AI document management: what the Domesday Book teaches anyone who can't find their files

"It was definitely in that folder." How many times have you heard that at the office? Contracts saved on a colleague's desktop, delivery notes buried in email, certificates with names like "scan_final_2.pdf". AI document management promises to find everything in seconds, and it largely delivers. But only if you lay the foundations first, as a Norman king discovered more than nine centuries ago.

The Domesday Book: the same questions, everywhere

At Christmas 1085 William the Conqueror, King of England for almost twenty years, ordered a survey of the whole kingdom. As the UK National Archives explain, he wanted to know who held what before and after the Norman conquest, and what land and property were worth. Royal commissioners toured the counties asking questions; local juries answered. The result, copied mostly by a single main scribe in the following years, describes about 13,000 settlements and roughly 270,000 inhabitants.

Three details make the Domesday Book a lesson in document management.

• The same questions for everyone. Places were described following a common pattern. Today we'd call it a metadata scheme: the same fields for every document, which make comparing and searching possible.
• Built to be consulted. Royal administrators often used it to establish rights and privileges. It wasn't a document graveyard: it was a working tool.
• Controlled access. Kept first in the royal treasury at Winchester and later at Westminster, from around 1600 it was locked in an iron-bound wooden chest with three locks, each key held by a different official.

Your company archive needs the same three things: a common scheme, a practical purpose, and keys handed out with judgment. AI makes everything faster, but it replaces none of the three.

What AI really does in document management

A traditional document system stores files and lets you search them by name or folder. AI adds new capabilities:

1. Automatic classification. The system recognizes that a PDF is a contract, an invoice, a delivery note or a certificate and files it in the right place.
2. Data extraction. It reads the useful fields from each document: parties, dates, amounts, deadlines, codes. It's the Domesday Book done automatically: the same questions, asked of every file.
3. Natural-language search. Instead of remembering the file name, you ask: "the maintenance contract with the forklift supplier, the one renewed last year". The system finds the document and shows where the information is.
4. Summaries and comparisons. A summary of a long contract, differences between two versions, clauses that deviate from your standard. Always to be checked, never signed blindly.
5. Automatic deadline calendar. Renewals, warranties, certifications, inspections: extracted from documents and put on the calendar with a reminder.

Practical table: what to extract and who checks

In logistics especially, automatically reading transport documents is often the first project with a clear return: we explain it in AI for logistics and warehouses.

Document management isn't legally compliant archiving

Clarity matters here, because confusion is common. Managing documents means organizing, finding and using them. Preserving them in a legally compliant way means guaranteeing over time that a digital document stays intact, authentic and readable, with legal value.

• In Italy, preservation of digital documents is governed by the AgID guidelines on creating, managing and preserving digital documents, in force since January 1, 2022.
• Italian accounting records and related documents must be kept for ten years, under Article 2220 of the Civil Code.
• For e-invoices, Italy's Revenue Agency offers a free preservation service within its "Fatture e Corrispettivi" portal, which you activate by signing up.

Other countries have their own rules, but the principle holds everywhere: an AI document management tool is not automatically a compliant archiving system. The two can coexist: compliant archiving guarantees legal value, AI-powered management helps you find and use documents in daily work. Always ask the vendor which of the two they're selling.

The five-step method

Applied to documents, the method from our guide to business process automation looks like this.

1. Take inventory

Where are your documents today? Shared folders, email, ERP, USB sticks, paper. How many, of what type, who uses them. No inventory, no project.

2. Write your own "Domesday questions"

For each document type, decide the fields you really need: for a contract, maybe parties, subject, expiry, automatic renewal yes or no. Few fields, the same for everyone. That's your classification scheme.

3. Hand out the keys

Who can see what? HR documents shouldn't show up in the warehouse team's searches; confidential contracts shouldn't be summarized for anyone who asks. An AI system that "knows everything" and answers everyone is a risk, not a convenience. Think of the Domesday Book's three locks.

4. Start with one document type

Supplier invoices, delivery notes or supplier contracts. A one-month trial alongside your current method, measuring time saved and errors.

5. Decide what to keep and what to delete

GDPR requires you not to keep personal data longer than necessary. A good document system has retention rules: what's kept for ten years, what's deleted after a relationship ends, who decides.

Privacy and security: questions to ask your vendor

• Where are documents processed and stored? On servers in the European Union?
• Are our documents used to train models? Can that be excluded by contract?
• Do system permissions mirror those of the original folders?
• Is there a log of who searched for and opened what?
• How do we export our documents if we switch vendors one day?

A tidy archive is also a valuable target. Personal credentials, two-factor authentication and caution with emails asking you to "urgently share a document" are the bare minimum: more in phishing, AI and cybersecurity for SMBs.

Mistakes to avoid

• Uploading everything with no rules. An AI system on a messy archive produces messy answers, just more convincing ones.
• Relying on summaries for important decisions. A summary helps you find the point; decisions are made on the full text.
• Forgetting paper. The most important documents are often still on paper. You need a scanning workflow, with someone responsible for it.
• Buying a system nobody uses. If searching with the new tool is slower than asking a colleague, the colleague wins.

A typical scenario

Picture a professional firm of eight people with fifteen years of documents spread across shared folders and email. Every week hours go into hunting for the latest version of a contract or a client's certificate. You start with the inventory and write the "questions" for three document types. Then you switch on automatic classification for incoming documents and natural-language search on the archive, with permissions that mirror the firm's roles. After a month you measure: how many searches, how much time, how many documents not found. Only then do you extend to other document types. Similar approaches work for accounting firms and law firms.

Want an archive that answers instead of hiding?

We help professional firms and SMBs design document management around how people actually work. Explore our services or book a free intro call: in half an hour we'll work out which document type to start with.