AI assistant

Describe it. It builds it.

Every powerful system has the same problem: the capabilities are there, but they are hidden behind menus, fields and jargon. Whoever works with it daily finds everything. Whoever needs it twice a month searches afresh every time.

The assistant in Caymland M4 turns the operation around. You say in your own words what you want to achieve. It knows the system, finds the way and creates it.

Not as a gimmick at the edge, but as a second, equal way of working with the platform.

Two related capabilities have pages of their own: the MCP interface, through which your own AI tools work with Caymland M4, and AI & machine learning for the patterns in your data.

It operates the system, not just the chat

That is the difference from a chat window that suggests copy. The assistant reaches the same capabilities you reach through the interface.

Creating, adjusting, deleting, searching, listing, analysing, checking, sending and explaining: for segments, campaigns, emails, contacts, companies, tags, dashboards, forms, templates and evaluations.

The goal is clear: the assistant should be able to operate the whole system. With every step of the build-out, further areas are added.

Emails that are genuinely finished

When the assistant builds an email, you do not get a draft to work on further. You get a designed email ready to send.

For that it draws on your templates and, when none is selected, on the appearance of your previous emails: your typefaces, your colours, your structure. The result looks like you, not like a default template.

Afterwards everything can be worked on in the editor as normal. The assistant is the fast start, not a one-way street.

It sees multi-step tasks through

Most assistants can do one step. Ask for something that needs three things in the right order and they lose the thread.

"A campaign for contacts in Zurich with a welcome email" becomes a plan here: first the segment, then the email, then the campaign that refers to both. Each step is proposed individually and confirmed individually, and the plan survives you doing something else in between.

You keep control of every intermediate step, without having to explain the whole task again.

It knows the manual too

Alongside building, it answers questions about how to work the system. "Where do I set the frequency rules?" "What does this unsubscribe reason mean?"

The answers come from the system's manual, not from general internet knowledge. And the search is built for daily language: you may say "add an action", even when the manual speaks of "nodes".

That replaces a considerable share of support requests and makes new staff self-sufficient sooner.

It remembers what you tell it

When you tell it something that should hold permanently, it keeps it. "Our newsletters always go out on Tuesdays." "We address customers formally as a rule."

And that is deliberately narrow: it remembers only what you tell it expressly. It interprets nothing out of your behaviour and builds no secret profile. What is stored can be viewed and changed.

You keep control

An assistant writing into live data on its own authority would be a risk, not progress. Therefore:

• Preview before every change. You see what would come into being, and approve or amend. Nothing is executed without confirmation.
• Your permissions apply unchanged. Whoever may not edit campaigns cannot create them in conversation either. The assistant has no special rights.
• Everything is logged. What was asked, what was proposed and what was approved.
• Personal data protected. Personal details are replaced before processing and put back only in the answer to you.
• It works with your real structures. It does not guess field names but reads your actual fields, categories and event types before proposing anything.
• When unsure, it asks, instead of inventing something.

When a service is unavailable

The platform carries on normally. The assistant falls back on rule-based suggestions instead of showing error messages. And the AI capabilities are optional as a whole: without them, Caymland M4 is fully usable.

What this looks like day to day

The new colleague. Instead of two weeks learning the filter logic, she describes her audience in one sentence and checks what the assistant builds from it. She learns the system while working with it.

Friday afternoon. A short-notice campaign has to go out. Segment, email and sequence are standing in twenty minutes instead of two hours.

The campaign that does not get going. Instead of hunting through conditions and logs, the marketing lead asks where it is stuck and gets the concrete step with a reason.

The question in the meeting. "How many is that, actually?" is answered in the conversation, not afterwards by email.