Agent Observability Changes the Way Developers Work
Agent Observability Changes Developer Workflows
Agent observability is changing how software teams build, run and improve applications in the AI era. In this Techstrong TV interview from SplunkConf 2026, Alan Shimel talks with Greg Leffler, Director of Developer Evangelism at Splunk, about how developers, SREs and operators are adapting to a new way of working.
Leffler says AI is not simply removing work from practitioners. It is changing the work they need to manage. Engineers are moving from manually writing every line of code toward planning, prompting, coordinating and validating AI-assisted systems. That shift creates new responsibilities around precision, strategy and oversight.
Observability Starts Earlier
The conversation explores how observability is becoming part of the development workflow instead of something added after deployment. Leffler describes Splunk Observability Studio, which integrates with the IDE and helps teams evaluate whether applications have the right OpenTelemetry instrumentation.
That matters because teams often intend to instrument applications but do not always follow through. AI-assisted instrumentation can help developers make applications observable from the start. It also helps ensure that business logic, transactions and key application behaviors are visible before software reaches production.
AI Agents Need Guardrails and Visibility
Agent observability also introduces a new layer of operational visibility. Organizations need to know whether AI-generated responses are accurate, compliant and safe to send to customers. They also need to understand how much agents cost to run.
Leffler connects this to tokenomics and the need to manage AI consumption. Visibility into token usage gives developers feedback while they build. That feedback can help teams make smarter choices about models, prompts, data stores and workflows.
Recursive Security Moves Closer to Reality
The interview also touches on recursive security, where observability becomes part of a continuous improvement loop. Security systems observe what is working, identify what needs to change and feed those lessons into the next iteration of defenses.
For technology leaders, the takeaway is clear. Agent observability is not only a monitoring feature. It is becoming a foundation for better software delivery, safer AI deployments, cost awareness and more resilient operations across Splunk and Cisco environments.
