Uploaded July 2026 | Updated September 2026, 2 weeks ago
Companies are spending heavily on AI agents, but many deployments still fail to deliver measurable business value. The problem may not be the technology. It may be that organizations are asking AI agents to perform jobs that were never clearly defined.
In this episode of Today in Tech, host Keith Shaw speaks with Satyen Sangani, CEO and co-founder of Alation, about why companies should treat AI agents more like employees and give them clear responsibilities, boundaries, success metrics and escalation rules.
They discuss:
• Why successful AI pilots often fail in production
• How unclear goals lead to unreliable AI agents
• Why AI agents need job descriptions
• The role of data, context, guardrails and evaluations
• Why 95% accuracy may still be unacceptable
• How governance can enable AI instead of slowing it down
• Why companies should start with high-value business problems
• Whether AI agents are being held to a higher standard than humans
The companies that succeed with AI agents will not be the ones deploying the most technology. They will be the ones that clearly define the problem, measure performance and continuously improve the system.
Subscribe to Today in Tech for weekly conversations about artificial intelligence, cybersecurity, enterprise technology and the future of work.
Follow TECH(talk) for the latest tech news and discussion!
#todayintech #agenticai #artificialintelligence #podcast
----------------------------------
Keith Shaw
linkedin.com/in/shawkeith
SUBSCRIBE: youtube.com/@CWTECHtalk
TIKTOK: tiktok.com/@todayintechpodcast
Companies are spending heavily on AI agents, but many deployments still fail to deliver measurable business value. The problem may not be the technology. It may be that organizations are asking AI agents to perform jobs that were never clearly defined.
In this episode of Today in Tech, host Keith Shaw speaks with Satyen Sangani, CEO and co-founder of Alation, about why companies should treat AI agents more like employees and give them clear responsibilities, boundaries, success metrics and escalation rules.
They discuss:
• Why successful AI pilots often fail in production
• How unclear goals lead to unreliable AI agents
• Why AI agents need job descriptions
• The role of data, context, guardrails and evaluations
• Why 95% accuracy may still be unacceptable
• How governance can enable AI instead of slowing it down
• Why companies should start with high-value business problems
• Whether AI agents are being held to a higher standard than humans
The companies that succeed with AI agents will not be the ones deploying the most technology. They will be the ones that clearly define the problem, measure performance and continuously improve the system.
Subscribe to Today in Tech for weekly conversations about artificial intelligence, cybersecurity, enterprise technology and the future of work.
Follow TECH(talk) for the latest tech news and discussion!
#todayintech #agenticai #artificialintelligence #podcast
----------------------------------
Keith Shaw
linkedin.com/in/shawkeith
SUBSCRIBE: youtube.com/@CWTECHtalk
TIKTOK: tiktok.com/@todayintechpodcast
