Extensions
AI Hiring Compliance Pack
Adds human-review checkpoints, decision logging, draft candidate notices and an auditor-ready data export to AI Resume Screening or AI Interview Intelligence; helps…
Extensions
Historical Document Backlog Processing
Your archive of past documents runs through the extraction product you already bought, checked and loaded.
A new extraction product handles documents from go-live, but the old pile still needs typing in. This add-on runs your existing archive through the Upmark document or invoice product you already use. Up to five thousand pages are processed and a sample is checked by hand, along with every low-confidence item. Results load into your system or a CSV.
- Past documents processed with your configured document types
- Random sample and low-confidence items checked by hand
- Exceptions list with reasons and an accuracy summary
$1,500
10–20 working days
What it is
Historical Document Backlog Processing is a one-off run of your existing documents through your Upmark document or invoice extraction product. Up to five thousand pages are processed with your configured document types. A random sample and every low-confidence item are checked by hand, and results load into your system or a validated CSV.
Why businesses buy it
Most teams want their history in the new system too, for reporting, audits and supplier records. Typing it in by hand is slow and error-prone, and it rarely gets finished. This add-on reuses the extraction rules you already approved, so the old documents are read the same way as the new ones. A hand check shows how reliable the results are before you rely on them. The written scope sets the page limit, checks, revision round and support in advance.
Problems this solves
Your new system only handles documents from go-live onwards
The add-on runs up to five thousand pages of historical documents through your existing product. It uses the document types and fields already configured, so old and new records match.
Typing old invoices and forms in by hand never gets finished
The batch run loads results into your destination system or a validated CSV. Staff only deal with the items on the exceptions list, each with its reason.
Nobody knows whether bulk-processed data can be trusted
A person checks a random sample and every low-confidence item by hand. The accuracy summary shows how the sample performed before you rely on the data.
A good fit if
Your new extraction product handles documents from go-live, but years of older files still sit unprocessed.
Staff would otherwise type data from past invoices, forms or contracts in by hand.
Your archive is already digital, as PDFs, images or emails.
The old documents match the document types your product already handles.
Not the right fit if
You do not yet run an Upmark document or invoice product. Start with AI Document Extraction or AI Invoice Intelligence.
Your archive is on paper or handwritten. It needs scanning or manual entry first, which this add-on does not cover.
The old documents are a different type from your configured ones. Add a new document type to the parent product through a Change Quote first.
You want to search and ask questions across past contracts, not extract fields. See AI Contract Intelligence.
How it works
- You share the parent product's order reference and put the historical files in a shared folder you control.
- We batch-process up to five thousand pages using the document types and fields already set up in your product.
- A person checks a random sample of the results by hand, plus every item the product marked as low confidence.
- Results load into your destination system, or arrive as a validated CSV if you prefer.
- You get an exceptions list with reasons and an accuracy summary for the sample.
- You review the results, ask for changes in one revision round, and get a short written walkthrough.
What you provide within 72 hours
Your portal lists these the moment you order. Delivery starts the working day after everything is in and your access is checked.
- The order reference of your Upmark document or invoice product.
- Historical documents as digital files in a shared folder you control.
- Access to the destination system, shared by invite or a key you control, never by email.
- Your business information (company basics, hours and brand voice), shared during intake.
- One decision-maker who approves the results.
The 72-hour rule. Send everything on your checklist, including access, within 72 hours of ordering so your build starts on time.
Timeline
Start
Order reference, file folder and destination access confirmed.
Week 1
Files reviewed against the configured document types; out-of-scope pages flagged.
Week 2
Batch run of the backlog through your product.
Week 2
Random sample and low-confidence items checked by hand; exceptions listed.
Week 2
Your review and one revision round.
