1,250 lots.
Typed by hand.
Now filled from photographs — 400+ hours off the client’s desk. Scroll the whole story.
Weeks of typing before
a single item went live.
Every auction meant 1,250 lots entered by hand. Photograph, edit, renumber, then type a title, a description, a start price and an estimate range — one lot at a time, to lot 1,250.
- 01
Ingestion
32,000 past items — titles, descriptions, estimates and realized prices — ingested into a hosted backend as the model’s reference set.
- 02
The engine
Photographs in, complete listing rows out. Repeat items pull their own history; new items get identified and described from the image.
- 03
Review triage
Confidence-scored output. Only uncertain items are flagged for a human; everything else passes straight through.
- 04
The handoff
The filled template uploads straight to LiveAuctioneers, and existing automation cross-posts to BidSpirit and the direct bidding site.
Every one previously typed by hand · First live auction figures [PENDING]
Realized prices sat below estimates — systematic inflation, invisible until the data was in one place.
The kind of thing nobody sees until software reads everything at once
“Cataloguing a sale used to take me four to eight hours a day for at least a couple of weeks. This app cut that two weeks of work into less than a full day. I never had to repeat myself with Umer. He delivers very quickly, and what comes back is beyond what I asked for.”
Tell us the
number. We’ll tell
you if we can.
We’ll tell you if we can hit it. If we can’t, we’ll say that too — and point you at whoever can.