RegenITAD · listing engine
Planned MVP — not live

Messy asset inputs in.
Reviewed listing drafts out.

The RegenITAD AI Asset Listing Pipeline turns photos, asset labels, and operator notes into normalized, reviewed, marketplace-ready listing drafts. Built for the ITAD, recommerce, and surplus IT equipment world — where every server, switch, and workstation arrives in a slightly different state and has to be listed accurately without burning a human on each one.

Why this exists

IT asset disposition and recommerce are bottlenecked at the listing desk. Each unit has factory specs, current configuration, photos, condition notes, and a service tag that may or may not still be legible. Producing a clean marketplace listing from that mess is slow, repetitive, and expensive — and inconsistent listings cost real money on resale.

The Listing Engine attacks the slow part: extraction, normalization, and draft synthesis. It does not attack the judgment part — a human still approves every draft before it can be exported to eBay, the RegenITAD storefront, or a CSV pipeline.

Domain shape

  • ITAD / recommerce / surplus IT equipment workflow
  • Servers, switches, workstations, laptops, daughter cards
  • Factory configuration vs current configuration matters
  • Multiple export targets: marketplace, storefront, CSV
  • Human review is required, not optional

The pipeline

Seven explicit stages. The AI handles extraction and synthesis; the accountable human handles judgment, exception review, and final action.

End-to-end story

Asset photos / labels / notes OCR / service tag extraction Dell/vendor enrichment factory-vs-current configuration review listing draft generation human review export to eBay / RegenITAD storefront / CSV.

01

Asset capture

Photos of the chassis, asset labels, and operator notes. Sample / synthetic inputs only — no real customer asset tags.

02

OCR & tag extraction

Optical character recognition pulls service tag / model / serial candidates out of the label photos and notes.

03

Vendor enrichment

Tag candidates are looked up against vendor specs (e.g. Dell) to recover the factory configuration and product family.

04

Config diff

Factory-vs-current comparison highlights what was changed: RAM, drives, CPU, GPU, daughter cards — the listing differentiators.

05

Draft generation

AI composes a normalized listing draft: title, condition, spec table, photos selection, and suggested price band.

06

Human review gate

An operator approves, edits, or rejects the draft before anything leaves the system. Nothing publishes without sign-off.

07

Export targets

Approved drafts can be exported to eBay, the RegenITAD storefront, or CSV for downstream tools. Export, not auto-publish.

What the AI actually does

The AI is doing extraction, lookup, normalization, and draft synthesis — not pricing decisions, not publishing, not selling.

Photo / OCR pipeline

Image preprocessing plus OCR to lift text off asset labels and operator notes, with confidence thresholds on each field.

Vendor lookup & normalization

Service tag / part number candidates are normalized and resolved against vendor metadata to attach a factory spec.

Configuration diff

Structured comparison of factory-vs-current configuration so the listing reflects reality, not a stale datasheet.

Listing draft synthesis

Templated draft generation that pulls from the normalized spec, the diff, and the operator notes — review-ready, not publish-ready.

Safety boundary

Planned MVP — what the system will and will not do.

  • Sample / synthetic assets only. No real customer asset tags, serials, or service tags are used or displayed.
  • No marketplace publishing in MVP — humans review and export drafts, the system does not auto-publish to eBay or storefronts.
  • Human-review gate is part of the pipeline, not bolted on. Closeout is blocked until a reviewer signs off.
  • Vendor enrichment runs against public spec data and operator notes; no customer-confidential records are wired in.
Planned MVPTracker access available on request

Implementation tracker

The listing-engine work is tracked in the private RegenITAD implementation repository. The page you are reading is the public proof surface; the tracker is where the work itself happens. Tracker access is private — reviewers can request access via the demo-access form.

Request tracker access
Future protected demo

Want to walk through the listing engine?

When the protected MVP is ready, allow-listed reviewers will be able to step through the pipeline on sample / synthetic assets. Register interest now — operator-reviewed, out-of-band approval, no marketing list.

Available for evaluation

Work with me on workflow automation

I take on ITAD, recommerce, ops-heavy listing pipelines, and AI-assisted document generation through scoped work with concrete artifacts.