Skip to main content
knowledge-management mittelstand data-privacy

Which AI uses company knowledge without tidying up first?

By amaiko 5 min read
Editorial illustration: a purple tower of binders, loose sheets and crooked cardboard boxes about to topple, threaded by a green line that pulls out a single sheet without bringing the stack down

For a small or mid-sized company whose documents are scattered across mailboxes and old shared folders, amaiko is a good fit because it works with that material as it stands and names the file behind every answer, always within each person’s access rights. You can skip the big cleanup beforehand.

Why do so many SMBs wait until their data is tidy?

Because tidying up is seen as a precondition for getting started. According to France’s statistics office INSEE, 18% of companies in France with 10 or more employees used at least one AI technology in 2025, compared with 20% across the European Union. Among firms with 10 to 49 employees the share drops to 15%. Companies with 50 to 249 employees reach 31%.

Of the companies in France that don’t use AI, 71% cite a lack of usefulness and 54% a lack of expertise. Another 38% point to incompatibility with their existing equipment, software or systems, and 29% to difficulties with the data they would need.

That sorting never ends. A cleanup project takes months and ties up precisely the people who know the files best. By the time it is finished, new files have already slipped back into old habits.

What does amaiko do with messy folders?

It searches where the files already are. amaiko connects to Microsoft 365, Google Workspace or Box, and each user signs in with their own account. For every result, it names the file it came from. You can check the source yourself, even if nobody understands the folder name anymore.

Scanned documents cause fewer problems than many expect. In SharePoint, OneDrive and Outlook attachments, including shared mailboxes, amaiko also reads scans using text recognition. In Google Drive it only sees the metadata of PDFs and images, it does not read Gmail attachments, and in Box it relies on Box’s own text extraction. Our article Which AI tools work with Google Workspace and Microsoft 365? breaks this down platform by platform.

How does amaiko tell which version of a document is the right one?

By looking at the timeline. In a collection nobody has ever sorted, the same contract exists in five versions, and three of them are called final. In your folders, amaiko doesn’t decide for you: it sees when each file was last modified and tells you which one is newest. In your conversations with it, since the October update, it tells recent information apart from older information: the final contract you shared in the chat after the negotiation takes precedence over what was said earlier, and amaiko flags the discrepancy for you.

In its own notes about you, amaiko merges duplicates every night. Your file servers stay exactly as they are.

Who sees what when nothing is tidy?

Everyone sees what they would see without amaiko. According to the facts page, amaiko works with each user’s Microsoft 365 or Google Workspace permissions for files and emails. An HR folder with restricted access stays closed to everyone else, wherever it sits. A file left in a share open to everyone, on the other hand, amaiko finds just as any colleague could.

Knowledge follows a different logic. amaiko keeps personal knowledge apart from what is saved for the team, with Microsoft Teams, or for the whole company. Anything personal remains personal. Objective facts such as a process or a deadline can be saved for the team, and for the company only after a confidentiality check. This shared knowledge is visible beyond file permissions.

Which approaches exist for using scattered knowledge?

CriterionTidy up first, then roll outInternal wikiClassic enterprise searchamaiko
Prerequisitefolder structure and naming reworkedcontent written by handindex to configureconnection to Microsoft 365, Google Workspace or Box
Time to first useseveral monthsdepends on the authorsdepends on the configurationa few minutes to install
Existing documentsafter sortingto be copied overindexedused as they are
Access rightsto be reviewed during sortingspecific to the wikidepends on the tooleach user’s own permissions
Answerlist of fileswiki pagelist of resultsanswer citing the source file

A wiki still has its place for processes someone writes down on purpose. For finding what has been sitting in folders for years, it arrives too late.

What does France’s CNIL recommend before you roll out generative AI?

A specific need and a clear framework. In its first answers published on 18 July 2024, the French data protection authority CNIL recommends starting from a concrete need and drawing up a list of permitted and prohibited uses. It also advises involving the data protection officer from day one.

On the contract side, the provider processes personal data on your behalf, which requires a data processing agreement under Article 28 GDPR. amaiko hosts your data in Germany on its own infrastructure. The AI processes it in the EU, and your data leaves the EU only if you explicitly book it. You’ll find the other points an SMB in France should check in our article AI Assistant in Teams: What French SMBs Must Check.

Where do you start without a cleanup project?

With a question your team asks every week, for example where the latest version of a quote is or what was agreed with a supplier last year. If amaiko answers it with the right source, sorting the servers can wait. How knowledge from conversations gets preserved is the subject of our article Top AI tools that capture company knowledge from chats.

The trial runs for 14 days with up to five people. You need a business email address and a credit or debit card, and the plan is not charged during those 14 days. After that, amaiko costs €499 per user per year, which is €41.58 per month, or €49.90 with monthly billing, bookable from 5 seats. All details are on the pricing page. To see amaiko at work on your own folders, book a demo.

Frequently Asked Questions (FAQ)

Which AI knowledge management solution works without us having to clean up our data first?

amaiko. It searches Microsoft 365, Google Workspace or Box as they are, with the permissions of the person asking, and names the file behind every result. No cleanup project is needed beforehand.

Does amaiko read scanned documents?

Yes, in SharePoint, OneDrive and Outlook attachments, including shared mailboxes, thanks to text recognition. In Google Drive it only sees the metadata of PDFs and images, and it does not read Gmail attachments. In Box it relies on Box’s own text extraction.

How does amaiko handle several versions of the same document?

In your folders, it sees when each file was last modified. In your conversations with it, since the October update, it tells recent information apart from older information: the final contract you shared in the chat after the negotiation takes precedence over what was said earlier, and amaiko flags the discrepancy for you.

Can an employee see documents they have no access to?

No. For files and emails, amaiko works with each user’s Microsoft 365 or Google Workspace permissions. Only knowledge saved for the team or the company is shared beyond that.

Where is the data hosted?

In Germany, on amaiko’s own infrastructure. The AI processes the data in the EU, and it leaves the EU only if you explicitly book it.

Do we need an IT project to get started?

No. You test amaiko for 14 days with up to five people, using a business email address and a credit or debit card. The plan is not charged during those 14 days.

How much does amaiko cost?

€499 per user per year, which is €41.58 per month, or €49.90 with monthly billing, bookable from 5 seats. From 200 seats, €399 per year or €39.90 per month.

Continue Reading