Hundreds of vehicle titles a week, processed while the dealership sleeps

A multi-store auto dealer group buys hundreds of vehicles a week. Every one of those vehicles has a title, and every title has to be transcribed by hand onto a state application form before the dealership can move the vehicle on. Titles come from all fifty states, in every layout imaginable, with the most important field, the odometer reading, usually handwritten by whoever signed the vehicle over.

By the client's own estimate, that was costing the titles team 20 to 30 hours a week. It is exactly the kind of work AI vendors promise to make disappear, and exactly the kind of work where a silent AI mistake becomes a legal document with a wrong odometer reading on it.

What the system does

The pipeline runs overnight. It reads each scanned title packet, works out the chain of ownership across reassignments (auction to wholesaler to dealer is common, and the rules for who counts as the previous owner are not obvious), extracts the fields the state form needs, and fills the form. By the time staff arrive in the morning, the batch is done and sorted into two piles: completed forms, and titles flagged for human review.

The flagged pile is the point. This system was designed around a rule I set with the client on day one: zero unreviewed errors beats maximum automation. Over-flagging is acceptable. Guessing is not.

The guardrails

• Every extraction is cross-checked against a second, independent source the business already had: their own purchase records. Any mismatch, even by one mile, sends the title to a human.
• A second AI pass re-reads each title blind and has to agree with the first. Disagreement means a flag, not a coin flip.
• Titles with no cross-check coverage get flagged automatically. If the safety net cannot see a title, that title does not ship unreviewed.

Building the QA layer took longer than building the extraction. That ratio is normal, and anyone who tells you otherwise has not run one of these in production.

What happened

• The validation corpus grew to roughly 600 real titles before go-live. On the final out-of-sample batch, the cross-check caught every genuine odometer misread. Not most. Every one.
• The AI cost per title is about fifteen cents. The labor it replaces was the equivalent of most of a full-time position.
• The first fully unattended overnight cycle ran a 58-title batch with zero pipeline errors and had forms ready by morning. 24 of the 58 were flagged for a human look, which is the design working, not failing.
• An unexpected bonus: the system kept catching mistakes in the client's historical paperwork. Nine separate errors in their existing completed forms, plus mispaired records, surfaced during testing because the pipeline disagreed with them and turned out to be right.

Two months in production

Since go-live the pipeline has processed more than 2,000 titles across nearly 40 overnight batches. The review portal logs every action, so the human side of the job can now be measured.

• Review averages under half a minute per title, counting every title in the batch.
• About 60% of titles pass every check on their own and print with the batch. They need essentially no review time.
• The other 40% are flagged. A reviewer typically clears a flagged title in under 30 seconds, or about a minute each once everything around it is counted.
• A 100-title batch takes about 45 minutes to review. A full week comes to roughly 4 hours, in place of an estimated 20 to 30 hours of manual entry.

That is an estimated 80 to 85% cut in data-entry time. The “before” figure is the client's own estimate, not something we measured. Neither number counts the parts of the job that did not change: printing, handling paper, and the trip to file them.

What this means for your business

If your team spends hours transcribing structured documents, this class of system is mature enough to trust, provided it is built with a review lane and an independent cross-check, not as a black box. The interesting design work is not the AI reading the document. It is deciding, mechanically and provably, when the AI is not allowed to be trusted.

Client details are anonymized and some specifics are generalized to protect confidentiality. The numbers and the failure stories are real. If you want this kind of system for your back office, start with a roadmap session.

Updated 7 Oct 2026 with measured review times.