Manually transcribing a utility bill is the single most boring task in brokering. The bill parser does it in about 10 seconds. Drop a PDF in; structured fields come out.
What it extracts
For any TDU bill (Oncor, CenterPoint, AEP North, AEP Central, TNMP) and most REP bills:
- ESI ID — the 17-digit meter number
- Service address — full street + city + ZIP
- Annual usage (kWh) — projected from the billing-cycle usage shown
- Demand (kW) — for accounts on demand tariffs
- Current REP — who's selling them power today
- Current rate (¢/kWh) — what they're paying right now
- Contract end date — if disclosed on the bill
Step-by-step
Open the parser
Go to Bill Parser. Drop a PDF onto the upload zone, or click to browse. Multi-page bills, scanned bills, and even photos of bills work (clarity varies).
Drop a bill PDF here. Multiple bills can be uploaded in a batch. Review the extracted fields
The parser shows you each field it pulled, alongside the snippet of the bill where it found that data. Confidence is color-coded — green high, yellow medium, red unsure.
Extraction result — every field has a confidence indicator and a source snippet. Push to a customer
Click Save to customer. If the bill's ESI ID matches an existing customer, the parser updates that record. Otherwise, it creates a new customer pre-populated with everything it found.
Tips & gotchas
A clean PDF download from the customer's online utility account beats a phone photo every time. Photos at an angle, in low light, or with reflections drop accuracy noticeably.
For a new book of business with 50 bills to load, batch upload all of them. The parser runs them in parallel and you review/save at the end. Faster than one-at-a-time.
Bill PDFs are sent to the AI vision model (Google Vertex Gemini) for extraction, then deleted from the AI provider's side. The PDF itself is stored in your Documents and only visible to your brokerage.