In business document management many checks are repetitive: does this document belong to the right person? When does it expire? Who still has to send it? These are simple questions, but multiplied across hundreds of documents and parties they become a full-time job.
Artificial intelligence can take on part of these checks. In this article we look at what this means in practice, how master data checks and expiry date calculation work in ThorIA, and where human oversight remains essential.
What applying AI to documents means
In a document management system, AI can carry out five main activities:
- Reading: it analyses the content of PDFs, Excel files and images uploaded to the system.
- Master data check: it checks that the details in the document match the party it is linked to.
- Expiry date calculation: it determines the document's validity date.
- Reminders: it sends automatic emails to suppliers and employees about missing or expiring documents.
- Natural-language questions: it answers requests such as "which documents expire this month?".
The aim is not to replace the people who manage documents, but to free them from repetitive checks so they can focus on exceptions and decisions.
How the master data check works
When a document is uploaded, the system links it to a party: a company or a person. The AI reads the document and compares the details it contains, such as the company name or the person's name, with those of the registered party.
If the details match, the document continues along its workflow. If differences emerge, the system flags them: it could be an upload error, a document referring to another company in the same group, or master data that needs updating. In every case, an operator decides how to proceed.
This check is particularly useful in supplier and subcontractor qualification, where documents arrive from many different parties and a document linked to the wrong supplier can give a false impression of compliance.
How ThorIA determines expiry dates
Not all documents state an explicit expiry date. Validity often depends on the type of document and on rules set elsewhere. ThorIA follows a multi-step process:
- Semantic database: first it consults its own semantic database, which maps document types to known validity rules.
- Official sources: if the type is not present, it checks official sources such as, in Italy, the Gazzetta Ufficiale, State-Regions Agreements and Normattiva.
- Human validation: the calculated expiry date is always validated by a human operator before it is recorded.
Once validated, the expiry date feeds automatic alerts and reminders: the system notifies the people managing the documents in good time and writes to the supplier or employee concerned when it is time to send the updated version.
Limits and human oversight
AI is a powerful tool, but not an infallible one. A barely legible document, an unusual type or a special case can lead to results that need correcting. That is why human oversight remains essential:
- Expiry dates must be validated by an operator before they take effect.
- Inconsistency flags must be reviewed by a person, who decides how to proceed.
- Natural-language answers must be checked when they are used for important decisions.
- The rules the AI works on must be reviewed periodically, because regulations and processes change.
The most effective model is one where the AI proposes and the person decides: the AI reduces repetitive work, while the person keeps responsibility for the choices.
Data is not used to train models
Applying AI to business documents means entrusting it with information that is often confidential: contracts, staff data, supplier documents. With ThorIA data is hosted in the EU (Milan, Italy) and is never used to train models. ThorIA is built into the IntraThor DMS document management system, where it works on the platform's own documents, permissions and workflows.
For a broader overview, read our 10 AI use cases for SMEs and the meaning of DMS in the glossary.
A practical example: qualifying a new supplier
A new supplier uploads its documents. Instead of opening them one by one, the office receives an initial automated check: the master data is compared with the records on file and the relevant dates are proposed as expiry dates. The operator checks the result, corrects it if necessary and approves it.
Manual work does not disappear, but it shifts: less time copying data, more time checking the cases that need attention.
Checklist for introducing automated checks
- Choose one type of document to start with, for example supplier documents.
- Decide who approves the results proposed by the AI.
- Define what to do when a check flags an anomaly.
- Review the first results carefully before extending use to other documents.
- Collect the cases where the proposal was corrected, to understand where more attention is needed.
Two useful terms
Master data check: comparing the details in a document with those registered in the archive. Human oversight: the step in which a person reviews and approves the result before it becomes final.
F.A.Q. - Frequently asked questions
What does AI do in document management?
It reads documents, checks master data, calculates expiry dates, sends reminders and answers questions in natural language.
How does the master data check work?
The AI compares data such as the company or person name in the document with that of the party it is associated with, and flags any inconsistencies.
How does ThorIA calculate expiry dates?
It first consults its own semantic database and, if needed, official sources such as the Italian Official Gazette, State-Regions Agreements and Normattiva.
Are expiry dates calculated by AI reliable?
They are a proposal: every expiry date is validated by a human operator before being recorded.
Are documents used to train the AI?
No. With ThorIA, data is never used to train models and is hosted in the EU (AWS Milan, Italy).
Which documents should I start with?
With a frequent, repetitive document type, such as supplier documents, so you can assess the results on a concrete case before extending the use.
What happens if the AI flags an anomaly?
The flag must be reviewed by a person, who decides whether to correct the data, request a new document or approve it anyway.
Request a demo
Find out how ThorIA checks documents and calculates expiry dates in IntraThor DMS.
General information only, not legal advice: always check the current legislation.