> Markdown version of [/videos/100249-ai-wrote-the-code-now-what?t=685](https://www.wearedevelopers.com/videos/100249-ai-wrote-the-code-now-what?t=685). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# AI Wrote the Code. Now What?

Jyoti Bansal warns that generating 2.5x more code is useless if delivery remains bottlenecked. Discover how risk-based autonomy and AI agents solve the velocity paradox.

- **Speakers:** [Jyoti Bansal](https://www.wearedevelopers.com/@jyoti-bansal), [Daniel Gebler](https://www.wearedevelopers.com/@daniel-gebler)
- **Event:** World Congress 2026 Europe
- **Published:** July 10, 2026
- **Duration:** 31:27
- **URL:** https://www.wearedevelopers.com/videos/100249-ai-wrote-the-code-now-what

## Summary

The rapid adoption of generative AI tools has birthed the AI velocity paradox: developers are generating 2.5x more code, yet production delivery has only increased by 15 percent. Jyoti Bansal, CEO of Harness, explains that while the inner loop of software creation has drastically accelerated, teams are severely bottlenecked by a human-speed outer loop encompassing testing, security audits, governance, and compliance. Solving this imbalance requires shifting toward risk-based autonomy, where a system dynamically evaluates code complexity, test coverage, and author experience to decide whether an AI agent can safely automate the deployment or if human-in-the-loop oversight is necessary. However, deploying multi-agent systems demands immediate codebase hygiene; organizations must refactor legacy monolithic applications into smaller components with clean APIs to reduce the strain on AI context windows and implement explicit AI guidance files inside repositories. As AI assumes the burden of raw code generation, the modern developer's role is evolving from manual implementation to architectural orchestration—acting as a manager who directs specialized agents responsible for coding, testing, and validation. This deep integration is rendering traditional five-to-seven-person agile units obsolete in favor of highly leveraged, micro-agile teams of up to three engineers. Finally, to ensure these digital transformations are economically viable, engineering leaders must discard arbitrary token budgets in favor of granular ROI tracking, meticulously correlating the aggregate cost of agent and developer tokens to the tangible business value of successfully shipped features.

**Keywords:** ai velocity paradox, software delivery lifecycle, outer loop bottlenecks, risk-based autonomy, ai token roi tracking, codebase refactoring for ai, autonomous agent orchestration, human-in-the-loop governance, context window optimization, micro-agile engineering teams, shift-left security, production deployment automation, software compliance checks, harness platform, ai adoption metrics

## Chapters

1. **Understanding the AI velocity paradox in software delivery** (00:00) — Generating significantly more code with AI creates a bottleneck when outer loops remain constrained to human speeds.
1. **Implementing risk-based autonomy for automated quality control** (02:43) — Applying AI automation based on code risk levels ensures safe deployments without manually blocking low-risk changes.
1. **Assessing code risk with multi-variable scoring systems** (04:16) — Evaluating factors like reviewer provenance, test coverage, and code complexity determines the exact level of required human oversight.
1. **Refactoring legacy codebases for effective AI agent integration** (05:57) — Breaking large repositories into smaller units with clean APIs allows AI agents to reason effectively within limited context windows.
1. **Tracking token usage for accurate AI cost management** (08:14) — Tying developer and agent tokens directly to specific business outcomes and issue trackers accurately measures software delivery returns.
1. **Addressing the post-coding bottlenecks in software delivery pipelines** (11:25) — Automating mandatory post-coding tasks like security scanning, compliance checks, and deployments is required to handle increased code volumes.
1. **Adapting shift-left principles for high-volume code generation** (13:49) — Empowering individual developers with automated testing and security pipelines ensures rapid processing of exponentially larger AI code outputs.
1. **Transitioning to smaller, autonomous full-stack engineering teams** (15:35) — Empowering engineers to act as managers guiding multiple automated tools allows highly capable small teams to operate more efficiently.
1. **Embedding intelligent agents into continuous integration pipelines** (17:12) — Upgrading traditional automation scripts with intelligent workflow agents provides active reasoning across every step of the continuous deployment pipeline.
1. **Establishing multi-agent architectures for trustworthy code validation** (19:59) — Using separate and audited models for writing code and testing code prevents automated shortcuts and ensures strict policy governance.
1. **Deploying AI session monitoring tools across enterprise environments** (21:59) — Distributing lightweight tracking tools via mobile device management solutions enables unified visibility into developer AI interactions across the organization.
1. **Preventing organizational coding knowledge gaps during AI adoption** (22:57) — Enforcing strict code review practices ensures both junior and senior engineers remain deeply connected to the complex architectures they manage.
1. **Managing code ownership for internal citizen developers** (24:43) — Allowing business users to rapidly prototype applications requires a clear handover protocol to professional engineers for reliable production-grade scaling.
1. **Balancing artisanal coding skills with automated agent oversight** (26:29) — Combining deep algorithmic expertise with efficient oversight of automated assistants produces highly capable and competitive modern software engineers.
1. **Focusing on true business returns for AI adoption initiatives** (28:14) — Measuring software delivery success through concrete efficiency improvements and revenue growth prevents wasteful spending on arbitrary AI consumption goals.

