Big Technology Podcast AI Agents: Mirage Or Real Revolution? — With Dmitry Shevelenko
136 snips
May 7, 2026 
Dmitry Shevelenko, Perplexity’s chief business officer focused on AI search and agentic products, digs into whether AI assistants can become real digital workers. He gets into computer use, trust and permissions, multi-model orchestration, enterprise adoption, Chinese open models, pricing, and why revenue matters more than hype.
AI Snips
Chapters
Transcript
Episode notes
Consumer AI Hit A Curiosity Bottleneck
- Shevelenko says consumer AI growth slowed because model capabilities outran user behavior, leaving many people stuck in basic information-retrieval habits.
- He calls this a capabilities overhang: people still use powerful models for weather, scores, and simple news instead of deeper agentic work.
Computer Use Looks More Like Payroll Than Novelty
- Shevelenko says computer use is stickier than voice or image novelties because users treat it like payroll, not entertainment software.
- Perplexity sees customers consuming more computer credits each week, tying usage to tax prep, analysis, and other economically productive tasks.
A Midnight Snowflake Query Without SQL
- Shevelenko used Perplexity Computer with Snowflake at midnight to analyze internal model usage without knowing SQL or asking a data scientist.
- He says a task that once required ten emails returned in minutes, showing how nontechnical employees can self-serve analysis.