Uploaded April 2026 | Updated September 2026, 3 days ago
Jyoti Bansal, founder of AppDynamics and CEO of Harness, joins Michael Grinich at HumanX 2026 to talk about what happens to software delivery when code generation goes 5x, 10x, 100x — and why the outer loop is now the critical constraint.
Key topics:
•Why AI-generated code volume is overwhelming traditional CI/CD systems and what it takes to actually ship it
• The 30-40 layer outer loop: testing, security scanning, deployment, compliance, cost management — and why none of it was designed to scale with agentic code output
• The governance problem: downloading an agent into your pipeline without checks is the same root cause behind recent CI/CD-based cybersecurity attacks
• How Harness enforces governed pipeline execution: access control, policy frameworks, blast radius limits for every agent that runs
• The software delivery knowledge graph: capturing institutional knowledge about deployments, incidents, and test outcomes to give agents the context AI coding tools are missing
• How the SDLC is changing: blurring boundaries between spec, design, and code; shorter cycles; and why CI is now the place where you find out if an idea was worth building
• Harness shipping 3-4x more capabilities per month than 18 months ago, with the same engineering team
• Whether software engineers are still needed — and why the answer is yes, but the ratio of humans to code is shifting fast
Learn more about WorkOS: workos.com and Harness: harness.io
Jyoti Bansal, founder of AppDynamics and CEO of Harness, joins Michael Grinich at HumanX 2026 to talk about what happens to software delivery when code generation goes 5x, 10x, 100x — and why the outer loop is now the critical constraint.
Key topics:
•Why AI-generated code volume is overwhelming traditional CI/CD systems and what it takes to actually ship it
• The 30-40 layer outer loop: testing, security scanning, deployment, compliance, cost management — and why none of it was designed to scale with agentic code output
• The governance problem: downloading an agent into your pipeline without checks is the same root cause behind recent CI/CD-based cybersecurity attacks
• How Harness enforces governed pipeline execution: access control, policy frameworks, blast radius limits for every agent that runs
• The software delivery knowledge graph: capturing institutional knowledge about deployments, incidents, and test outcomes to give agents the context AI coding tools are missing
• How the SDLC is changing: blurring boundaries between spec, design, and code; shorter cycles; and why CI is now the place where you find out if an idea was worth building
• Harness shipping 3-4x more capabilities per month than 18 months ago, with the same engineering team
• Whether software engineers are still needed — and why the answer is yes, but the ratio of humans to code is shifting fast
Learn more about WorkOS: workos.com and Harness: harness.io
