Skip to main content
AI file ingestion reads gig paperwork and proposes structured records: a timeline, a crew list with fees, ledger items, and leads. You review the proposals and select what to apply. Extracted records are applied only when you commit; uploaded files are attached to the gig immediately. It produces the same Universal Event Format (UEF) as the spreadsheet importer and partner feeds, so a file dropped on a gig and a row imported from a spreadsheet end up as the same kind of record.
CheckAI AI file ingestion is in Private Preview, available only to the existing beta tester group. Signing up, choosing a paid plan, upgrading, contacting support, or viewing a public demo does not grant access. Ordinary gig attachments are a separate workflow and do not require AI ingestion.

​
The flow

1

Open the gig's Files tab

Existing testers with an Owner, Admin, or Gig Coordinator role can open the enabled Files preview. Files are grouped into an ingestion batch for that gig; uploading creates a batch when needed.
2

Add files

Drop files or choose Choose files. Supported inputs include PDFs, DOCX documents, images, text files, spreadsheets, and saved emails. Uploading attaches the source files to the gig before you review extracted records.
3

Per-file analysis

CheckAI reads each file and extracts a set of proposals: run-of-show cues, crew members, ledger items, leads, and event details. A file moves from analyzing to ready for review.
4

Review proposals

For each file you see exactly what was found, with a preview: which rows are new versus which already exist (a crew member already on your roster, a venue already in your CRM). Check the proposed values and select only the records you want to apply. Leave incorrect or uncertain rows unselected.
5

Review combined files

A batch can combine proposals across files. Check overlapping or conflicting details, such as two different load-in times, against the original documents. Refresh a review if files or combined proposals change.
6

Commit what you chose

Commit the proposals you selected, by file or from the combined batch review. Commit selected applies those choices to the gig, so review everything first. An AI client uses the separate confirmation step described below.

​
What gets extracted

A single file can propose any of: Crew rows without a contact channel can’t become an invitation, so they’re skipped when you commit. The same rule the importer follows. Positions are matched to your organization’s positions the way imports match them. A position that matches none is kept in the crew record’s fee notes as “Position: …” so you can assign it later. A lead found without an email is left out of individual-file and batch reviews and their proposal counts, because it can’t be committed. This also applies when reviewing previously extracted files. A lead with no date, event type, or location takes the gig’s. The same lead (email and date) is added to a gig once, even when two files both list it, and an existing lead on that gig is never overwritten. The same person asking about another gig on the same date gets their own lead on that gig. Ledger items are written the way a UEF import writes them. A line with no status is paid if it’s income or an expense, and unpaid if it’s a payable or receivable. A status is read ignoring case and punctuation. Words such as paid, paid in full, fully paid, settled, received, payment received, cleared, complete and completed mark the line paid. Words such as unpaid, due, outstanding, open, pending, invoiced, overdue, not paid, partially paid and partial mark it unpaid. Any other status is saved as unpaid, and the line gets a warning that names the status. A payable marked paid is saved as an expense marked paid, and a receivable marked paid as income marked paid, with a warning. Its payment is recorded in Soundcheck, not on import. A line keeps its invoice or reference number as its ID, and a line with one is matched by it. A reference that repeats within one file keeps every line: the first keeps the reference, and each repeat gets its own key under the reserved uef:occurrence: prefix, numbered in file order across the whole file. Choosing only some rows to import does not change those keys, so a repeat committed on its own is matched the same way later. A reference that starts with uef:occurrence: is refused. A reference already on the gig with the same type, description, amount, date, currency and counterparty is a duplicate. The same reference with any of those different is refused and reported with both values, and nothing on the gig is changed. A currency must be an ISO 4217 code such as USD or EUR. A date must be a real day, in any format a spreadsheet date column accepts. A line with a currency or date it can’t read is refused and reported; it is never saved with another currency or the gig’s day. A line with no date or currency takes the gig’s. An account name in the file is used only when it matches an account of the right kind (income or expense); otherwise the line goes to the default account. The same line (type, description, amount, and date) is added to a gig once, even when two files both list it. You can’t commit files to an archived gig. Unarchive it first.

​
New vs. duplicate

Before you commit, the review summarizes new and matching records against data you already have. Matching records can reuse an existing crew member, venue, or customer. Review the matches carefully when re-ingesting an updated document, especially if names, contact details, or external identifiers have changed. Identity follows the same external-ID and natural-key rules described in Identity and safe re-imports.

​
Review before applying extracted records

In the web review, Commit selected is your approval to save the selected records and event fields. A combined review that has changed must be refreshed before it can be committed. Check the result afterward: if something goes wrong partway through, some selected records may be applied while others still need attention. When using CheckAI or an external AI client, the MCP commit tools require an explicit confirmation of the proposed changes. If the selected data or merge changes, review the updated proposal and confirm it again. See the MCP security model.

​
From an AI client

AI file ingestion also has MCP tools for authorized external clients. This integration has its own authentication and available tools; connecting an MCP client does not grant CheckAI Private Preview access. The ingest_gig_files prompt walks the whole batch, review, confirm, and commit loop. See the available tools on the MCP server page.

CheckAI overview

The assistant behind ingestion.

Spreadsheet imports

The other way data gets into Soundcheck.

Universal Event Format

The shape every ingestion produces.