AI-powered email auto-reply — Proposal or car shipping.
The DigElite email chatbot connects to the customer's service mailbox via IMAP, classifies incoming messages, and generates a draft response based on its own knowledge base—either as a suggestion in the mailbox (classic "Drafts" folder) or as an automatic send when a response is highly reliable. Response reliability thresholds are configurable for each request. The default is "suggestion, no automatic sending." Every automatically sent response includes a note about the use of AI and a fallback contact person.
How the email bot relieves the burden of standard requests without losing trust.
Level 1 — Classification
Incoming emails are sorted into categories (default: inquiry / complaint / appointment request / spam / unclear). For each category, it is determined whether an AI response will be generated or whether the email will be forwarded directly to a human handler.
Stage 2 — Draft answer from the knowledge base
If the category is AI-enabled (e.g., "Inquiry about opening hours" or "Appointment request"), the bot generates a draft response from the shared knowledge base—with internal source attribution, but without attribution in the reply email body (otherwise it would seem odd). The draft is then saved in the inbox's drafts folder.
Level 3 — Shipping mode configurable
Two modes: "Suggestion" (default) — the sender reviews the suggestion and clicks send. Or "Auto-send with threshold" — the bot sends automatically when the likelihood of a response exceeds a defined threshold. Auto-sent emails contain mandatory information: AI notification, model identification upon request, and fallback contact with a human sender.
Workflow for auto-response.
1. Connect mailbox & define categories
IMAP access, classification categories, processing path per category (AI suggestion / AI auto-sending / directly to human).
2. Test knowledge base
The bot answers 20–50 historical queries; we check quality, sources, and thresholds. False positives are fed back into the category rules.
3. Live with threshold „proposal"
Auto-shipping will only be enabled once the suggestion mode is running stably and the team has confidence in it. The threshold for auto-shipping will be raised category by category.
Example: Property management company with several hundred units — double-digit number of emails daily.
A property management company with several hundred units under its care receives a mid-double-digit number of emails daily at info@. More than half are standard questions: "When is the next owners' meeting?", "How do I submit a repair request?", "When will I receive the utility bill?". We're creating categories, indexing house rules and management FAQs. In the first quarter, everything runs in suggestion mode, saving administrators a few minutes per email. In the second quarter, automated sending for "Owners' Meeting" and "Standard Repair Instruction" will be enabled—with an AI notification and fallback contact information in the footer.
„"Email automation is a matter of trust. That's why we're building the bot to start as a drafting assistant and only send it independently once it's been proven, category by category, that it can actually provide the answer.""
— Philipp Herrmann, founder of DigElite
What potential customers should ask before deployment.
How do I prevent the bot from sending an incorrect answer?
Three layers of security: First, by default only the suggestion mode is active—a human approves each answer. Second, the auto-sending threshold is set high (typically: only when the retrieval score is above 0.85). Third, when auto-sending, the bot always sends with an AI-generated note and a fallback contact so that recipients can quickly correct their mistakes.
Which email providers work?
Anything that supports IMAP/SMTP: your own mail server, Microsoft 365, Google Workspace, mailbox.org, IONOS Mail, Strato Mail, Posteo. For Microsoft 365, we recommend OAuth2 integration instead of an app password. Login credentials are stored encrypted in the customer's WordPress database.
What happens to personal data in the emails?
Email content is sent to the selected LLM for classification and response generation. With Aleph Alpha Luminous, this remains in Germany; with Mistral, in France; and with Llama on-premise, on the customer's hosting. We strongly recommend Aleph Alpha or Llama on-premise for email use cases—email content is often more sensitive than web chat requests.
Isn't the AI reference perceived as unprofessional?
The AI Act requires transparency. We phrase the notice discreetly—typically: "This response was generated automatically. For further questions, please contact [Name] directly at [Email]." Recipients appreciate the honesty, and you are legally protected. This notice is omitted in suggestion mode (human sending).
Where you can continue reading.
This feature is part of the DigElite chatbot family — check it out. Product Overview or the thematically related clusters.
Service chat on the website
Same body of knowledge, different interface.
Learn more →
Knowledge Database (RAG)
Response basis for every email draft.
Learn more →
GDPR-compliant chatbot
AI transparency in car shipping.
Learn more →
DigElite Chatbots — Overview
Pillar with all cluster topics.
Learn more →
15 minutes is enough to get an impression.
We'll be live-chatting with our own chatbot on nordzypern.live and showing you how it responds to real documents, when it honestly says "I don't know," and how it hands the call off to a human. No sales pitch, no Slide 47.
Watch the chatbot live & get an initial consultation