Today
Managers and clerks spend hours a day in their inbox. In long threads the actual question gets buried. A good reply needs data from several systems.
An example
Monday morning at Talwerk Antriebstechnik: 140 new e-mails in the head of sales' inbox. The agent marks 12 as urgent, condenses a 38-message thread about a delayed project into five open points and drafts 20 replies with the current order status, such as the new delivery date of November 21 for order 4711-A.
Your data, wherever it lives
Today requests sit in full inboxes with long threads and PDF attachments, order status and delivery dates in the ERP, customer history in the CRM, and contracts and drawings in the DMS. demi reads, understands and links them per case, with your rights. The AI sorts by urgency, sums up long threads and drafts replies with order numbers and dates from the systems in seconds instead of minutes per e-mail, and you review and send them yourself.
Where the data lives today
- E-mails
- PDFs
- ERP
- CRM
- DMS & archive
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Data sources in this case: mailboxes, ERP, CRM and DMS. demi replaces nothing and links every e-mail to its customer, supplier, order, project or document, so to the right case.
- 2 · PermitPermission Layer
The agent reads only the mailboxes a person already has access to, shared ones included. The person reviews every draft and sends it themselves, and every step is recorded.
- 3 · ActIntelligence Layer
The agent sorts by urgency and case and sums up long threads with their open points. Order numbers, dates and amounts in the drafts come exactly from the systems, not from the model's memory.
How it runs
- 1
Clarify the data sources: mailboxes, ERP, CRM and DMS
- 2
The agent sorts new e-mail by urgency and case
- 3
It sums up long threads with their open points
- 4
It drafts replies with data from the ERP, CRM and DMS
- 5
You review, adjust and send
What changes
Less time sorting and searching
Nothing urgent slips through
Replies with the right data
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
+450 hfreed per year
Assumptions and calculation
demi sorts by priority, sums up threads and drafts replies with current data; you send.
5 min → 3.5 min per case
How we calculate
1,500 e-mails × 5 min × 30% × 1.0 (manual) × 12 months = 27,000 min = 450 h
Value per year, before demi costs
Freed per year+450 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
+$29,250
gained per year
450 h ÷ 1,760 h per role = 0.3 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
Does the agent send e-mails on its own?
No. It writes drafts that sit in your mailbox. You review, edit and send them yourself. Every e-mail that leaves the company stays your responsibility.
Which mailboxes can the agent read?
Only those a person can already access today: their own mailbox and shared mailboxes they have been given, such as sales or purchasing. Rights come from your existing mail system. Other people's mailboxes stay off limits.
Can it file e-mails to the right case?
Yes. It recognizes the customer, order or project from the sender, numbers and content and assigns the e-mail to the case. When it is unsure, it suggests the match instead of making it.
Where does the data in the drafts come from?
The case needs data from ERP, CRM and DMS: order status and delivery dates, contacts and history, documents. Every value in a draft links to its source. If something is missing, the agent marks the gap instead of making something up.
