Documents — Upload, Review, Approve

Common flags and what to do with them

New vendor. First time Tally has seen this vendor. Confirm the name, category and account mapping. Suggestions become available for a vendor once its live-review conditions are met — see Vendor Autonomy Journey.
Possible duplicate. Tally scored this document against earlier documents on vendor, amount, date, document number, card tail and line items, and the combined score crossed the duplicate threshold. Review both, approve the one you want to keep, reject the other. The comparison is re-run on each new document against the documents already stored, so a genuinely identical document will be flagged again — approving one does not switch the check off for that pattern.
Possible anomaly. The amount sits outside the range Tally has recorded for this vendor. The flag names the comparison it made ("Stripe usually $12–15/month, this is $250"). You decide.
Missing field. Something Tally needed wasn't extractable from the document. Fill it in and approve.
Low confidence. The image was hard to read. Verify the extracted fields and approve, correct, or reject.

Common issues

Problem: A receipt won't extract or extracts wrong fields.
Why: handwriting, glare, or unusual receipt layouts.
What to do: correct the fields manually and approve. The correction is recorded against that vendor and is used on later documents from it.
Problem: The same invoice keeps appearing as a duplicate.
Why: the vendor sent multiple copies, or the document was uploaded more than once. Each newly-arrived copy is scored against what is already stored, so each one is flagged on arrival.
What to do: approve the copy you are keeping and reject the others.
Problem: A vendor I've never used appeared on the list.
Why: it was in your CSV import, your forwarded inbox, or your bank feed.
What to do: review the document. If it's a one-off, approve or reject as needed. If it's not yours, reject it and the vendor won't be remembered.
Problem: Tally flagged an amount as anomalous but it's correct.
Why: the charge is outside the range recorded for this vendor (annual renewal, refund reversal, a real price change).
What to do: approve it. Your decision is recorded against that vendor and the vendor's baselines are recomputed from the recorded history, so the range it is compared against reflects what you have approved. Whether a specific future document is flagged depends on that recomputed range and on the other checks.

If you're a bookkeeper reviewing client work

The document list and review actions work the same — the difference is scope. From the client drill-down view, you see only that client's documents. The Bookkeeper Command Center shows you which clients have queues that need attention so you can prioritize across your portfolio. See Bookkeeper workspace.

When you correct a document for a client, two things happen:

That client's Tally learns from the correction (per-client memory).
If the correction is one your firm makes consistently across clients (e.g., "Adobe Creative Cloud always maps to Software Subscriptions"), it can become firm-level memory on Business plans — see How Tally learns.

If you're an SMB owner reviewing your own documents

Same actions, smaller scope — you only see your own business. A new workspace has no vendor history, so nothing is pre-filled at first. As live reviews accumulate on a vendor, that vendor becomes eligible for suggestions and then for automatic handling, under the conditions in Vendor Autonomy Journey.

Just upload your receipts and invoices — Tally tells you what needs attention

Yes — you can simply upload your receipts and invoices (drag and drop, single files or batches), and Tally extracts each one and tells you exactly what needs your attention: it surfaces a review queue where every flag comes with a clear reason (new vendor, unusual amount, possible duplicate, missing field). This everyday upload-and-review workflow is different from importing a CSV of historical records (see Importing CSV history).

Reviewing

Open a document to see Tally's full extraction. The detail panel shows the source PDF/image alongside the structured data. Three actions:

Approve — confirm Tally's extraction is correct. The document moves to Approved. If the document is part of a flagged pair (possible duplicate), you can override the flag with a note.
Correct — edit any field inline (vendor, amount, date, category, line items) and then approve. Corrections are recorded against that vendor and feed the vendor's stored pattern and its approval rate.
Reject — if the document is wrong, irrelevant, or shouldn't be in your books at all. Use this for accidental uploads or misidentified items.

The status taxonomy

Each document gets a status that tells you what to do next:

Pending — Tally is still extracting. Nothing for you to do yet.
Needs Attention — extracted, but Tally flagged something you should look at (missing field, low confidence, possible duplicate, possible anomaly, new vendor).
Suggested — Tally has a suggestion (often based on your history) that she'd like you to confirm with one click.
Approved — you've approved this document. Tally remembers your decision.
Auto-approved — Tally handled it on her own (only happens for vendors where she's earned that level of autonomy).
Rejected — you've rejected the document. It's archived, not deleted.

What's next

How Tally learns — why your reviews matter and how Tally earns autonomy over time
Connect QuickBooks — push approved documents through to your books
Bank feed — automatic transaction import
Troubleshooting — when something doesn't work

What Tally extracts

For each document, Tally captures:

Vendor name (and remembers your canonical spelling — "AWS" / "Amazon Web Services" become one vendor over time).
Amount, currency, tax, subtotal.
Date (invoice date, due date, document date).
Category (suggested from your history; you can correct).
Line items (when present and clear).
Confidence — Tally's own assessment of how clean the extraction was.

What you'll do

Upload documents (drag and drop, CSV import, email forwarding, or bank feed).
Read the statuses Tally assigns and understand which need your attention.
Review and approve, correct, or reject documents.
Handle duplicates, anomalies, and missing fields.
Find documents fast with search and filters.

Where documents come from

Tally accepts documents from four sources:

Direct upload — drag PDFs, images, or scans into the upload area on the Financials page. Single files or batches.
CSV / history import — upload a CSV exported from QuickBooks, Xero, or another system. Tally treats these as historical context (see How Tally learns for why this matters).
Email forwarding (Inbox) — if you have the Inbox feature, forward invoices from your email and Tally extracts them automatically. See Inbox.
Bank feed — a connected bank syncs transactions as reconciliation records. They are matched against your documents; an unmatched one becomes a document only when you explicitly import it as an expense. See Bank feed.