Turn Tariffs into Data: Verified Extraction for Utility Rates, Interconnection Agreements, and Regulatory Filings

Utilities and energy teams live in documents that behave like law:

• tariffs and rate schedules
• interconnection agreements
• regulatory filings and approvals
• compliance reporting requirements
• service rules and exceptions

These documents are long, versioned, full of tables, and packed with “it depends” clauses.

The cost of getting them wrong is high:

• billing errors,
• compliance exposure,
• delayed interconnections,
• and operational confusion.

The high-value use case is transforming these documents into verified, reviewable data:

• extract key fields,
• attach citations,
• and keep an audit trail of what changed.

Where the value shows up first

1) Rate schedule extraction (tables that drive billing)

Extract:

• rate class identifiers
• usage tiers
• unit rates
• demand charges (if applicable)
• fixed charges
• riders/surcharges references
• effective dates and expiration/supersession notes

Citations matter because rate tables are error traps: rows are easy to misread, footnotes change meaning, and a single wrong number can cascade into billing issues.

2) Interconnection agreements and technical requirements

Extract:

• project/customer identifiers (where present)
• milestone dates and deadlines
• cost responsibility language
• technical standards referenced
• curtailment and operational requirements
• termination/default provisions

Citations matter because obligations are legal/contractual language, not simple fields.

3) Regulatory filings and approvals (what is actually authorized)

Extract:

• filing docket references
• approved effective dates
• conditions or carve-outs
• compliance reporting deadlines
• scope language

Citations matter because approvals often hinge on exact phrasing.

Practical workflow: version-aware and defensible

Step 1: Ingest + classify by document type and version

A tariff without version control is a liability. Track:

• effective date
• revision/version
• supersedes/replace references (when present)

Step 2: Extract “operational fields” + citations

Don’t try to extract the whole document. Extract what your teams need to act:

• rates
• effective dates
• key obligations
• exceptions

Step 3: Build a change log (“what changed?”)

When a new version arrives:

• extract again
• diff the structured fields
• show “changed fields” with citations from both versions

This is incredibly high value for teams managing billing logic, customer communications, and compliance deadlines.

Step 4: Route review to the right owners

• billing team reviews rate table changes
• regulatory/compliance reviews obligations and conditions
• operations reviews interconnection milestones

Citations make review fast because each change is linked to the exact source.

Schema sketch: rate schedule + obligations

{
  "schema": {
    "document_type": { "type": "string", "description": "tariff, rate_schedule, interconnection_agreement, regulatory_order, other" },
    "jurisdiction": { "type": "string" },
    "effective_date": { "type": "date" },
    "version_identifier": { "type": "string" },

    "rate_classes": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "rate_class_name": { "type": "string" },
          "fixed_charge": { "type": "number" },
          "usage_tiers": {
            "type": "array",
            "items": {
              "type": "object",
              "properties": {
                "tier_start": { "type": "number" },
                "tier_end": { "type": "number" },
                "unit_rate": { "type": "number" },
                "unit": { "type": "string", "description": "e.g., kWh" }
              }
            }
          }
        }
      }
    },

    "obligations": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "topic": { "type": "string", "description": "e.g., billing, interconnection, reporting" },
          "obligation_text": { "type": "string", "description": "Short extracted clause; must be cited" },
          "deadline_date": { "type": "date" }
        }
      }
    }
  },
  "options": { "confidence_threshold": 0.85 }
}

What to measure

• time-to-implement a rate change after a new version is issued
• number of billing incidents caused by misread tables
• time-to-review regulatory changes
• completeness of version tracking (no “mystery” documents)

Tariffs and filings don’t need to be unreadable. Evidence-backed extraction turns them into verifiable, operational inputs—without losing legal defensibility.