Knowledgeassist
AI Skills: Using Recurring Tasks Permanently
What AI Skills are, which processes are suitable for them, and how to teach your AI once how your company works.
AI Skills are maintained capabilities within an AI work environment: recurring processes, rules, or task patterns from your company – such as “This is how we write proposals” or “This is how we respond to complaints”. Once stored as a Skill, the working method doesn’t need to be explained anew in every conversation.
The Problem: Good Prompts Get Lost
Anyone who works regularly with AI knows the pattern: For a task, after some trial and error, a really good instruction emerges – tone, structure, rules, examples. Next time, it’s gone, buried in an old chat history or in a text file that only one person on the team knows about. The result: The same explanation work happens again and again, and quality varies depending on who asks.
Skills solve exactly that: They turn a fleeting prompt into a permanent, maintained capability.
How a Skill is Structured
In webRichtung assist, Skills are part of the same work environment as chat, dictation, tasks, and artifacts. A Skill brings together the information needed for its application:
- a clear work assignment
- the necessary rules and examples
- recognizable boundaries and responsibilities
Added to this is the question of who the Skill applies to and who maintains it. That way, a proven working method can be described uniformly for the intended users without blurring responsibilities or exceptions.
Which Tasks Are Good Skill Candidates
The rule of thumb: Good candidates are things you regularly explain in the same way. For example:
- Texts with fixed rules: proposals, rejections, payment reminders, replies to typical inquiries
- Processes with steps: how a complaint is recorded, how an order is documented
- Formats: how a weekly report, minutes, or handover is structured
- House rules: form of address, tone, what cannot be promised, when a person must take over
One-off special cases are unsuitable – you’ll explain those in conversation faster than you’d maintain a Skill for them.
How to Build a Good Skill
- Observe: Which explanation are you giving for the third time? That’s your candidate.
- Formulate concretely: Not “write good proposals,” but structure, required information, tone, and a successful example.
- Name boundaries: What the Skill must not decide – such as discounts or legal commitments.
- Test and refine: Check results, feed deviations back into the Skill.
- Share: Make proven Skills an organizational capability so the whole team benefits.
Why the Effort Pays Off
A good Skill requires care in structure, rules, and boundaries. In return, it captures knowledge that would otherwise remain tied to individual people or old chat histories. How your company writes proposals then stands as a maintained capability in the AI work environment – a building block of what distinguishes an AI chat with company knowledge from a generic chatbot. How to create Skills is shown in the documentation at docs.webrichtung.de/assist/.
Frequently asked questions
What is an AI Skill?
A capability you teach your AI: a recurring process, a rule, or a task pattern from your company – defined once, then usable without explaining it again.
Which tasks are suitable as Skills?
Anything you regularly explain the same way: how you write proposals, how you respond to complaints, how a weekly report is structured.
Do Skills apply only to me or to the whole team?
That depends on the work environment and your access rules. What matters is that it remains clear who the Skill applies to and who maintains it.
How does a Skill differ from a good prompt?
A prompt applies to one conversation, a Skill is permanent. You don't have to repeat the explanation – the knowledge is stored and maintained in one place.