From Gut Feeling to Predictive Marketing: How We Built Botilia’s AI-Powered CRM
When we started working with Botilia’s marketing team, their CRM was a glorified address book. Customer records existed, but they were static. Names, emails, purchase history. The marketing team decided who to target, what to offer, and when to send it based on experience and instinct.
That’s not a criticism. It’s how most enterprises operate. And for a while, it works well enough. But as Botilia scaled, the cracks started showing. Campaign response rates were declining. Customer segments were based on demographics rather than behavior. The team was spending more time deciding who to target than actually creating campaigns.
They didn’t need a better CRM. They needed a CRM that thinks. This case study is one chapter in a longer body of AI work we delivered for Botilia across every department.
What we built
Predictive customer segmentation
Traditional CRM segments are static: age group, location, purchase tier. Our system creates dynamic segments based on behavior patterns. It identifies clusters of customers who act similarly, not just who they are, but what they’re likely to do next.
Instead of “women aged 25-34 in Athens,” the system creates segments like “customers whose engagement has dropped 40% in the last 30 days but who responded to discount offers in the past” or “high-value customers who browse frequently but haven’t purchased in 60 days.” These segments update daily as behavior changes.
Automated action plans
The CRM doesn’t just show data. It tells the marketing team what to do. Each segment comes with a recommended action: send this type of offer, at this time, through this channel. The recommendations are based on what worked for similar customers in the past, continuously refined by outcomes.
The team can accept, modify, or override any recommendation. The system learns from all three responses. Over time, the recommendations get sharper because they’re trained on this specific team’s judgment, not just generic marketing playbooks.
Predictive marketing automation
Some actions don’t need human approval at all. When the system identifies a high-confidence trigger, a cart abandonment pattern that’s recovered 70% of the time with a specific follow-up, it can execute automatically. The team sets the rules and confidence thresholds. The AI handles the execution.
This isn’t a black box. Every automated action is logged, traceable, and reversible. The team has full visibility into what the system is doing and why.
The results at Botilia
Within six months of deployment:
Campaign response rates increased by 34%, because targeting shifted from demographics to behavior. The right message reached the right person at the right time.
Time spent on campaign planning dropped by 60%, because the CRM does the segmentation and targeting work that used to take the team two days per campaign.
Customer lifetime value increased by 18% for the cohort managed by the AI CRM, compared to a control group managed with the previous process.
The system identified three customer segments that the team had never considered targeting. One of these became their highest-performing campaign of Q4 2025.
Why we didn’t use an off-the-shelf CRM
We evaluated the major platforms before building. Every one of them had AI features. None of them worked for Botilia’s specific situation. The build-vs-buy math is structural, not incidental, we walk through the economics here.
The data structures didn’t match. The AI features were generic, trained on aggregate data from thousands of companies rather than Botilia’s specific customers. The Greek-language content handling was poor. And the integration with Botilia’s existing e-commerce platform would have required months of custom development anyway.
Building our own CRM was a bigger initial investment. But the result is a system that’s trained on Botilia’s actual data, integrated into their actual workflows, and improving based on their actual outcomes. That’s not something you get from a vendor’s AI checkbox.
How this applies to your business
The specific implementation is always different. Your customer data lives in different systems. Your marketing team has different workflows. Your customers behave differently.
But the pattern is the same: map the current process, identify where human judgment is being spent on decisions that data could inform, build AI that augments the team rather than replacing it, and create a feedback loop that makes the system smarter over time.
If your CRM is still a database and your marketing team is still targeting by demographics, the gap between you and your AI-equipped competitors is growing every quarter. Let’s talk about closing it. Reach us at inbusiness.gr.