SUMM

Eno Reyes introduces himself and shares his background in LLMs, noting experience since the release of GPT-3.5.
Factory focuses on agent-driven software development, believing it will transform the industry.
Current approaches mostly add AI to old human-centric tools, but a paradigm shift is needed.

True power of AI in software comes when most software lifecycle tasks are delegated to agents.
This requires intuitive task management, centralized context, reliable agent outputs, and infrastructure for parallel agents.
Factory has partnered with large organizations over two years to solve these hard problems.

Agents should not replace human ingenuity but be seen as climbing gear for scaling complex projects.
Demo provided: delegating a task to an agent ("droid") that gathers codebase data, references prior interactions, proposes a plan, seeks clarification, and generates a PR passing CI.
Success with agents hinges on providing complete and relevant context; lack of context is a key reason for AI failures.
Getting context into AI systems is critical, integrating everything from meeting notes to whiteboard photos as data sources for agents.

In agent-native organizations, agents aid in planning and product design, not just coding.
Agents can research, interpret product goals, and handle technical architecture references to assist with planning.
Collaboration with agents turns organizational documentation into living knowledge bases for current and future use.
Example: Using feedback transcripts and documentation to inform PRDs (Product Requirement Documents) and convert those into actionable tickets and roadmaps.

Typical processes and documentation (PRDs, design docs, meeting transcripts) become foundational knowledge for agents.
This shift turns documentation from a burden into an asset that future teams and agents can leverage.
Proper documentation enables agents to mimic and enhance team workflows by providing historical context and rationale.

AI agents contribute to site reliability engineering by aggregating incident data and organizational knowledge.
Droids can reduce incident investigation from hours to minutes by accessing distributed logs, documentation, and discussions.
Organizations experience faster incident response, fewer repeat incidents, and improved onboarding via persistent organizational memory.
Move observed from reactive to predictive operations, as agents help create better processes and system reliability over time.

Agents do not replace engineers but amplify their capabilities, letting developers focus on high-leverage orchestration rather than manual coding.
Future developers need skills in clear thinking and communication, both with humans and AI systems, to effectively leverage agents.
Emphasis on the importance of transitioning to orchestrating agent-powered solutions rather than routine code writing.

Attendees are invited to sign up for Factory’s platform with free trial tokens.
The platform is enterprise-focused, with security, auditability, and ownership built-in; potential risks and ownership of agent actions are addressed.
Security professionals and legal teams should consider questions of responsibility, auditability, and safe operation when implementing agent-driven systems.

Ship Production Software in Minutes, Not Months — Eno Reyes, Factory