AI Agents for Rule
Rule Communication is a Stockholm-based marketing automation company whose platform handles email and cross-channel customer communication. Vilmate has been Rule’s development partner for more than ten years, working on the platform’s backend, database, and front end.
About a year ago, Rule’s CEO asked us what AI could actually do for the business. That question led to two agents. One answers the CEO’s questions across the whole company, and the other helps the support team answer tickets faster.
The problem
Rule’s numbers lived in two places. Campaign statistics, reports, and conversion data, including Shopify conversions, sat in the platform’s AWS infrastructure, while support history sat in Intercom. Answers depended on reports built in advance. Any new question, especially one that crossed both sources, meant building another report or pulling and joining numbers by hand.


What Vilmate built
One Vilmate data scientist built both agents on Databricks. He worked closely with the Vilmate team that works on Rule’s platform day to day, and that collaboration is a big part of why the project worked.
The executive agent
We built the ingestion that moves Rule’s statistics and reports—including emails sent, open rates, and opens by customer and by industry—from AWS into Databricks. We also turned raw application data into structured SQL tables so that Genie builds answers from correct numbers. That data includes conversions from Rule’s Shopify integration, so for customers who run a Shopify store, it shows whether a campaign email led to a sale.
The CEO asks any question in plain language and gets the answer in chat, without anyone building a report first. Answers can also draw on context from text sources such as past support conversations. Dashboards still cover the reports that recur.
One example shows what this makes possible. The CEO can ask for the customers who raise the most support tickets, ranked, with each one’s revenue and email volume beside it. The answer combines support history from one agent with revenue and campaign data from the other, in a single report.
The support agent
We ingested Rule’s Intercom ticket history and split it into chunks sized so the agent can find the past tickets most related to a new one. For each new ticket, the agent drafts a reply based on how similar cases were answered before. A support specialist reviews the draft and sends it, so a person stays responsible for every reply that goes out.
“I basically just chat with our data now. Customers, campaigns, support, revenue: if I’m curious about something, I ask, and the answer’s just there. Nobody has to build a report first. It’s the first thing I open when I want to know how the business is doing.”