Go-live
Results loaded or CSV delivered, with accuracy summary and walkthrough.
| Option | Delivery time | Price |
|---|---|---|
| Standard | 10–20 working days | $1,500 |
What you receive at handover
- A batch run of up to five thousand pages using your configured document types
- A hand check of a random sample and all low-confidence items
- Results loaded into the destination system or delivered as a CSV
- An exceptions list with reasons and an accuracy summary for the sample
- A a short written walkthrough and 14 days of support
Expected benefits and how to measure them
- Can bring past records into the same system as new ones
- Helps avoid weeks of manual data entry from old files
- Pages processed versus exceptions
- Compare the number of pages loaded with the number on the exceptions list after the run.
- Sample accuracy
- Read the accuracy summary for the hand-checked sample, and spot-check a few records yourself.
- Staff hours saved on data entry
- Time how long staff take to key in a small batch by hand, then compare with the exceptions they handle after.
Before you start: Count the pages in the archive and note how long staff take to enter a typical document by hand.
These are things to track, not promised results. Outcomes depend on your data, your team and your market.
Running costs
There are no running costs after the run. The run itself uses AI model usage in your own account, and the amount depends on the number of pages. We charge once for the add-on and nothing per page within the limit.
Ask about running costsCustom requirements
Paper scanning, handwritten documents or pages outside your configured document types are not covered. Add a note at checkout for a written Change Quote. Larger or bespoke projects go through the Custom AI request form instead.
Request a custom versionKeep it running
This is a one-off run, so nothing needs to keep running afterwards. New documents keep flowing through the parent product as usual, under its own maintenance if you have it. Questions about the results are covered during the 14 days of support.
Adds $75/mo to maintenance
Covered by the main product's maintenance, which grows by this amount.
How maintenance worksHow this compares
| Do it yourself with a SaaS tool | Hire a custom agency | Upmark Labs | |
|---|---|---|---|
| Price | Doing it yourself costs staff hours per page | A data-entry bureau usually charges per page or per hour | Ours is one fixed price for the page limit |
| Method | Typing data in by hand | Outsourced keying by a bureau | Ours uses the extraction product you already run |
| Checking | Whatever time allows | Depends on the bureau | Ours hand-checks a random sample and every low-confidence item |
| Problem items | Found later, one by one | Often returned without notes | Ours lists every exception with a reason |
Questions about this product
Which products can the backlog run be added to?
Any live or same-order Upmark document operations, invoice intelligence, accounts payable or financial document product. Common examples are AI Document Extraction, AI Invoice Intelligence and AI Accounts Payable Agent. The run uses the document types and fields already set up in that product, so it needs to be configured first.
What happens if our archive is bigger than the page limit?
Add the backlog run again for each further block of pages. Each run follows the same steps: batch processing, a hand check, loading and an exceptions list. You can split the archive by year or supplier so each run is easy to check.
How accurate is the historical document backlog run?
It depends on the quality of your files and how closely they match your configured document types. That is why a person checks a random sample and every low-confidence item. You get an accuracy summary for the sample, so you can judge the results before relying on them.
Can the backlog run process paper or handwritten documents?
No. The run handles digital files only, such as PDFs, images and emails. Paper documents need scanning first, and handwritten documents are excluded because extraction is unreliable on them. If part of your archive is on paper, scan it before ordering or leave it out of the run.
What happens to documents the extraction cannot read?
They go on the exceptions list, each with a reason, such as a poor scan or an unexpected layout. Data entry for those documents is not included. Your team can enter them by hand, or you can ask for a Change Quote if many share the same problem.
Where do the backlog results end up after the run?
In the destination system your parent product already writes to, such as your accounting system or document store. If you prefer, we deliver a validated CSV instead, so you can review or import it yourself. Either way, the exceptions list and accuracy summary come with it.
Clear the archive your new system was never going to touch.
Add Historical Document Backlog Processing and your past documents are processed, sample-checked and loaded, with every exception listed and explained.
Works well with
Extensions
E-Invoice Intake
Adds XRechnung, ZUGFeRD/Factur-X and Peppol e-invoice intake to AI Invoice Intelligence or AI Accounts Payable Agent, validated and sent through your existing…
Extensions
Offline Export and Backup
Adds scheduled export and backup of an existing product's data. Requires an existing or simultaneously purchased Upmark Labs product.