Data & privacy
Your documents.
Your decisions.
Less repeated work. Clear checks. You stay in control of what becomes a business record.
Updated 3 October 2026
The intended workflow
A draft first.
A decision from you.
- 01
Source document
Start with a supplier invoice.
- 02
Editable draft
Read details into clear fields.
- 03
Your review
Resolve questions and mistakes.
- 04
Your approval
Only then update the records.
Reading a document alone does not change stock, create an amount owed or send a payment.
01 / Purpose
Business information deserves careful handling.
TotalUp is business software in development for small-business owners. It is designed to bring purchases, products, stock and supplier payments together, with clearer everyday workflows and fewer repeated steps.
Supplier invoices can contain business names, contact details, negotiated prices, product information and other commercially confidential information. Our approach is to use that information for the task the user intends, keep important decisions visible, and establish the required data controls before enabling external AI processing.
02 / Today
Local reading.
A manual path when needed.
The development application can read embedded text from supported digital PDFs without sending it to an external AI service. It uses recognised fields and previously confirmed supplier items to help prepare an invoice.
Its current reader supports a limited set of document structures. When an invoice cannot be read reliably, manual entry remains available. Scanned pages and photographs are not currently processed by AI.
03 / Proposed AI use
Extraction with a specific purpose.
We are evaluating the OpenAI API to help extract supplier invoice details into structured draft fields: the supplier, invoice reference, dates, line items, quantities, prices and stated tax amounts.
The intended users are business owners and the people they authorise to manage purchases and review invoices. The purpose is to reduce repetitive typing while keeping those users in control.
The proposed initial workflow uses the Responses API for a bounded extraction request after local text reading and confirmed-item lookup. Only information necessary for the unresolved extraction task would be submitted. The model and its supported features must be verified before release.
File Search is not planned. Provider file storage is not part of the proposed initial text workflow; any later use would require its own review.
04 / Your control
A draft is not
a transaction.
Extracted information stays editable. Users can check the source, resolve missing information and correct mistakes before explicitly approving and posting an invoice.
AI would not independently create suppliers or products, confirm item relationships, authorise payments or decide tax recovery. Calculations and posting follow application rules. Stock and the amount owed change only through the applicable confirmed posting workflow. Posting a purchase invoice does not send a payment.
05 / Before external processing
Requirements first.
Activation later.
Before enabling external AI processing, we require approval and verification of Zero Data Retention and European regional processing for the actual provider project, model, endpoint and features used.
These are requirements we are pursuing, not approvals or protections we claim to have in place. Zero Data Retention is subject to provider eligibility review and applicable limitations. Scanned invoices and photographs would need a separate review of image support, regional processing and retention exceptions.
06 / Retention
Provider processing.
Your business records.
Provider retention controls concern information sent to the AI service. They are separate from the invoice records and source documents retained in TotalUp.
The development workflow retains accepted source documents privately so users can refer back to them. Cancelling an invoice or discarding a draft does not automatically delete its accepted original. Retention and deletion arrangements for the public service will be defined before launch; we do not promise that all business records disappear after extraction.
Public document uploads are not enabled. Production document protection, release checks and the required provider controls must be completed before customer processing begins.
A clear distinction
Built in development.
Planned for the product.
The product tour includes future workflows. Here is where each area stands.
| Area | Current position |
|---|---|
| Products, suppliers & stock | DevelopmentImplemented in the development application, including stock history and physical counts. |
| Purchase invoices & supplier payments | DevelopmentDrafts, posting, payment records and allocations. No money transfer service. |
| Digital invoice reading | DevelopmentLimited local PDF text extraction, guided questions and manual fallback. |
| AI invoice extraction | ProposedExternal processing is disabled pending approval and verification. |
| Invoice photographs & phone stock scanning | ConceptRequire further development and review. Image processing is separately gated. |
| Customer invoices, selling prices & accounts | PlannedFuture features illustrated in the product tour. |
| Till/POS connection | ConceptNo live integration or device compatibility is claimed. |
| Public availability | Pre-productionNo public signup, invoice uploads or live customer processing on this website. |
For the provider’s current terms and feature limitations, see OpenAI’s API data controls .