• 6 agentic knowledge base patterns emerging in the wild AI agents have become the software industry’s latest fascination. • Backed by large language models (LLMs), this new class of AI is unlocking data-driven decision-making and autonomous actions, transforming enterprise software practices and business workflows in the process. • However, it wasn’t always this way. • According to Ajay Prakash, a senior staff software engineer at LinkedIn, AI agents initially faced a major gap. • “Out of the box, AI coding agents weren’t effective,” Prakash tells The New Stack. • They lacked context and awareness of internal systems, frameworks, and practices, he adds.

Article Summaries:

  • AI agents are reshaping enterprise software, but early versions lacked context and internal system awareness. To address this, companies are building “agentic knowledge bases” that supply AI agents with institutional data, runbooks, tools, and project histories, enabling more reliable, verifiable outcomes. The New Stack reports six emerging patterns, from LinkedIn’s Contextual Agent Playbooks and Tools (CAPT) that enforce coding style and automate bug fixes, to integration knowledge centers that standardise data‑automation workflows. These systems vary in architecture and scope, but all aim to embed organization‑wide standards into AI agents, improving consistency and accountability across domains.
  • AI agents are increasingly integral to enterprise software, but early versions lacked the context needed for effective operation. To bridge this gap, organizations are building “agentic knowledge bases” that supply AI agents with internal data, runbooks, tools, and project histories. These systems enable agents to understand company‑specific conventions, access internal tools, and execute tasks such as debugging or integration maintenance. The New Stack reports six emerging patterns, including playbooks for coding assistants, integration knowledge centers, and purpose‑built layers that enforce accountability across domains. While architectures vary, the common goal is to embed organization‑wide standards into AI workflows, improving consistency and verifiability.

Sources: