The Challenge
The platform processed millions of daily transactions across banking, investing, and payments. AI agents needed to access user context from 7 different systems to handle support inquiries, fraud alerts, and personalized financial insights.
At peak load, the existing context assembly pipeline buckled — response times exceeded 3 seconds, and AI agents frequently timed out. Users received generic responses that undermined trust in the platform and drove support ticket volume even higher.
The Solution
mantle handled the full query load — 50,000 context queries per minute at peak — with consistent sub-60ms latency. Zero-copy connectors read transaction, KYC, and event data in place without adding load to production databases.
Quality scoring ensures AI agents receive the most relevant context for each interaction type — prioritizing recent transactions for fraud alerts, account history for support inquiries, and behavioral patterns for financial insights. All with full audit trail compliance.
Results
50K
Queries per minute
99.99%
System uptime
7
Data sources connected
<60ms
P95 latency
“We tested every context solution on the market. mantle was the only one that could handle our volume without compromising latency or quality. At 50K queries per minute, it never flinched.”
— CTO, Consumer Fintech Platform