DOCUMENT OPS
Low-confidence receipt OCR to human review queue
Routes receipts whose extracted fields fall below a confidence threshold into a Slack review queue and writes corrected values back to BigQuery.
How it runs
The automated pipeline, trigger to output.
- TriggerNew receipt added to Dropbox folderDropbox
- ActionExtract fields and confidence with Hugging FaceHugging Face
- LogicBranch on confidence vs. threshold
- ActionSend low-confidence receipt to Slack review queueSlack
- OutputWrite final values to BigQuery expense tableBigQuery
What it does
Splits receipt processing into a fast lane and a review lane. Each scanned receipt is extracted with a Hugging Face model that also returns per-field confidence; clean extractions go straight to BigQuery, while anything ambiguous is sent to a Slack review queue with the receipt image so a person can correct it.
When to use it
Use it when crumpled, handwritten, or low-light receipts produce unreliable OCR and you want automation without silently saving wrong numbers. It gives finance a controlled queue instead of trusting every extraction blindly.
How it works
- 1A new receipt in the Dropbox folder triggers the run.
- 2The Hugging Face model extracts fields and reports a confidence score for the total and date.
- 3A logic step compares confidence against the threshold.
- 4High-confidence receipts are inserted directly into BigQuery.
- 5Low-confidence receipts post to a Slack review channel with the image; the reviewer's corrected values are written back to BigQuery, completing the record.
Set it up
What you configure once, before turning it on.
- 1Connect DropboxFiles and folders.
- 2Connect Hugging FaceModels, datasets, spaces — the open-source hub.
- 3Connect BigQueryDatasets, queries, schemas.
- 4Connect SlackChannels, DMs, threads, mentions.
- 5Set each agent's modelWe leave models unset so you pick the tier — fast + cheap, or top-quality.
- 6Tune it to your dataEdit the prompts, filters, and field mappings so it matches how your team works.
- 7Test, then turn it onRun once against a sample, confirm the output, then enable the trigger.
More Document Ops workflows
Narrate new Dropbox PDFs into MP3 audio versions
When a PDF lands in a watched Dropbox folder, extract its text and generate an ElevenLabs voice narration.
On-demand PDF narration via webhook with emailed audio link
Accepts a PDF URL through a webhook, generates an ElevenLabs narration with the requested voice, stores the MP3, and emails the requester a download link.
Triage emailed contract redlines and route by risk
When a counterparty emails a redlined contract, extracts the attachment, diffs clauses against approved templates.
Batch-narrate a Google Drive PDF folder in multiple languages
On a schedule, finds PDFs in a Google Drive folder that lack audio, then generates ElevenLabs narrations in each configured language and files them into per-language subfolders…
Executed Contract Exhibit & Initials Completeness Gate
When a signed contract lands in a Dropbox intake folder, verify every required exhibit, schedule, and initialed page is present.
Draft a negotiation brief from contract clause deviations
An agent reviews a contract against approved templates, researches each deviation.
Run it inside a business
This workflow drops into a full company template. Import the org, and this is one of the playbooks its agents run.
Run this workflow in your colony.
14-day trial. No DevOps. No Sales call. Provisioned in under a minute.