Skip to main content

How AI Descriptions Work in iCat

Accuracy Expectations, Workflow Rules, and Review Best Practices

Support MD avatar
Written by Support MD
Updated this week

AI-generated descriptions in iCat are designed to help contents teams work faster without compromising billing accuracy.

When AI Descriptions Are Applied

AI descriptions in iCat are never applied over human-entered data.

They are added only when:

  • No description is entered during packout, or

  • An admin intentionally runs AI descriptions post-packout

Manually entered descriptions always take precedence. AI is used to fill gaps, not undo field work — keeping your team fully in control while still benefiting from AI-assisted efficiency.


Is AI Expected to Be Perfect?

Yes — when it matters.

Accuracy expectations in iCat are determined by how a description is used.
If a description impacts estimating, billing, or reporting, AI is held to a near-perfect standard.


Where AI Must Be Accurate (and Is Held to That Standard)

Billable & Accuracy-Critical Items

This category includes any item that is not billed via flat-rate box pricing, including:

  • Single, non-boxed items

  • Individually inventoried items placed into a regular box but billed separately

  • Non-Salvage inventory (items requiring reliable identification for documentation, valuation, or workflow clarity)

For all items in this category:

  • AI descriptions are expected to be accurate and dependable

  • With a clear photo of a single item, current accuracy is measured at 99%+

  • These descriptions directly impact estimating, billing, reporting, and documentation

  • Precision is mandatory — not optional


Core Rule (Summary)

If an item is not in a cleaning box and not inventoried for storage-only / Return-As-Is (flat-rate by box type), the description must be accurate — and it is.


Where AI Does Not Need to Be Perfect

Cleaning Boxes

For cleaning boxes, AI is intentionally used in a more approximate way.

The goal is to:

  • Provide a general understanding of what’s in the box

  • Improve searchability and packback context

  • Support documentation — not billing

Important clarifications:

  • Billing is based solely on the box type selected

  • Box descriptions have zero impact on billing

  • Descriptions are informational only

  • Photos remain the primary source of record and evidence

Because these descriptions do not affect billing, item-level precision is not required in this workflow.


Recommended Best Practice: Human Review

While AI descriptions in iCat are held to a near-perfect standard where accuracy matters, human review is still a recommended best practice.

This is not because AI is unreliable — but because input quality matters.

Common factors that can affect results include:

  • Blurry or poorly framed photos

  • Multiple items captured in a single image

  • Obstructed or partially visible contents

For this reason, iCat workflows are designed with the expectation that teams will review AI-generated descriptions as part of a standard packout or QA process, especially for billable items.

A quick review helps ensure:

  • Descriptions correctly reflect what was actually inventoried

  • Edge cases caused by photo quality are caught early

  • Estimating and billing remain accurate and defensible

AI accelerates the work — it does not replace accountability.

Did this answer your question?