We continue our CEO series with Douwe Kiela, the CEO of Contextual AI, who is addressing the challenges of building effective agentic applications. The shift to agentic amplifies the need for enterprises to improve their data management capabilities and infrastructure scaling. The best models won't perform well, if there isn't well built context to support them. Much like people, if there's not enough of the right information, decision making is going to suffer. There's an evolution from the prompt engineering needed to generate better results from LLM's, to the context engineering that crafts the right data to feed agents.

This is an area where agents can also help to tackle the data quality problem that many enterprises face. Standing the old computing paradigm on its head, effective agentic applications ought to take garbage in and put information out. Well built agentic architectures can understand data characteristics and not only evaluate its quality, but also classify it and apply the appropriate security controls to its use. The scope and scale of agentic potential demands much greater thought to achieve its full value. We need only look to the recent Open Claw project to see both the up and downsides.

More S&P Global Content:

• Next in Tech | Ep. 244: Agentic Customer Experience
• Next in Tech | Ep. 222: FinOps – Managing Cloud and AI Costs
• Next in Tech | Ep. 205: Agentic AI Impacts
• AI for security: Agentic AI will be a focus for security operations in 2025

For S&P Global subscribers:

• Agents in the enterprise: Laying the groundwork for automation
• The CX AI Agent Index 2025
• Agents are already driving workplace impact and agentic AI adoption – Highlights from VotE
• Big Picture 2026 AI Outlook: Unleashing agentic potential

Credits:

• Host/Author: Eric Hanselman
• Guest: Douwe Kiela
• Producer/Editor: Feranmi Adeoshun
• Published With Assistance From: Sophie Carr, Kyra Smith