weighty).

Best for: Every retailer with material monthly flow. Payback typically inside 4 to 6 months on the unrecovered variance reduction.

Related reading: Billing management: DSO, REE, OMIE and MEFF.

Where these products sit together

The four products are designed to be implemented together, but each can run standalone. The most common sequence:

  1. Margin first. Working capital weight, payback in weeks rather than months, sets the operational discipline the rest of the suite relies on.
  2. Reconcile next. Largest single margin recovery, forces the canonical book of record to be in place, which the rest of the products use.
  3. Forecast third. Foundation for any improvement in the purchasing decision quality.
  4. Execute last. Builds on the three above and needs them to be operationally clean.

The diagnostic engagement covered in the products hub sets the actual sequence based on the retailer's specific operational position.

Related