How Rengage Power the Future of Marketing Automation with TiDB

In the world of marketing technology, “real-time personalization” has been the holy grail for years. But the team at Rengage, founded by ex-Uber growth platform veterans, is aiming for something far more ambitious: a future where marketing is not just personalized, but fully autonomous, driven by AI agents that think, act, and optimize on their own.

This vision, however, collided with a fundamental industry problem. The team knew from experience that the intelligence of any AI system is ultimately constrained by the accessibility and integrity of its data. To build a truly autonomous system, they first had to deconstruct the traditional marketing stack and rebuild it on a more logical and efficient foundation.

The Architectural Challenge: A System Divided by Design

The primary architectural challenge in any AI-native application is data fragmentation. High-throughput marketing systems must process two distinct categories of data, which are almost always managed in isolated environments:

1. Hot Data (Transactional): High-velocity, real-time event streams such as user clicks, opens, and purchases.This data requires low-latency writes and reads for immediate decision-making.
2. Cold Data (Analytical): Aggregated historical data, including engagement metrics, conversion funnels, and user journey analysis. This data requires complex, ad-hoc queries for strategic optimization.

Rengage’s initial architecture, like many others, utilized a multi-system approach. This conventional design produced predictable and unacceptable inefficiencies for a company with their ambitions:

• Query Latency at Scale: Hybrid queries that required both hot and cold data were slow and computationally expensive, creating a bottleneck that made true real-time responses impossible.
• High Operational Overhead: Managing, scaling, and monitoring multiple, disparate database systems consumed significant engineering resources, diverting focus from core product innovation.
• Data Synchronization Failures: The requisite ETL pipelines between the transactional and analytical systems introduced latency and created a critical point of failure, compromising the data integrity required for AI decision-making.

This bifurcated model was fundamentally illogical for a system designed for AI-driven autonomy. The agents’ decision-making capabilities were being throttled by the inherent latency and complexity of the underlying data infrastructure.

The Strategic Solution: A Unified Data Core

The logical solution was to eliminate the data silos. Rengage’s engineering team, led by CTO Shujun Liu, made the strategic decision to migrate to a single, unified data platform capable of handling both transactional and analytical workloads. They selected TiDB based on a clear set of technical and operational criteria.

The decision was predicated on three core requirements:

1. Mixed Workload Processing Capability (Transactional + Analytical): This was the non-negotiable prerequisite. The ability to collapse the hot and cold data pipelines into a single system was the primary architectural goal.
2. Proven Horizontal Scalability: The system had to scale out to handle the massive data ingestion and query throughput required by enterprise-level marketing campaigns.
3. Developer Efficiency via MySQL Compatibility: Leveraging the existing MySQL protocol minimized the need for retraining and retooling, thereby accelerating the migration and future development cycles.

“This unified architecture is the logical prerequisite for enabling our AI agents to act in near real-time,” states Liu.

Unleashing AI Agents

With a unified data core provided by TiDB, Rengage’s vision for autonomous marketing was finally unlocked. Their AI agents could now access a complete, up-to-the-millisecond view of every user and campaign. This empowers them to operate as an integrated, autonomous system.

Rengage shared its unified architecture on SCailE Summit, 2025.

As shown in the above diagram, user actions and data flow into TiDB. It serves as a single data core for both real-time analytics and immediate action by the AI agents, powering a continuous loop of personalized engagement. Additional benefits include:

• Autonomous Segmentation: Creating dynamic user segments based on both real-time behaviors and historical patterns.
• Intelligent Journey Orchestration: Moving users through personalized journeys, adapting the path based on live insights.
• Automated Content Generation: Crafting and tailoring messages for individual users.
• Real-Time Performance Evaluation: Continuously monitoring campaign performance and making adjustments on the fly.
• Smart Traffic Adjustment: Reallocating campaign resources to maximize outcomes without human intervention.

The entire workflow, from event ingestion to multi-channel message delivery, now flows through a single, coherent data environment. This represents a fundamental leap forward in system efficiency and capability.

Lessons for Building the Next Generation of AI

Rengage’s architectural evolution provides a powerful blueprint for engineering the next generation of AI-native applications. The core lesson is that architectural simplicity is a strategic advantage.

By unifying their data architecture, Rengage not only solved their immediate latency and complexity issues. They also constructed a more scalable andfuture-proof foundation. Their journey demonstrates that the most effective path to building highly autonomous AI systems is to design a data infrastructure predicated on logic, efficiency, and the aggressive elimination of systemic complexity.

Ready to Build the Next Generation of AI?

Scaling agentic workloads are the defining challenges of modern AI development. See how leaders like Manus are solving them in our AI Solution Gallery.