## Related Moments

- [Balancing developer autonomy with the adoption of coding agents](https://www.wearedevelopers.com/videos/100198-the-last-mile-of-ai-from-prototype-to-production) (from "The Last Mile of AI: From Prototype to Production")
- [The promise and risk of AI coding agents](https://www.wearedevelopers.com/videos/100277-what-production-knows-closing-the-loop-between-ai-agents-and-the-systems-they-build) (from "What Production Knows: Closing the Loop Between AI Agents and the Systems They Build")
- [Motivations for adopting AI to enhance developer productivity](https://www.wearedevelopers.com/videos/1266-navigating-the-ai-revolution-in-software-development) (from "Navigating the AI Revolution in Software Development")
- [Shifting developer workloads and realistic AI productivity gains](https://www.wearedevelopers.com/videos/1830-wearedevelopers-live-speculaitions) (from "WeAreDevelopers LIVE - SpeculAItions")
- [Impact of AI on the software development lifecycle](https://www.wearedevelopers.com/videos/100086-unlocking-the-ai-black-box-building-trust-in-the-era-of-agentic-production) (from "Unlocking the AI Black Box: Building Trust in the Era of Agentic Production")
- [Navigating developer bottlenecks and human accountability](https://www.wearedevelopers.com/videos/100265-fireside-chat-in-conversation-with-werner-vogels-cto-of-amazon-com) (from "Fireside Chat - In conversation with Werner Vogels, CTO of Amazon.com")

## Related Articles

- [Transforming Software Development: The Role of AI and Developer Tools](https://www.wearedevelopers.com/magazine/527-transforming-software-development-the-role-of-ai-and-developer-tools)
- [What is Agentic Programming and Why Should Developers Care?](https://www.wearedevelopers.com/magazine/625-what-is-agentic-programming-and-why-should-developers-care)
- [Exploring AI: Opportunities and Risks for Developers](https://www.wearedevelopers.com/magazine/522-exploring-ai-opportunities-and-risks-for-developers)
- [What is Software Engineering in the Age of AI?](https://www.wearedevelopers.com/magazine/640-what-is-software-engineering-in-the-age-of-ai)

## Related Jobs

- [Staff Software Engineer, Copilot Experiences](https://www.wearedevelopers.com/jobs/ext/164361-staff-software-engineer-copilot-experiences) at **GitHub**
- [Principal Product Manager, Agent Platform](https://www.wearedevelopers.com/jobs/ext/277541-principal-product-manager-agent-platform) at **GitHub**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**
- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**
- [Tribe Lead - ( Software) Engineering Centre of Excllence](https://www.wearedevelopers.com/jobs/ext/1475530-tribe-lead-software-engineering-centre-of-excllence) at **SD Worx**