Six use cases with eight specialists

Your car mechanic does not do your tax return either.

Every employee has clear tasks, tailored permissions, individual skills, and optimized tools. That is what an ordinary Tuesday looks like.

Need different employees? Just hire them.

Who works here

Eight job descriptions — but only one point of contact: Conrad.

Conrad
Coordinator
Takes the job, breaks it up and hands it to the right people.

Anni
Analysis
Checks the numbers, compares, finds the contradiction — and would rather ask than guess.

Resi
Research
Searches, reads, cites. Brings sources back, not just claims.

Salvatore
Sales
Writes the approach, holds the tone, and knows the customer from the inbox.

Paul
Product Owner
Keeps the product in view and formulates its instructions for the next sprint.

Harry
Hacker
Focuses on security topics, investigates vulnerabilities and turns findings into practical next steps.

Arthur
Author
Writes clearly and purposefully, from concise notes to considered long-form texts.

Codula
Coder
Shapes software concepts and develops reliable solutions from the first idea to working code.

01The morning report

At 6:30, a job runs a stored work instruction.

Resi searches, reads, summarizes, and sends the result to your Telegram chat. The work instruction is a skill — a Markdown file that the agent can write itself on request and reuse afterwards.

If your monthly limit is already used up, the job does not run. You get a notification instead of a surprise on your bill.

Prompt

Dear Resi, find 25 medium-sized companies headquartered in
Germany that already use generative AI successfully,
and the successes they report from it. Include sources.

Schedule06:30 Europe/Berlin
Job
Skill
Web search
Response
Telegram
Log

02The clarifying question

Anni is working through an export and encounters something unclear. Rather than guess, she pauses the run and asks.

You reply in the app or Telegram — the run continues exactly where it stopped.

Every employee can ask questions and tries to make answering them as easy as possible for you.

Answer

Anni asks:
  12 rows in the export have no cost centre.
  How should I handle them?

  [ ] Skip the rows and report them
  [ ] Book them to cost centre 9999
  [ ] Cancel the run
  [ ] Enter a custom answer ...

03The lead research

Conrad takes the job and breaks it up. Four stages, four specialists, one result.

Resi gathers candidates, Anni assesses them against your criteria, and Salvatore writes the approach. Conrad himself does not write anything one of the three could do better.

Intermediate results are handed on as artifacts, not pasted chat text. That means everyone builds on the previous result; nobody needs to know the whole story. It saves time and money.

Conradtakes it, breaks it up
Resigathers candidates
Anniassesses
Salvatorewrites the approach

04The inbox

One email account, several rules, first rule wins.

Mail from @customer.com goes to Salvatore; anything with “invoice” in the subject goes to Anni. Every message becomes a chat with the right agent.

Only employees with permission may reply directly; the others do their work and pass the information on.

Rule 1: Sender contains @customer.com → Salvatore
Rule 2: Subject contains “invoice” → Anni
Rule 3: Folder “Press” → Resi
Otherwise: It stays put. No rule means no agent.

05The notebook

Your workspace consists of Markdown files and Excalidraw drawings — and agents may read and write in it.

Files are searched literally, including what appears in a drawing. Vector databases, embeddings, and RAG can be added at any time through MCP or an external tool.

That may sound old-fashioned, but it has an advantage: you can always see what the agent found because it is a file. You decide per workspace whether it may only be read or also written to.

06The task board

You write a card into the backlog and drag it to “Todos”. That sends it.

Employees move tasks from one Kanban status to the next until they are complete.

And what did that cost?

You can ask the agent in the middle of a conversation. It knows and tells you.

Cost

Model        gpt-5.6-terra
Provider     Claude
Tokens       42,708 input · 1,153 output
Context      64,139 of 80,000 tokens free
Cost         ~ $0.043  (estimate, not an invoice)

The same information appears afterwards in the console — per run, per user, and per model.