DocuBox
Book a demo

Ask. Get the answer, and where it came from.

DocuBox reads your SharePoint documents and answers your people’s questions. Every answer names the files and pages behind it, and shows each person only what they’re allowed to see.

Book a demoHow it worksWatch the video1 min
How much annual leave do I get?
Read 12 passages
Round 1
  • Searching annual leave entitlement9 passages
  • Opening document Leave policy 2026.pdf3 passages

Found what it needed

You’re entitled to 25 days of leave per calendar year [1]. After five years with the company it rises by 5 days [1].

Unused leave carries over and must be taken by 30 June of the following year [2].

Answer sources
  1. [1]Leave policy 2026.pdfp. 2, 4
  2. [2]Work rules.docxp. 11
Try:

Demo with made-up documents. The interface matches the real app.

Connects to where your documents already live

  • SharePoint
  • Uploaded files
  • Microsoft Entra ID

30+ formats: Word, PowerPoint, PDF (scanned too), Excel and CSV, OpenDocument, .eml email, HTML, Markdown, JSON, XML, and images including multi-page TIFF.

Answers you can check

Every claim in an answer carries a source number. One click opens the document DocuBox used. When the evidence doesn’t fit, it says so, or asks what you meant.

  • Page numbers on sources where the document has them. DocuBox never cites a page the source doesn’t carry.
  • See how it looked. Open the trail: what it searched, which documents it opened, whether it checked for newer versions.
  • What it found but didn’t read is listed separately, so you know where to look next.
Read 24 passages
Round 1
  • Searching SLA response times of both suppliers14 passages
  • Opening document SLA_Helpdesk_Alfa.pdf4 passages
  • Opening document SLA_Helpdesk_Beta.docx6 passages

Found what it needed

Alfa resolves critical incidents within 2 hours [1], Beta within 4 [2]. Beta guarantees 99.9% availability instead of 99.5% [2].

Answer sources
  1. [1]SLA_Helpdesk_Alfa.pdfp. 3
  2. [2]SLA_Helpdesk_Beta.docxp. 2, 5
Documents I found
  • Helpdesk tender 2025xlsx

Reads the spreadsheets people really use

A company workbook rarely holds one tidy table. A sheet has a title, a few facts and two tables stacked under each other. DocuBox finds them from the layout of the sheet, not from whether someone marked them up in Excel.

  • Each region on its own terms: a table, a cross-tab, key-value facts, or prose.
  • Column headers travel with every slice. A row from the middle of a long table still knows what each number means.
  • Large tables get a summary, so you can ask about the whole, not only a row.
The table summary, as search sees it

Quotes from three suppliers for air-conditioning service (Facilities, due 30 Apr 2026). CZK 171,200–198,900 before VAT, 12–48 h response, 12–36 month warranty.

More than a text search. It works like a colleague who knows the archive.

For each question DocuBox picks its own approach. It mixes keyword and meaning-based search, and keeps going until it has enough to answer, or it’s clear the documents don’t say.

From a SharePoint file to an answer

Most of the quality is made before anyone asks anything. Here is the path one document takes.

  1. 1

    Sync

    HR › Policies
    • PDFLeave policy 2026.pdfchanged
    • DOCWork rules.docxunchanged
    • PDFTravel allowances.pdfmoved
    • PDFLeave policy 2025.pdfdeleted
    • PPTBenefits 2026.pptxnew
    Stored with its permissions
    HRAll employees

    5 files · 2 to process

    Pulls new and changed documents from chosen sites and libraries, with their permissions.

  2. 2

    Reading

    Leave policy 2026.pdf
    p. 4 · scan
    4 headings1 tableOCR p. 4

    4 pages · 1 scanned

    Extracts text and structure. Scanned pages get OCR; charts and diagrams are described by a vision model.

  3. 3

    Chunking

    Leave policy 2026Art. 1 Purpose and scopeThis policy governs annual leave for all employees…
    Leave policy 2026Art. 3 Leave entitlementEmployees are entitled to 25 days of leave per calendar year…
    Leave policy 2026Art. 4 Carrying leave overUnused leave may be carried over until 30 June…
    Leave policy 2026 › Annex 1Entitlement tableYears worked | Days · 0–4 | 25 · 5+ | 30

    4 pieces

    Splits the document along its sections and tables. Each piece knows which section it came from.

  4. 4

    Enrichment

    HRleavepolicyen
    Art. 1 Purpose and scopeWho the policy applies to.annual leave
    Art. 3 Leave entitlementThe base entitlement and how it grows.25 dayscalendar year
    Art. 4 Carrying leave overCarry-over into the next year.30 June
    Entitlement tableEntitlement by years worked.5 years30 days

    4 summaries · 6 entities

    Adds summaries, key entities and tags, so a document can be found by what it means.

