Skip to content

AI Features

MeterBase integrates with Anthropic Claude to provide AI-powered capabilities across the platform. These features automate manual data entry, extract insights from unstructured documents, and provide natural language analysis of complex rate structures.

API Key Required

AI features require an ANTHROPIC_API_KEY in your environment configuration. Without it, AI features are disabled but all other platform functionality remains available. See the Configuration Reference for setup details.


Bill Analysis

Drop a PDF utility bill into MeterBase and get structured data in seconds.

What It Does

The AI reads the full utility bill PDF and extracts:

  • Billing period dates
  • Account and meter numbers
  • Total kWh usage (with peak/off-peak breakdown if available)
  • Peak demand in kW
  • Every individual line-item charge, categorized by type
  • Rate schedule name
  • Service address
  • Taxes and riders

How It Works

sequenceDiagram participant U as User participant G as MeterBase participant AI as Anthropic Claude U->>G: Upload bill PDF G->>AI: Send PDF with extraction prompt AI-->>G: Structured JSON response G->>G: Validate and store extracted data G->>G: Run anomaly detection against history G-->>U: Extracted data + AI insights
  1. The PDF is sent to Claude with a structured extraction prompt that specifies the exact fields and format expected.
  2. Claude processes the document and returns structured JSON with all extracted fields.
  3. MeterBase validates the response (checks for reasonable values, consistent totals, valid dates).
  4. If bill history exists for the property, MeterBase compares the new bill against historical patterns to detect anomalies.
  5. The extracted data is presented for user review before being saved.

Accuracy and Limitations

Document Type Expected Accuracy
Utility-generated PDFs 95%+ field extraction accuracy
Scanned paper bills (good quality) 85-90% accuracy
Poor quality scans or photos 60-75% accuracy; manual review recommended
Multi-page bills Fully supported
Multi-meter consolidated bills Supported with meter-level breakdowns

Best Results

For highest accuracy, upload the original digital PDF provided by the utility (from their online portal or email). Scanned paper bills work but may have lower accuracy for small text and numbers.


Rate Schedule Extraction

Extract structured tariff data from a utility's published rate schedule PDF.

What It Does

Upload a rate schedule PDF (the document that defines a tariff's rates, tiers, TOU periods, and charges) and MeterBase's AI converts it into a structured tariff record that can be used for cost calculations and savings analysis.

Extracted Tariff Components

Component Example
Rate schedule name TOU-GS-2-B
Effective date January 1, 2026
Customer class / sector Commercial
Energy rates by period and season Summer On-Peak: $0.12/kWh
Demand rates by period and season Summer Demand: $15.00/kW
Tier definitions and boundaries Tier 1: 0-500 kWh, Tier 2: 501+ kWh
TOU period definitions On-Peak: 12pm-6pm weekdays
Fixed charges Customer Charge: $300/mo
Riders and surcharges Fuel Adjustment: $0.008/kWh
Eligibility criteria Demand between 20 kW and 500 kW

Usage

  1. Navigate to Tariffs > Extract from PDF
  2. Upload the rate schedule PDF
  3. Review the extracted tariff structure
  4. Edit any fields if needed
  5. Save to add the tariff to your database
curl -X POST https://api.meterbase.io/api/ai/extract-tariff \
  -H "Authorization: Bearer $TOKEN" \
  -F "file=@rate-schedule-tou-gs-2.pdf"

Custom Tariffs

This feature is particularly useful for adding tariffs that are not yet in the OpenEI database, such as newly published rates, municipal utility rates, or special contract rates.


Tariff Finder

Tell the AI what utility and state you are looking for, and it suggests where to find the official rate schedule documents.

What It Does

The Tariff Finder uses its knowledge of utility company websites, regulatory filings, and rate case databases to help you locate the official source documents for rate schedules. This is useful when:

  • You need the latest version of a tariff (rates may have changed since the last database update)
  • You want to verify that a tariff in the database is current
  • You need a tariff from a utility not well covered by OpenEI

Usage

curl -X POST https://api.meterbase.io/api/ai/tariff-finder \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "utility_name": "Pacific Gas & Electric",
    "state": "CA",
    "tariff_name": "B-20",
    "sector": "commercial"
  }'

Response:

{
  "utility": "Pacific Gas & Electric",
  "tariff": "B-20",
  "sources": [
    {
      "type": "utility_website",
      "url": "https://www.pge.com/tariffs/electric.shtml",
      "description": "PG&E's official electric tariff book, updated with each rate case."
    },
    {
      "type": "regulatory_filing",
      "url": "https://apps.cpuc.ca.gov/...",
      "description": "CPUC filings for PG&E rate cases."
    }
  ],
  "notes": "Schedule B-20 is a medium general demand-metered service rate for demand between 75 kW and 500 kW. Last major rate change was effective January 2026."
}

AI Tariff Comparison

Ask questions about tariffs in natural language and get structured, expert-level analysis.

What It Does

Instead of manually comparing rate tables, describe what you want to know and the AI provides a detailed comparison:

curl -X POST https://api.meterbase.io/api/ai/compare-tariffs \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "question": "Compare SCE TOU-GS-2-B and TOU-GS-2-D for a 48-unit apartment complex using about 38,000 kWh per month with 65 kW peak demand. Which is cheaper and why?",
    "tariff_ids": [4521, 4523]
  }'

Response:

{
  "analysis": "For your usage profile (38,000 kWh/month, 65 kW peak demand), **TOU-GS-2-D is approximately $480/month cheaper** than TOU-GS-2-B.\n\nThe primary driver is the demand charge difference: TOU-GS-2-D charges $11.50/kW in summer vs. $15.00/kW under TOU-GS-2-B. At 65 kW, that saves $227.50/month in summer alone.\n\nTOU-GS-2-D also has slightly lower on-peak energy rates ($0.11 vs $0.12/kWh), saving another ~$130/month based on your estimated on-peak usage.\n\nThe tradeoff: TOU-GS-2-D has a higher customer charge ($350 vs $300/month). But the demand and energy savings far outweigh this.\n\n**Estimated annual savings: $5,760.**",
  "summary": {
    "recommended_tariff": "TOU-GS-2-D",
    "annual_savings": 5760,
    "primary_driver": "Lower demand charges"
  }
}

Example Questions

  • "Which rate is best for a property that uses most of its electricity at night?"
  • "If we reduce our peak demand by 20%, how much would we save on TOU-GS-2-B?"
  • "Explain the difference between TOU-GS-2-B and TOU-GS-3-B. When does it make sense to be on TOU-GS-3?"
  • "Our bill went up 25% this month. Usage was flat. What changed?"

AI Usage and Costs

Feature Tokens per Request (approx.) Typical Latency
Bill analysis 3,000 - 8,000 5-15 seconds
Tariff extraction 5,000 - 15,000 10-30 seconds
Tariff finder 1,000 - 2,000 3-5 seconds
Tariff comparison 2,000 - 5,000 5-10 seconds

Rate Limits

AI requests count toward your tier's daily AI request limit (Free: 10/day, Pro: 200/day, Enterprise: unlimited). See the Configuration Reference for details.