Cobalt Credit Union

Cobalt Credit Union, a Nebraska-based institution managing over $1.3B in assets and serving more than 116,000 members — including military personnel stationed globally — faced a fundamental service gap: members in distant time zones needed account access at hours when contact center staff were unavailable. High volumes of routine inquiries like balance checks, transaction confirmations, and authentication requests consumed agent capacity during business hours, creating call queues and elevated abandonment rates. Without scalable automation, the credit union could not meet the always-on expectations of a geographically dispersed membership or redeploy staff toward higher-value interactions.

Cobalt Credit Union partnered with Eltropy to deploy a unified multi-channel AI strategy anchored by an AI contact center assistant named Coby. Rather than launching all channels simultaneously, the credit union followed a phased rollout: video banking first, then AI chat and text messaging, and finally Eltropy AI Voice in August 2024. AI Voice launched with an intent-based architecture, then added generative AI capabilities two months later. Coby handles both authenticated and unauthenticated member intents — performing real-time authentication conversationally, processing transactions, and delivering confirmations — without requiring a human agent. The voice channel integrates with existing Eltropy chat and text infrastructure, giving members consistent AI-assisted service regardless of how they choose to reach the credit union.

Coby achieved an 83% session containment rate, meaning the vast majority of member interactions are fully resolved by AI without escalation to a live agent. Key outcomes include:

• 116,000+ members now have 24/7 access to account services through AI-assisted voice, chat, and text channels
• Lower abandonment rates as members no longer wait in queues for routine requests
• Faster resolution times through real-time transaction processing and instant authentication
• Contact center agents shifted focus to complex, high-touch cases requiring human judgment

Member trust in AI-assisted service grew incrementally as Cobalt refined Coby's responses through ongoing prompt engineering, correlating improved containment with sustained satisfaction scores.

• A phased rollout — video banking, then chat and text, then AI Voice — builds member familiarity and surfaces integration issues before full deployment.
• Launching with intent-based AI before introducing generative capabilities allows teams to validate accuracy on predictable queries first.
• Ongoing prompt engineering is not a one-time task; tuning AI responses to balance technical precision with natural conversational flow is what drives containment rates above industry averages.
• Covering both authenticated and unauthenticated intents in AI Voice maximizes deflection; limiting AI to unauthenticated calls leaves significant automation opportunity unrealized.
• Multi-channel AI coherence — voice, chat, and text sharing the same underlying assistant — reduces member friction and simplifies ongoing maintenance.