Conference Presentation, Keynote
Cresta: The Path to 90% Service Automation: Pitfalls, Secrets & Lessons from Enterprise Deployments
- Automation in contact centers is projected to be one of the two most useful AI use cases for enterprise transformation over the next two years, with the majority of interaction touchpoints expected to shift from human-only to AI-led over a five-to-ten-year horizon.
- While high-wage roles like coding and consulting are considered ideal fits for Large Language Models, lower-wage contact center roles face unique challenges due to reliance on messy, private, frequently changing internal data rather than evergreen public datasets, as well as the lack of verifiable properties for reinforcement learning.
- Future LLMs will likely require multimodal context including screen visual data and CRM information, necessitating a transition where "X percent automation" signifies AI handling 90% of specific tasks within 100% of conversations rather than 90% of calls end-to-end.
- Enterprises plan to improve knowledge base hygiene and adopt API-based interactions for AI agents instead of graphical user interfaces, while platforms like Cuesta will learn from all conversations to identify 43 distinct reasons for calls and analyze customer intents.
- New tools such as the AI Analyst product, currently deployed at United Airlines, will enable natural language discovery of automation opportunities, while agent assistants will work alongside humans to automate subtasks like data entry, transcribe conversations in real-time, and suggest context-aware actions.
- Native AI agents are expected to handle complex, turn-by-turn conversations, leverage multiple channels to resolve issues like router troubleshooting, and continuously learn from automation opportunities, mistakes, and human recovery steps rather than remaining static SaaS products.
- As service automation reaches a 70% to 90% threshold, contact centers are predicted to increase business volume through reduced interaction costs, shifting customer experience from a defensive cost center to a proactive engagement driver.
- User habits regarding bot interactions are expected to evolve as agent capabilities improve, with systems offering consumers a choice between AI assistance or human agent wait times, while high-value customers may be fast-tracked directly to human support without AI intervention.