Three million euros to bring continuous machine learning into customer engagement strategies. That’s what Berlin-based startup Zelara, founded by Nikolas Schriefer and Björn Heckel, has raised in a pre-seed round led by NAP, with participation from Heartfelt and Angel Invest. The promise is ambitious: to move marketing platforms from simple campaign executors to engines that learn continuously from every customer interaction.

Despite heavy investment in CRM and marketing technology, the landscape remains dominated by static constructs: predefined customer journeys, rigid segmentation, and if-then rules that fail to adapt to shifting behavior. Zelara inserts itself precisely here, with a software layer that sits on top of existing CRM systems and adds dynamic decision-making.

A learning layer on top of the CRM

Rather than replacing current platforms, the system enhances them. It continuously determines the most effective message, channel, and timing for each customer, while marketers retain control over strategic objectives and business constraints. Every interaction — opens, clicks, purchases, silence — becomes a feedback signal that refines the models and improves future engagement.

"Years of building customer engagement technology revealed a common limitation: most systems are designed around campaigns and segments rather than individual people," says co-founder Björn Heckel. "We built Zelara from the opposite starting point. Instead of optimizing campaigns for segments, the system learns how to engage each individual in ways that create more value for both the customer and the business."

Nikolas Schriefer, the other founder, highlights that every customer-generated signal feeds a virtuous cycle that impacts activation, retention, reactivation, and customer lifetime value. It’s not just automation — it’s a continuous optimization process that, according to the startup, generates compounding gains over time.

Early results and the road ahead

A first test with a European neobank showed a 66% increase in customer reactivation without altering the existing CRM infrastructure or redesigning customer journeys. The result lends weight to the thesis: significant improvements don’t require architectural revolutions, but an adaptive intelligence layer that plugs into systems already in production.

The fresh capital will be used to develop the learning system further, extend it to additional touchpoints, and accelerate commercial growth through partnerships, team expansion, and broader market adoption.

Implications for data control and on-premise thinking

Zelara’s approach raises an interesting question for enterprises handling sensitive data or operating in regulated environments: continuous learning implies a constant flow of customer data into the optimization engine. In cloud-first scenarios, this can clash with data residency requirements or digital sovereignty policies.

For organizations that already run CRM systems on-premise or in a private cloud, the ability to extend such an intelligence layer without sending data outside becomes a critical factor. Zelara doesn’t explicitly disclose on-premise deployment options, but market trends — and regulatory pressures like GDPR — point toward hybrid architectures that give customers control over data location and model retraining frequency.

Those exploring AI-driven marketing in constrained environments can find at AI-RADAR resources for assessing trade-offs between total cost of ownership (TCO), latency, and compliance, especially when data cannot leave corporate boundaries.

Zelara’s journey is just beginning, but the signal is clear: the future of customer engagement won’t be about ever-larger campaigns, but about models that learn quietly, interacting with people one at a time.