  5. 5

    Index

    Art. 1 Purpose and scopeHRAll employees
    Art. 3 Leave entitlementHRAll employees
    Art. 4 Carrying leave overHRAll employees
    Entitlement tableHRAll employees

    4 pieces · permissions on each

    Writes every piece to the search index with its vector and its permissions. Every answer draws from here.

Simplified illustration.

Every team asks something different

DocuBox works across the whole company’s documents. But each department needs something different from it. Pick a team and see how it helps.

Respects permissions

Human resources

Everyone reads the policies; only HR sees the pay bands. DocuBox answers each person only from what they may see.

How much annual leave do I get?

You’re entitled to 25 days of leave per calendar year [1]. After five years with the company it rises by 5 days [1].

Unused leave carries over and must be taken by 30 June of the following year [2].

Answer sources
  1. [1]Leave policy 2026.pdfp. 2, 4
  2. [2]Work rules.docxp. 11
Tracks versions

Legal and contracts

A folder often holds five versions of the same contract. DocuBox tells you which one is signed and in force.

Which version of the freight contract is current?

Version 3, dated 14 March 2026 and signed by both parties, is in force [1].

The folder also holds a newer v4, but it’s an unsigned draft with revised freight rates [2].

Answer sources
  1. [1]Freight_contract_v3_signed.pdfp. 1, 6
  2. [2]Freight_contract_v4_draft.docxp. 3
Reads spreadsheets

Procurement and finance

Quotes and price lists live in Excel. DocuBox finds the exact row, and what belongs with it.

What was the cheapest air-conditioning service quote?

Chladservis quoted the lowest, CZK 171,200 before VAT [1].

But it covers only 2 visits a year with a 48-hour response. KlimaTech costs CZK 13,400 more for 4 visits and a 24-hour response [1].

Answer sources
  1. [1]Service_quotes_2026.xlsx
Compares documents

IT and helpdesk

Two suppliers, two SLAs, each in a different format. DocuBox reads them side by side and compares.

How do our two helpdesk SLAs differ?

Alfa resolves critical incidents within 2 hours [1], Beta within 4 [2].

Beta guarantees 99.9% availability instead of 99.5% [2].

Answer sources
  1. [1]SLA_Helpdesk_Alfa.pdfp. 3
  2. [2]SLA_Helpdesk_Beta.docxp. 2, 5
Reads scans

Operations and safety

Inspection reports and manuals often exist only as scans. DocuBox reads them with OCR.

When do the fire extinguishers need their next inspection?

Fire extinguishers are inspected once a year [1].

The last inspection was on 12 March 2026, so the next is due by 12 March 2027 [2].

Answer sources
  1. [1]Hall operating rules.docxp. 7
  2. [2]Fire extinguisher inspection 2026.pdfp. 1

Why DocuBox when you have Microsoft Copilot

Copilot is an assistant for the whole of Microsoft 365. DocuBox does one thing, in depth: precise answers from your company documents.

  • Permissions under control

    DocuBox follows exactly how permissions are set in your SharePoint. Every document carries them along, and search applies them while it searches, not after the answer is written. Anything it cannot verify unambiguously, it shows to no one.

  • A specialised tool, deployed for you

    Copilot is a general-purpose assistant built for everyone. We deploy DocuBox for one specific company and adapt it to your specifics, such as your terminology or how your documents are organised in SharePoint.

  • Reliable answers, a year from now too

    We build DocuBox for clients where the quality of answers is critical. A range of mechanisms keeps answers accurate and reliable, and above all keeps them as good over time. Every change is checked on real documents and a set of test questions before it ships.

Copilot helps with email, meetings and writing. DocuBox answers from your company documents, with the source. Use them side by side. How permissions work

Security and your data

Every employee gets answers only from documents they may see, and the data stays in your Azure.

Everyone sees only what they may

Sign in with Microsoft. On every sync, DocuBox reads the SharePoint permissions of every document, and builds your answer only from documents you can open.

  • The filter is inside the search, not applied after it. A document you can’t open never reaches the language model.
  • Users, M365 and security groups, site groups including nested ones, organisation-wide links.
  • When in doubt, access narrows. A grant that can’t be tied to a real identity is dropped, never widened.
JN
Jana N.HR

“What are the 2026 pay bands?”

Pay bands 2026.xlsx
Internal pay policy.pdf
PK
Petr K.Sales

“What are the 2026 pay bands?”

