Case Study: AI in
FleetDrive
360 for PrePass
Introduction
MetroMax has two AI features live in production inside PrePass’s FleetDrive 360 compliance workflow: an outbound voice agent that schedules DOT drug and alcohol tests, and AI-OCR that turns driver documents into bulk onboarding files.
At a glance
The challenge
FleetDrive 360 users run regulated compliance work that was heavily manual in two places:
Test scheduling.Staff had to phone drivers one by one to book DOT drug and alcohol tests under 49 CFR Part 382, then log each outcome.
Driver onboarding. Adding drivers in volume meant re-keying data from licences and other documents that vary by state and issuer.
In a regulated product, errors carry more weight. A mis-keyed field or an unrecorded call is a compliance error, not a typo.
Feature 1: AI outbound voice agent
The voice agent calls drivers directly to schedule DOT drug and alcohol testing, confirms a slot and records the outcome back into the compliance workflow, with no human placing the call.
Speaks to drivers unsupervised, inside a 49 CFR Part 382 process
Captures outcomes as structured data, not free-text notes
Hands off to a human through a defined escalation path
Keeps an auditable record of what was said and agreed
Feature 2: Bulk upload AI-OCR
AI-OCR scans driver documents and generates the bulk upload sheet used to add drivers in volume, replacing manual re-keying.
Handles document variety across states and issuers, the real engineering cost of the feature
Extracts data field by field and validates it before output
Adds a human checkpoint before any record is committed
How we built it safely
Both features follow one pattern: AI generates and proposes, humans approve anything that reaches a compliance record, and every step leaves a trail.
Structured over free text. Outcomes and extracted fields land in defined slots the workflow can check.
Validation before commit. Extracted data is validated, then confirmed by a person before it becomes a driver record.
Clear escalation. The voice agent hands off to a human when a call goes outside its script.
Auditability. What was said, agreed and extracted is recorded, which matters for audits in a regulated product.
Outcomes
Both features are live in production. Test-scheduling calls no longer need a person to place them, and bulk driver onboarding no longer relies on manual re-keying.
Measured results are not yet in this draft. Figures to confirm with PrePass before publishing:
Calls placed and tests scheduled by the voice agent per month
Share of calls completed without human escalation
Drivers onboarded through AI-OCR and time saved per batch
Extraction accuracy before the human checkpoint
Client quote and approval to name PrePass publicly
