The Times Tech Podcast Bonus episode: Agentic AI explained – The next phase of artificial intelligence
8 snips
Feb 23, 2026 Lilia Christoff, Partner for AI and Data at PwC, advises financial firms on AI strategy and governance. She explains what agentic AI is, how multiple agents coordinate and make decisions, and how systems can be designed and overseen. Short-term rollout challenges, cost and build-vs-buy trade-offs, oversight at scale, and workforce shifts are discussed in clear, practical terms.
AI Snips
Chapters
Transcript
Episode notes
Agentic AI Is Multiagent Orchestration
- Agentic AI equals multiple specialized agents plus an orchestration engine, not a single chatbot calling a language model.
- Agents act like team members with roles, communicate, and a manager agent deduces outcomes from their inputs.
Agents Turn Messy Data Into Deduced Conclusions
- Agentic systems excel at research by consuming structured and unstructured data across sources and synthesizing a conclusion.
- A manager agent aggregates specialist agents' findings to present a deduced truth to a human.
Run Real-Time Tests And Self-Healing For Agents
- Test agentic systems in production continuously and build self-healing and business continuity plans for model drift and hallucinations.
- Monitor model versioning and external infra changes and switch to human or alternative models when thresholds hit.