Customer communication
Sorting emails and enquiries
Your inbox sorts itself. A ready-made reply draft awaits enquiries - send it with one tap.
Running at fotobazos.cz
The problem
A shared inbox mixes enquiries, complaints, questions about order status, system notifications, and spam. Every email has to be opened, understood, and acted on - and an important message can easily get lost among twenty unimportant ones.
Who it is for
- Online stores
- Companies with an info@ inbox
- Customer support teams
- Real estate and service offices
How it works
Rules where rules are enough
Automated messages (payment gateway notifications, website forms, system notifications) are sorted by headers and templates. AI is not needed for that, and it is more reliable.
AI where understanding is needed
A language model reads and categorizes the rest: enquiry, order status, complaint, sales offer, notification, spam. For an enquiry, it immediately prepares a reply draft in your tone.
Card for approval
A card arrives on your phone: who is writing, what they want, and a reply draft. Send it, refine it with an instruction ("mention free shipping"), or mark it as resolved. The reply goes to the original thread.
You decide
No email is sent without approval. Before deployment, the system runs in shadow mode - it sorts and drafts, but sends nothing until you see that it sorts correctly.
What you get
- Connection to your inbox (IMAP/SMTP, Google Workspace, Microsoft 365)
- Categories and rules based on your operations
- Reply drafts in your company's tone
- Approval in Telegram or by email
- Shadow mode before going live
Running in production
Fotobazoš.cz
A system that runs beside the store, not inside it. It prepares posts from the catalog, creates products from one photo, watches comments, reviews, and inboxes, and suggests replies. The owner approves everything that goes out with one tap on her phone.
View the case study →Frequently asked questions
What if AI makes a mistake?
That is why AI suggests and a person decides. A new system starts in shadow mode: it sorts and suggests, but sends nothing. Then you approve each step with a single tap. Only what has proven itself in production can run on its own, and even then there is a log and a way to intervene. When facts are involved (prices, stock availability), a second step checks the output against the data source.
Where does our data go?
It depends on the solution, and we agree on it in advance. It is never done without your knowledge. In general, we send a language model only what the task needs. For email sorting, for example, addresses and phone numbers are removed before the text leaves the server. The solution proposal describes where the system runs and which models it uses.
Do we need to change the systems we use for automation?
Usually not. We connect through what you already have: APIs, product feeds, exports, email or a shared folder. For example, marketing automation at Fotobazoš runs entirely outside their website. It reads the feed and sitemap, so nothing on the website needed to change.
Does it fit your operation?
Describe how you do it today. The process analysis is free and does not commit you to anything.