Building the Foundation for the Agentic AI Era

The podcast discusses the widespread adoption of AI within organizations, focusing on strategies to integrate AI tools across both technical and non-technical teams. A key initiative involved training 12,000 employees at Block in the use of AI agents, including the development of an internal agent named Goose to assist with engineering tasks. To drive adoption, a change management strategy based on the 1-9-90 rule was implemented, leveraging 50 early-adopter "champions" who experimented with AI tools and shared best practices. The approach emphasized practical benefits, incremental integration, and education to overcome resistance and ensure sustainable use across diverse departments.

Another major theme is the collaborative development of open standards for agentic AI through the Agentic AI Foundation, hosted under the Linux Foundation. This neutral space enables companies like OpenAI, Anthropic, Google, and Block to co-develop critical protocols such as MCP (Model Context Protocol), AgentsMD, Agent Gateway, and A2A (Agent-to-Agent Protocol). These efforts aim to ensure interoperability, security, and global applicability of AI systems, with working groups addressing challenges like identity, watermarking, and agentic commerce. The discussion highlights the rapid evolution of AI agents - from single-task tools to complex, collaborative networks - and stresses the importance of open, inclusive standardization to prevent monopolization and support ethical, human-centered AI advancement.

• What if you trained a small group of AI champions within your solo dev workflow to systematize agent use?

  • Move: Dedicate 30% of your weekly development time to experiment with AI agents on one high-impact project (e.g., automated testing or API generation), document what works, and embed successful patterns into reusable scripts or templates.
  • Why Now?: AI tools have crossed a threshold in late 2025 where reliability and context understanding make sustained integration feasible - waiting further risks falling behind in velocity.
  • Expected Upside: You'll create a personal "champion loop" that turns sporadic AI use into a repeatable, automated advantage - boosting output by 2 - 3x without increasing effort.

• What if you built your own lightweight AI agent runtime based on open standards like MCP?

  • Move: Use open-source projects like Goose and MCP to assemble a minimal agent runtime that automates one core business function (e.g., customer onboarding emails or invoice processing) using tool-calling and context routing.
  • Why Now?: MCP is now stabilized through neutral governance (Agentic AI Foundation), making it safe to build on without vendor lock-in - perfect for solo devs needing interoperable, future-proof systems.
  • Expected Upside: You gain a differentiated, automatable service layer that can be productized or reused across clients, reducing manual work while increasing scalability.

• What if you joined a public working group to co-develop AI standards relevant to your niche?

  • Move: Identify and actively contribute to one open working group (e.g., Agent Gateway security or AgentsMD for codebase instructions) by submitting documentation, bug reports, or integration examples from your solo projects.
  • Why Now?: These groups are actively shaping protocols used by Google, OpenAI, and Anthropic - early participation gives you insider insight and influence before standards solidify.
  • Expected Upside: You position yourself as a subject-matter expert, gain early access to best practices, and build trust that can convert into consulting opportunities or product differentiation.

• Identify and empower 1% of your user or developer community as AI champions, giving them dedicated time to experiment and integrate effective AI tools into shared workflows.
• Focus on low-risk, high-impact AI use cases first - especially those that automate repetitive tasks - to build trust and demonstrate tangible value to non-technical users.
• Embed proven AI practices directly into codebases or tools (e.g., templates, configs, scripts) so broader adoption becomes seamless without requiring individual initiative.
• Contribute to or leverage open standards like MCP (Model Context Protocol) to ensure AI agents can securely and interoperably connect with existing tools and systems.
• Organize small, informal knowledge-sharing sessions (e.g., brown bags) to disseminate learnings from early AI adopters and reduce resistance across teams.