Artificial intelligence & machine learning

Your data knows more than you can see in it

Millions of small observations sit in your system. Who opened what, and when. Who was on the same page three times. Who has done nothing for eight weeks, although they used to be there every week.

Individually, none of them is interesting. Together they form a picture no person could work out by hand.

That is exactly what the machine learning part of Caymland M4 is for.

Two related capabilities have pages of their own: the AI assistant, which operates the system for you in conversation, and the MCP interface, through which your own AI tools work with Caymland M4.

Interests that arise from behaviour

Ask a customer about their interests in a form and you get a polite answer. Watch what they actually click on and you get the truth.

Caymland M4 derives topic interests from real behaviour: from links clicked, pages visited, documents downloaded and forms submitted. That produces a profile which keeps correcting itself, instead of standing unchanged since the sign-up in 2019.

The result is usable straight away: as a filter in a segment, as a condition in a campaign, as a criterion for the content that gets delivered.

Recognising similar customers

Some things in common appear in no field. Two customers can come from different regions, be of different sizes, and still take the same path.

The analysis groups contacts by how they behave, not only by how they were recorded. That leads to audiences nobody would have arrived at by hand, and it helps with the most important question in new business: who looks like our best customers, before they became one?

Signals rather than scores

A score is a number. A signal is an observation with a reason.

Not "has 84 points", but "opened the pricing page twice yesterday, after six months of nothing". Not "score has fallen", but "has opened nothing for eight weeks, although every message used to be read".

Signals like these are immediately usable by people, because they explain why somebody deserves attention right now. They collect into a daily list that answers the question every sales day begins with: who do I call today?

The next sensible step

For a contact, a suggestion can be retrieved at the press of a button: what would be the right action now? Call, send a particular offer, take into a sequence, leave alone.

The suggestion rests on this contact's history and on the patterns of comparable cases. It is produced on request, not unasked on every view, and it is a suggestion, not an instruction.

If no AI service is available, a rule-based recommendation takes over. The list stays usable.

Small helpers in daily work

• Photograph business cards and have them created as a clean contact, instead of typing them up on a trade fair evening.
• Translations of content into the languages you run.
• Copy variants for A/B tests, in your tone.
• Drafts for personal outreach, when an account owner is to write to a contact.

Collected as it should be

Behavioural data is only valuable when it is collected cleanly.

Interest recognition and click analysis work with consent logic, with transparency about what is captured, and with configurable retention periods. Personal details are replaced before they reach a language model. Operation is designed for the revised Swiss Act and the GDPR, rather than built on hope.

And: the AI capabilities can be switched on and off. Caymland M4 is fully usable without them.

What this looks like day to day

The silent cancellation. A customer reduces their activity over weeks without saying anything. The signal stands out, the win-back starts before the cancellation arrives.

The reawakened prospect. After months of silence somebody opens the pricing page twice. They are at the top of the next morning's list, with a reason.

The new audience. Behavioural analysis reveals a group nobody had defined as an audience, but which behaves like the best customers. A campaign of its own comes out of it.