A digital assistant that keeps your inbox under control.
A digital assistant right in your email inbox. It sorts incoming messages, suggests replies, holds conversation context across weeks. The team only picks what needs judgment — the assistant handles the rest.
It's not an auto-responder. It's not a reply generator.
It's a digital assistant for the customer inbox.
The assistant doesn't imitate you. It sorts incoming mail, categorizes request types, holds conversation context and suggests replies in your company's tone. You approve or edit — or enable full autonomy for defined types.
What you customize on this assistant.
We don't build templates. We build the email assistant individually based on how your company communicates by email.
Style and tone
How the assistant writes replies — formal, friendly, technical. Brief or detailed. In your brand's tone.
What it can do
Sorting incoming mail, replies, follow-ups, categorization, handover. You define the scope.
What it knows
Your services, procedures, documentation, FAQs and internal rules. Trained on the reality of your company.
Connections
Email inbox, CRM, helpdesk, invoicing and internal software — it works with live data.
Handover to a human
Whatever needs judgment, the assistant hands to the team with the whole thread context. Rules to fit you.
Every solution bespoke
Plus anything else — it's up to you what the assistant should know, do and which features it has.
With every assistant you also get a free dashboard — conversation history, statistics, escalation overview and live conversations where your team can take over the chat directly from the overview.
The customer asks. The assistant finds it. You approve.
The question “where's my order” has a clear answer in 30 seconds. You stay at the Approve button.
Five types of communication that email carries in business.
Email is the most asynchronous channel. The customer expects a reply within hours, not minutes — but a quality one. The assistant categorizes and handles by type.
A customer query
Questions about a product, service, terms, availability. The assistant suggests an answer from the knowledge base.
Orders and statuses
Order confirmations, status queries, changes. The assistant knows the status from the connected system and suggests a reply.
Returns and sensitive topics
A sensitive topic — the assistant doesn't reply autonomously but categorizes, collects context and hands to the responsible person. By arrangement it can handle these itself too.
Multi-step transactions
Threads of several messages (booking, a negotiation process). The assistant holds context, doesn't start over each time.
Internal team communication
Forwards, FYIs, coordination. The assistant recognizes it's not for a customer and doesn't sort it as customer-facing.
Typical customer interactions.
Three examples of email communication across industries.
→ Technical query → specific solution + document + meeting proposal → escalation to a specialist.
→ Personalized reply with booking identification + an action in the PMS system.
→ Inquiry → date proposals from the calendar → connection with a specific person.
How the assistant sees your inbox.
Every incoming message gets a category, priority and a suggested next step. The team opens an already-sorted inbox — it doesn't start each day at zero.
What the company stops dealing with.
Email is the oldest and most overloaded customer channel. The assistant makes it manageable.
Inbox bombarded every day
An assistant spends hours sorting into categories.
Incoming messages automatically sorted, the question “what's important today” is solved by morning.
Answering the same things over and over
60% of emails are variations of 30 typical questions.
The assistant suggests replies for the team, or handles them autonomously per the setup.
Forgotten conversations
An email from last week falls through, the customer gets no reply.
The assistant tracks conversations, flags unresolved ones, escalates after a defined time.
The team doesn't know what the sales rep handled a month ago
The context of historical communication gets lost.
Full conversation history + CRM in one panel, available with one click.
Sensitive emails wait a long time
A sensitive email arrives at night, gets handled only in the morning.
The assistant recognizes the sensitivity and escalates immediately to a specific person.
A new team member doesn't know the tone
A junior replies differently than a senior, the client notices.
The assistant suggests in the tone defined for the brand, the junior approves.
How the assistant is trained on your company.
Email carries more context than other channels. Training focuses on detail, nuance, tone.
Tone and formality
B2B vs. B2C, formal vs. partnerly. We calibrate from your historical emails — at least 100 examples.
Knowledge base
Products, services, terms, FAQs, policies. The assistant draws solely from approved sources, doesn't improvise.
Categorization logic
What's a query, what's a complaint, what's an order, what's spam, what's internal communication. We define the rules together.
Decision rules
What to answer autonomously, what to send for approval, what to escalate. Granular by type and sensitivity.
Examples of good conversations
Training on 30+ historical conversations where the team marked “this is how it should be done”.
