#315 Jarrod Johnson: How Agentic AI Is Impacting Modern Customer Service

What It Covers

Jarrod Johnson, Chief Customer Officer at TaskUs, explains how his company deploys agentic AI solutions from partners Regal and Decagon to transform customer service operations. He details implementation challenges, cost savings of 20-70%, workforce evolution strategies, and predictions for AI replacing traditional phone systems within 24 months across enterprise customer support.

Key Questions Answered

• •Agentic AI Definition: Agentic AI differs from traditional chatbots by taking direct actions in backend systems rather than just providing information. Example: Instead of explaining how to lock a credit card, the agent authenticates into financial systems and executes the card lock immediately, eliminating human intervention for tasks that previously required trained representatives to access multiple systems and complete multi-step workflows.
• •Resolution Rate Improvement: Traditional chatbots resolve 40-50% of customer interactions, while agentic AI systems push resolution rates to 65-70% by handling actions previously requiring human agents. This 20-25% improvement in automated resolution translates to significant cost reduction even for work already offshored, with TaskUs guaranteeing clients minimum 20% savings and up to 70% for US-based operations at $22 per hour.
• •Implementation Challenges: Three primary obstacles block successful deployment: outdated knowledge management systems lacking accurate operating procedures for AI ingestion, unsophisticated backend systems that humans navigate but AI cannot, and rapidly evolving platform capabilities requiring clients to commit to multi-year technology partnerships. First implementations require extensive hand-holding from platform engineers who custom-build solutions in real-time on constantly updating systems.
• •Workforce Transformation Strategy: Rather than eliminating 62,000 human agents, TaskUs redeploys workers from simple tasks to premium support queues, fraud investigation, and retention activities. The AI safety and model training workforce grows at 50% annually versus 7-11% for traditional customer service. Training duration for RLHF and red teaming work requires higher language proficiency but shorter onboarding than the typical 2-4 week customer service training cycle.
• •Technology Replacement Timeline: Johnson predicts traditional IVR phone trees will be replaced by natural language AI agents within 24 months. Current deployments show clients taking cost savings from automated simple interactions and reinvesting in complex human-handled cases. Multi-agent orchestration will emerge where specialized AI agents handle distinct workflows like fraud investigation or dispute resolution, coordinating to resolve complete customer journeys without human intervention.