✦Skillties AI · The game changer for HR

AI for HR that makes nothing up and gives nothing away.

A real innovation: Skillties AI understands language without guessing it. That solves the two problems AI in HR has struggled with so far: hallucinations and data protection.

No hallucinations
No personal data
Runs in Frankfurt

A network of texts

Skillties AI breaks every text into four levels: whole text, sections, phrases, words. Each level becomes a data point in the network. Shared words connect the texts.

Overview

What does it help with?

Wherever texts about work meet skills. Skillties AI turns them into connections anyone can follow.

Map learning content

Every course, video and podcast finds the skills it builds. Nobody has to tag catalogs by hand.

The right recommendation

Everyone gets the content that fits their gap and their goal, instead of a catalog with a thousand entries.

Place new skills right away

Anyone who creates a skill in their own words sees right away which skills are close and which content fits. Duplicate skills stand out.

Bring in your own content

Your internal courses and documents grow into the network in an area of their own. Only your organization sees and uses them.

The difference

Why is this better than a language model?

A language model predicts the most likely next word. It sounds convincing, but it cannot be checked. For decisions about people and their development, that is not enough.

✦ Skillties AITypical language model

How it works: Breaks texts down and searches a network of termsPredicts the most likely next word
Same input: Always the same resultCan differ every time
Hallucinations: Not possible, nothing is writtenPossible, even when it sounds plausible
Reasoning: Every match follows a path back to the textHard to trace
Data about people: Never goes inOften goes to the provider
Where it runs: Our own servers in FrankfurtAt the provider, often outside the EU
Cost: No token costs, your price stays stableEvery request costs tokens, the provider sets the price
EU AI Act: Built to be traceable and auditableExplainability has to be added later

Why is this better than a language model?	✦ Skillties AI	Typical language model
How it works	Breaks texts down and searches a network of terms	Predicts the most likely next word
Same input	Always the same result	Can differ every time
Hallucinations	Not possible, nothing is written	Possible, even when it sounds plausible
Reasoning	Every match follows a path back to the text	Hard to trace
Data about people	Never goes in	Often goes to the provider
Where it runs	Our own servers in Frankfurt	At the provider, often outside the EU
Cost	No token costs, your price stays stable	Every request costs tokens, the provider sets the price
EU AI Act	Built to be traceable and auditable	Explainability has to be added later

Tokens in short

Language models count in tokens, small pieces of words. Every question and every answer costs tokens, and the bill grows with every piece of content, every person and every search. The provider sets the price per token and can change it at any time. Skillties AI searches its own network and needs no tokens to do so. That is why your price does not depend on how much content is mapped or how often people search.

The right tool

Where a language model makes sense for us

Language models are good at writing. Skillties AI is good at matching. We use each exactly where it works best.

Writing · Gemini

A quick first draft

Write descriptionsThe name of a skill or skills profile becomes a first text that you accept or adjust.
Suggest skills for a profileA new skills profile does not start empty but with suggestions to choose from.

Matching and recommending · Skillties AI

Everything you need to rely on

Connect learning content to skillsTraceable through paths in the network, always the same for the same input.
Recommend content to peopleFitting the gap and the goal, without data about people ever reaching a language model.

Gemini runs in the EU, in the Frankfurt data center
Gemini only receives names of skills and profiles
Never data about people
A person accepts, edits or discards

Built in-house

Not an API. Our own invention.

Dr.-Ing. Sebastian Lothary, CTO of Skillties — Dr.-Ing. Sebastian LotharyCTO, Skillties

From practiceSebastian knows skill management from practice in organizations, with everything that works there and everything that does not.
The questionWhy do so many skill management projects fail, even though the need is so great?
The doctorateHe pursued this question in his dissertation at FernUniversität in Hagen, as an ontologist looking at language.
Skillties AISkillties AI grew out of this work: a method of its own that breaks language down instead of predicting words. Method, code and operation are all at Skillties.

Our own methodFrom research, not bought in

Our own codeDeveloped by Skillties

Our own serversOperated in Frankfurt

Outlook

Where is this heading?

Learning content is only the beginning. The same logic works for any text that describes what someone can do or should be able to do.

Today
Learning content and skills
Map content, close gaps, place new skills right away.
Next
Texts from your organization
Your own courses, role and job descriptions become part of the network.
After that
CVs in recruiting
Every CV is read with the same logic. Every application is treated the same way.

Recruiting

The same rules for every application.

The EU AI Act classifies AI used to select applicants as high-risk. It requires results to be traceable and everyone to be treated equally. A logic that breaks every CV down in the same way delivers exactly that: same input, same result, every match explained. A person always makes the decision.

Start for free.
Or talk to us directly.

The free version is ready right away. If you would like to talk about your transformation, we are happy to take the time.

Book a call