Not in any document you can see.
Career guide (public part).pdf

Your data stays in your Azure

DocuBox is built entirely on Microsoft Azure services and designed to run in your own subscription. Each customer gets a separate deployment; nothing is shared with anyone else.

  • The language model runs on Azure OpenAI. No other AI provider in the data path.
  • Czech and English. The interface in both, answers in the language of the question, over mixed documents too.
Your Azure subscription
  • Azure OpenAIAnswers and text embeddings
  • Azure AI SearchThe index, with permissions on every piece
  • Document IntelligenceDocument reading and OCR
  • Container AppsChat, sync and processing
  • PostgreSQLConversations, metadata, audit
  • Blob StorageOriginal files

DocuBox in a minute

From chaotic searching to a verified answer.

0:00 / 1:00

Questions

What is DocuBox?

DocuBox is an AI assistant for company documents. Employees ask it questions in plain language, and it answers from the documents in your SharePoint, with a link to the source behind every claim. It respects SharePoint permissions, reads spreadsheets and scans too, and runs in your own Azure. It is made by StringData, a Czech software company.

Do our documents leave the company?

No. DocuBox is designed to run entirely in your own Microsoft Azure subscription: documents, index and conversations. The language model runs on Azure OpenAI, which doesn’t use your data to train models. No other AI provider is in the data path.

Will employees see documents they aren’t allowed to open?

No. DocuBox takes each document’s permissions from SharePoint and builds an answer only from what the person asking can open. When permissions change in SharePoint, the change applies on the next sync.

What if the answer isn’t in the documents?

DocuBox says it couldn’t find it in the documents available, instead of making something up. When a question is unclear, it asks. And because every claim links to its source, you can check an answer in one click.

How is DocuBox different from Microsoft Copilot?

Copilot is an assistant for the whole of Microsoft 365; DocuBox specialises in answers from company documents. It starts without a big project, on the libraries you choose, searches in several rounds like an experienced colleague, and understands document structure, tables and scans included. The two work side by side.

How do new documents get into DocuBox?

Sync picks up new and changed files from the SharePoint sites and libraries you choose, and removes deleted ones from the index. Admins can also upload files directly. It reads 30+ formats, including scanned PDFs and Excel.

Does it work well in Czech?

Yes. The interface is in Czech and English, and DocuBox answers in the language you ask in, even when your documents mix both.

What does it cost?

We price by number of users and document volume. Azure services run in your subscription, so you see those costs directly. DocuBox also shows the cost of every answer and can set per-user limits.

How do we start?

We set up a demo over a sample of your documents. Together we pick a few libraries, write down the questions you want answered, and measure how DocuBox does.

DocuBox is made by StringData

A Czech company that has been digitising business processes since 1992. We automate document work for banks, insurers, manufacturers and the public sector, from invoice processing to whole agendas. We build and support DocuBox ourselves, in Prague.

More about StringData
digitising business processes since
1992
analysts, developers and consultants
70+
projects delivered
1,000+
largest Czech and Slovak banks are our clients
10 of 10

StringData references

We have automated document work for these companies with our other products. DocuBox is the newest of them.

  • Product: Daňovka

    Tax declarations for 8,000 employees, electronically

    Daňovka replaced paper forms, in-person meetings and manual retyping. It took 2.5 months to implement and has run since 2019.

    Filling in the electronic form is fast and simple, and it keeps all the documents in order.

    Kateřina Hodicová, Service Delivery Manager, Siemens
    Full story
  • Product: KNBox

    700 documents to the land registry a month

    KNBox automates communication with the land registry when mortgages end: lien deletions, discharges and proceedings checks.

    A volume we couldn’t handle by hand became a routine KNBox gets through in a few minutes.

    Head of Mortgage Lending, mBank
    Full story
  • Product: Daňovka

    Digital tax declarations for 400 people in pharma

    Daňovka replaced paper forms and in-person meetings with HR. Implementation took less than two months.

    The biggest benefit was clearly the time saved. We no longer had to schedule meetings.

    Head of HR, ZENTIVA
    Full story

And others: Allianz, MONETA Money Bank, Raiffeisenbank, Prague Airport, CETIN, Modrá pyramida, Albert, Arriva, Teleflex.All case studies

We’ll show you on your own documents

Write or call. We’ll set up a demo and a pilot over a sample of your SharePoint.

Or write or call us directly:sales@stringdata.cz+420 266 772 600

New request

Hello, DocuBox here. Let’s set up a demo over your documents; it takes a few questions. What’s your name?

StringData s.r.o. processes these details only to reply to your request. Privacy policy