Skip to main content

AI Configurator reference

This reference documents the boundaries of the AI Configurator chat surface in Estimate Setup: what each chat action does, what it cannot do, and the…

S
Written by Styrbjörn Holmberg

Scope

This reference documents the boundaries of the AI Configurator chat surface in Estimate Setup: what each chat action does, what it cannot do, and the observable signals an operator can use to confirm each state. Guidance Settings (Category, Product, Suggestions, Knowledge, Communication Style) are documented in the Guidance reference, not here.

Chat dock actions

Action

What it controls

When to use it

What you will see

AI Configurator (collapsed pill)

Expands the chat dock back to full height after it has been minimised.

When you collapsed the chat to work in a table and want to resume the conversation.

The dock expands; the prior conversation is preserved row by row.

New conversation

Clears the current chat history and starts a fresh extraction context.

When you start a new import session and do not want prior file context to influence parsing.

The chat clears; processing state resets; previously-staged rows in Pending Changes are not affected.

Drop files here

Accepts a file drag-and-drop into the chat for parsing.

When you have CSV, XLSX, or PDF source data to import.

A file pill appears in the chat with a parsing progress strip; the strip names each stage as it completes.

Stage changes

Pushes extracted rows from the chat into Pending Changes.

When the extraction summary looks correct and you want to review the staged result in the sidebar.

The Pending Changes count in the sidebar increases by the number of staged rows; the chat closes the extraction.

Boundaries: what AI Configurator does

Capability

Notes

Parse machines, substrates, and finishing equipment from CSV, XLSX, or PDF

Files must contain the field values the importer needs; it does not invent numbers.

Match imported rows against existing setup to produce create / update / delete operations

The match key is the row name. Renamed rows are imported as create, not update.

Stage all changes in Pending Changes

No row is applied to live setup directly from chat.

Persist conversation state per shop

The dock remembers chat content across navigation within Estimate Setup.

Continue parsing while you navigate elsewhere

The progress strip remains accurate; the chat dock surfaces a notification when extraction completes.

Boundaries: what AI Configurator does not do

Limitation

Workaround

Guarantee 100% field accuracy without review

Always inspect new, updated, and removed rows before staging. The chat surfaces a row-level summary for this purpose.

Apply changes without operator confirmation

Open Pending Changes and select Apply Changes. There is no auto-apply path from chat.

Design product categories or production-step sequences

Categories and production steps are configured per category, not via chat. Use Products & Product Categories in the sidebar.

Fix a broken quote that already exists

The chat changes setup. To recalculate an existing quote, open the quote and select Recalculate after applying staged changes.

Invent missing values

If a press lacks a click rate in the source, the imported row lacks it too. Edit the row before applying, or routing will exclude it.

Replace manual edits

Manual edits to Print Machines, Substrates, and Finishing Machines tables also flow into Pending Changes alongside chat-staged rows. Treat the queue as the global staging area.

File handling

Behaviour

Detail

Accepted formats

CSV, XLSX, PDF. Other formats are rejected at upload time with an error in the chat.

File size handling

Large files are processed in named batches. The progress strip identifies which batch is active so you can track which subset has finished extracting.

Background processing

Extraction continues while you navigate to another page in Estimate Setup. The chat dock surfaces the completion event when you return.

Multiple files per conversation

The chat accepts more than one file per conversation. Each file is processed in sequence; the summary at the end aggregates rows from all files.

Observable signals

Signal

Meaning

Pending Changes count increased after Stage changes

The extraction succeeded and rows are queued for review.

Pending Changes count unchanged after Stage changes

Either the extraction failed or every row was rejected by validation. Re-open the chat and inspect the extraction log.

Chat dock shows an error icon on a row

The importer failed to parse or validate that row. Open the row to see the specific error.

Toast Change staged. Review in Pending Changes.

A staged row was written to the queue successfully.

Apply Changes dialog confirms apply

Staged rows are publishing to live setup. The count returns to zero on success.

Related articles

Did this answer your question?