KnowledgeCapital

KnowledgeCapital › Docs

AI and machine learning: what they do for you, and what they do not

Where the local AI comes into the software, what it needs to be useful, and how to check that nothing leaves your own server.

In short. The engine runs on the space's server: your entries, your CVs and your documents are not sent to any AI provider, confinement is active by default and every outbound network connection is logged. The AI proposes questions, links, definitions and forecasts; you are always the one who confirms or dismisses.

An AI that works in-house. The engine runs on the space's server. Your entries, CVs and documents are not sent to any AI provider; confinement is active by default and every outbound network connection is logged.

Where it comes in, screen by screen

ScreenWhat it doesWho decides
SearchUnderstands a question in everyday language; “Who knows?” finds people, not just documents—
Assessments › Question workshopDrafts questions from your validated contentYou validate each question
Untapped knowledgeReads CVs, groups knowledge by theme, proposes tasksYou transfer or dismiss
Smart repositorySpots emerging skills and levels to raiseYou apply them or not
ConnectorsProposes document ↔ knowledge linksYou confirm
GlossaryProposes definitionsYou review them
Predictive analyticsForecasts what will tip over, spots what is dropping off, forms knowledge communities, estimates how long a knowledge item will surviveYou decide on the actions
Search: Search, Who knows?, What we risk losing.
Search: Search, Who knows?, What we risk losing.

Predictive analytics

Predictive analytics: the horizon and the five tabs.
Predictive analytics: the horizon and the five tabs.

1 The horizon (6 to 36 months), then five tabs: 2 what will tip over, 3 what is dropping off, 4 knowledge communities, 5 survival of knowledge, 6 what drives reading.

Learning needs history. In a newly created space, forecasts are cautious (“Nothing tips over within 12 months”): there are not yet enough assessments, departures and reads to learn from. Allow a few months of real use. Dates of birth, repeated assessments and reads of entries are what feed them.

Checking what the machine does

1
Administration › Artificial intelligence: 1 what this machine can do, verified (the capabilities actually available on your server); 2 the confinement status; 3 the network outbound log.
Artificial intelligence: verified capabilities, confinement, the network outbound log.
Artificial intelligence: verified capabilities, confinement, the network outbound log.

Confinement can only be lifted deliberately, with explicit confirmation, and the action is logged. Connectors only open the addresses you connect.

An AI that works in-house.Search in everyday language, questions drafted from your validated content, proposed document links, predictive analytics: everything is computed on your server, and checkable in the network outbound log. The trial lasts 14 days, with no card. Start a free 14-day trial (no card needed) or request a demo.

See also: Step 10 — Connecting SharePoint, your DMS, Notion, your intranet or your own system · Keeping the repository up to date without spending your days on it · Reporting: the decision board, the ready-made reports, and your own · Sovereign knowledge management with local AI · A knowledge ERP: what your ERP does not manage

Frequently asked questions

Is our data sent to an AI provider?

No. The engine runs on the space's server; confinement is active by default, can only be lifted deliberately and with explicit confirmation, and every outbound network connection is logged.

What does the AI actually do in the software?

It understands a question in everyday language, drafts assessment questions from your validated content, reads CVs, spots emerging skills, proposes document links and definitions, and computes the predictive analytics. You confirm, transfer or dismiss every time.

Why do the predictive analytics show nothing?

In a newly created space there are not yet enough assessments, departures and reads to learn from. Allow a few months of real use.

Docs