Today
Service and internal help desks get the same questions again and again. The wiki is out of date and nobody has time to maintain it. So every question gets answered one more time by hand, often a little differently each time.
An example
At Talwerk Antriebstechnik the agent works through three months of service tickets. It drafts 25 FAQ articles, among them “Which oil does gearbox TW-480 need at −20 °C?”, and flags 12 wiki pages that contradict the current price list or manual rev. F. The head of service approves 21 articles and sends 4 back with comments.
Your data, wherever it lives
Today the answers are scattered across tickets, the service inbox, chats and the CRM; the wiki is out of date, and manuals and price lists sit next to it as PDFs. demi reads, understands and links them to products, processes and the document that currently applies, with your rights. The AI turns recurring questions into FAQ drafts with sources within hours and flags outdated wiki pages, and the person responsible decides what is published.
Where the data lives today
- Tickets
- E-mails
- Chats
- CRM
- Wiki & intranet
- PDFs
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
The case needs data from the ticket system, e-mail, chat, wiki and CRM. demi replaces nothing and links tickets, e-mails and wiki pages to products, processes and current documents such as manuals and price lists. demi writes back only after approval.
- 2 · PermitPermission Layer
Drafts are built only from tickets the article owner may read, and the agent removes customer data first. The article owner approves every article and every change, and this is recorded.
- 3 · ActIntelligence Layer
The agent groups similar questions into topics and drafts articles that cite the tickets and documents they come from; values are taken unchanged from the current document. It flags wiki pages that contradict newer documents and proposes the fix.
How it runs
- 1
Clarify the data sources: ticket system, mailbox, chat and wiki
- 2
The agent groups recurring questions by product and topic
- 3
Draft articles with sources and without customer data
- 4
Owners approve articles and corrections
- 5
Answer new tickets with the matching article
What changes
Fewer repeated questions
A wiki that matches the current state
Self-service for customers and staff
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+96 hfreed per year
Assumptions and calculation
demi turns recurring questions into sourced FAQ articles and flags outdated pages; you approve.
60 min → 36 min per case
How we calculate
20 FAQ articles × 60 min × 40% × 1.0 (manual) × 12 months = 5,760 min = 96 h
Value per year, before demi costs
Freed per year+96 h
Assumption: labor cost of $48.62 per hour in US manufacturing in June 2026 (BLS), plus workplace, IT and overhead.
= Value per year, before demi costs
+$6,240
gained per year
96 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
How does the agent find recurring questions?
It compares tickets, mail and chats by what they say, not by keywords. That way, differently worded questions about the same problem end up in one topic. For each topic you see the tickets, the product and whether the question is rising or falling.
Who approves the articles?
Every article has an owner, for example from service, product management or IT. The owner checks the draft with its sources, edits it and approves it. Nothing reaches the wiki without that approval, and every change is recorded.
Can we publish FAQs for customers?
Yes. You decide which articles stay internal and which appear in the customer portal or on the website. Before publication, the agent removes customer names, contact details and internal notes. External articles get their own approval from the responsible person.
What happens to wiki pages that are no longer right?
The agent compares wiki pages with newer manuals, price lists and product data. When it finds a contradiction, it marks the passage, names the newer document and proposes a correction. The owner decides.
Which systems does an FAQ built from tickets need?
A ticket system or a service mailbox, plus your wiki or file share. Chat and CRM complete the picture but are not required. How the data from your systems gets to demi, we clarify in a conversation. Nothing is replaced.
