Functional DemoAutomation System

FuelFlow

Fuel receipt capture, approval, and reimbursement automation for mobile teams.

What it solves

Companies with mobile drivers and field crews collect fuel receipts through texted photos, spreadsheets, and scattered payment tracking. Receipts get lost, reimbursements are delayed and disputed, and duplicate or suspicious submissions slip through. Admins have no clean, auditable path from a receipt to an accurate, on-time payout.

Built for Businesses with mobile drivers, field teams, and trucking operators, and the admins who reimburse them.

Overview

FuelFlow turns receipt scans into verified reimbursement batches and scheduled payouts. A driver scans a receipt, an OCR pipeline extracts the merchant, date, amount, and fuel details, and the driver confirms before submitting. Duplicate and fraud checks run across image hashes, perceptual hashes, and OCR field matching, and a rules engine decides auto-approval, review, or flag based on confidence and company limits. Admins review flagged receipts, group approved ones into weekly or monthly batches, and release payouts once batches are clean and drivers are payout-ready. Reports export to accounting-ready CSV and PDF.

What this demonstrates

Builds a receipt intelligence pipeline behind a provider abstraction (Veryfi, Textract, and a mock provider) with field normalization and confidence banding, a four-layer duplicate detection system, and a rules engine that gates auto-approval on confidence, limits, and fraud signals. Payout automation runs on background jobs and only marks receipts paid after provider confirmation, with audit logging throughout.

What makes it different

Not a generic expense platform. A focused fuel reimbursement automation system that turns receipt scans into verified batches and scheduled payouts.

Scan, confirm, submitThe driver flow is three steps and under a minute.
Provider-agnostic OCRA provider abstraction avoids lock-in to one vendor.
Four-layer duplicate detectionImage hash, perceptual hash, OCR fields, and provider signals.
Confidence-gated approvalMissing total or date blocks auto-approval.
Pay only after confirmationReceipts are marked paid only after payout confirms.
Never store raw payment dataSensitive account data is never stored directly.

Core capabilities · 14

Receipt scanner
Camera capture with client-side image quality checks.
OCR provider abstraction
Veryfi, Textract, and mock providers behind one interface.
Field normalization
Maps raw OCR into merchant, date, amount, and fuel fields.
Confidence scoring
High, medium, and low bands drive routing.
Duplicate detection
Exact hash, perceptual hash, and OCR field matching.
Fraud signals
Tampering, screenshot, and high-velocity detection flags.
Reimbursement rules engine
Decides auto-approve, review, flag, or reject.
Reimbursement batches
Groups approved receipts by weekly, biweekly, or monthly period.
Payout automation
Schedules and confirms payouts through a provider.
Admin receipt review
Queue with approve, reject, and request-clarification actions.
Driver dashboard
Pending, approved, scheduled, and paid totals in plain language.
Reports and exports
Driver, vehicle, and monthly reports in CSV and PDF.
Audit logging
Records before and after state for every decision.
Company workspaces
Multi-tenant companies with role-based access.

How it works

Driver

Log in
Tap Scan Receipt
Capture or upload a photo
Pass image quality checks
Review OCR-extracted details
Confirm and submit
Track status: pending, approved, scheduled, paid

Admin

Review pending and flagged receipts
Approve, reject, or request clarification
Move approved receipts into batches
Review batch totals by driver
Approve or automate payout
Pay drivers through the provider
Export accounting reports

How the score works

Approval routing decision

OCR confidence
High-confidence receipts under limit can auto-approve.
Duplicate risk score
Elevated risk routes to review or blocks submission.
Fraud risk score
Suspicious signals route to admin review.
Amount vs company limit
Over-limit amounts require review.
Receipt date window
Dates outside the allowed window require review.
Payout setup status
Missing payout onboarding blocks payment.

Key screens

Driver home
Scan receipt / camera
Receipt confirmation
Receipt status
Admin dashboard
Receipt review
Reimbursement batches
Payouts
Reports
Settings

Data model

Company
User
Vehicle
Receipt
ReceiptOcrResult
ReceiptFlag
ReimbursementBatch
Payout
AuditLog

Screens · 8

FuelFlow Receipts review list filtered to Pending, showing submitted fuel receipts by merchant, driver, vehicle, date, and amount — Receipts review list

FuelFlow receipt detail for a $73.76 Shell manual entry by Aisha Khan with gallons, price, scan confidence, no flags, and Approve/Reject/Request changes actions — Receipt detail

FuelFlow Team list of drivers and admins with email, receipt counts, and role/status badges (Owner, Manager, Driver, Ready or Pending) — Team roster

FuelFlow Vehicles list of ten fleet trucks with unit numbers, assigned drivers, receipt counts, and Active or Inactive status — Vehicles list

FuelFlow reimbursement batch detail for Jul 15–22 2026, Approved, totaling $1,156.03 across 4 drivers with per-driver totals and a Schedule payouts button — Reimbursement batch

FuelFlow Payouts page with a setup-incomplete notice and a recent payouts list showing one failed payout with Retry and several Paid entries — Payouts list

FuelFlow Reports for July 2026 with receipts and payment tiles, a reconciled banner, export buttons, and by-driver and fuel-spend-by-vehicle breakdowns — Reports and exports

FuelFlow Settings with company settings (reimbursement frequency, auto-approval limits, currency) and a system status panel showing healthy database and queues — FuelFlow settings

How it is built

A Next.js and React application with a Prisma data layer over PostgreSQL, Redis and BullMQ for background OCR and payout jobs, and S3-compatible object storage for receipt images. Receipt OCR runs through AWS Textract behind a provider interface, and PDF summaries are generated server-side.

APIs and services

AWS S3
AWS Textract

Classification

Building something with similar Document OCR and Extraction, Payments, Background Job Pipelines functionality?

Start a project and reference this one — the conversation begins with what you have already seen.

Start a project