Which systems it connects to.
The email assistant lives in the inbox ecosystem. It connects to what you already use.
Email inbox
Gmail (Google Workspace), Outlook (Microsoft 365), custom IMAP/SMTP. It works inside the inbox, not outside it.
CRM system
Raynet, Anabix, eWay-CRM or a custom API. Customer lookup, storing conversations, qualification.
Helpdesk
Daktela or your own helpdesk. Converting an email to a ticket if the team works in a ticketing system.
E-commerce
Shoptet, Upgates, WooCommerce. Order statuses, availability, returns.
ERP / invoicing
Pohoda, Money S3, Helios, ABRA, iDoklad. Invoices, delivery notes, statuses.
Booking system
Reservio, Reenio, Bookio, hotel PMS. Hotels, salons, restaurants, clinics. When a system has no API, we solve the connection another way too.
What the assistant actually does.
Actions it can perform on its own or with your approval.
- 01Classifies an incoming message (query / order / complaint / spam / internal)
- 02Suggests a full reply in your company's tone, ready to send
- 03Finds information in the CRM, e-shop, ERP or helpdesk in real time
- 04Sends a document (invoice, guide, terms) as an attachment
- 05Identifies that a conversation is waiting for your input and flags it
- 06Escalates sensitive messages immediately to a specific person per CRM rules
- 07Creates a ticket in the helpdesk from an incoming email with full categorization
- 08Saves a closed conversation to the CRM with a summary and outcome classification
- 09Monitors the SLA — if a conversation stalls longer than a defined time, it escalates
How the assistant handles a multi-message conversation.
Email communication often runs over 5–10 messages across weeks. The assistant holds context, tracks statuses, escalates on stagnation.
Classification
The assistant categorizes the message, identifies whether it follows an existing thread, loads the history.
Drafting
For typical queries it suggests a full reply. For complex ones it prepares materials — facts, context, suggestions.
Team / autonomous
Approval mode: 1-click approval. Or full autonomy for defined types.
Status tracking
The assistant records when the customer replies. If a conversation stalls, it flags it.
CRM record
After closing a conversation it saves a summary to the CRM, classifies the outcome, archives for audit.
When it hands over to a human.
With email the line between “reply autonomously” and “send for approval” is finer than in chat. We define it together.
Sensitive topic
Refunds, legal or health questions and other more sensitive situations — a human by default, or a digital assistant by arrangement.
Contractual change
Anything with legal weight (terms change, refund, extension) — always a human.
The customer wants a human
When a customer explicitly asks for a human in an email, the assistant prepares materials and hands to a specific person.
Outside the approved scope
A question outside training → the assistant prepares materials, hands to a human.
A stalled thread
If the customer doesn't respond to the assistant's question for 5 days, it escalates to a human for follow-up.
What deployment looks like.
Seven steps from the first consultation to production — 10–14 days.
Consultation
Mapping email volume, types, escalation logic, communication tone.
History audit
Analysis of the last 500+ historical conversations for patterns, categories, types.
Configuration
Tone, categorization, approval logic, routing rules.
Training
Knowledge base, approved answers, examples of good conversations.
Integration
Inbox, CRM, helpdesk, ERP, inventory — connection to real data.
Approval mode
Launch in approval mode — the team approves every reply, we calibrate.
Full operation
Gradual switch to autonomy for defined categories, weekly reviews.
Operational impact we measure from the first week.
Email efficiency is often neglected. The assistant makes it visible.
What went through the inbox
What went through the inbox — what's actually measured.
- Incoming by category
- Answered autonomously vs. by a human
- Response time (assistant vs. team before)
What moved off people
What moved off people — what's actually measured.
- Hours saved on sorting
- Hours saved on typical replies
- Shifting attention to complex cases
How you keep promises to customers
How you keep promises to customers — what's actually measured.
- Time to first response
- Time to resolution
- Stalled conversations (early warning)
How well the assistant works
How well the assistant works — what's actually measured.
- Approval rate (how many drafts pass without edits)
- Re-open rate
- Satisfaction after closing a conversation
Thirty minutes to know whether an email assistant makes sense for you.
No presentation. We'll discuss your communication volume, typical customer situations and how the assistant would fit into your processes.
30 minutes. Online. No commitment.