Skip to content
Agentic Relations

The patterns, reframed

The 37 DevRel Patterns, Reframed for Agentic Relations

The patterns don't become obsolete. They bifurcate — and understanding which bucket each pattern falls into is the first step in adapting a DevRel program for the two-layer operating environment.

Every activity pattern in Developer Relations Activity Patterns was documented for a human audience. None of the 37 patterns becomes obsolete in the Agentic Relations era. But every one of them bifurcates — some requiring only the addition of an agent-facing variant, some requiring substantive rewriting, and some growing more valuable precisely because they are irreducibly human.

The three categories

Gains an agent layer (15 patterns): the human variant continues largely unchanged. A parallel agent-facing variant is required.

Requires substantive rewriting (8 patterns): the core intent survives, but implementation, participants, metrics, and sometimes the name must change.

Grows more valuable (14 patterns): irreducibly human patterns that increase in strategic value as AI floods every content channel with generic material.

All 37 patterns at a glance

The full pattern catalog from Developer Relations Activity Patterns, tagged with Agentic Relations status.

Article

Gains agent layer

Blog Post

Gains agent layer

Book

Gains agent layer

Guide

Gains agent layer

Newsletter (technical)

Gains agent layer

Reference Documentation

Gains agent layer

Sample / Example

Gains agent layer

SDK

Gains agent layer

Social Media

Gains agent layer

Tutorial

Gains agent layer

Forum

Gains agent layer

Open-Source Project

Gains agent layer

Survey

Gains agent layer

Technical Support

Gains agent layer

Recorded Video

Gains agent layer

Ambassadors → Agent Champion

Requires rewrite

Hackathon → Agentic Hackathon

Requires rewrite

Office Hours → Integration Debug Sessions

Requires rewrite

Training → Agent Direction Training

Requires rewrite

Code Review

Requires rewrite

Live Playground

Requires rewrite

Case Study

Requires rewrite

Workshop

Requires rewrite

Conference

Grows more valuable

Conference Session

Grows more valuable

User Group Network

Grows more valuable

Parties

Grows more valuable

Live Streaming

Grows more valuable

Customer Check-In

Grows more valuable

Customer Pre-Sale

Grows more valuable

Sponsorship

Grows more valuable

Webinar

Grows more valuable

Podcast

Grows more valuable

Partnerships

Grows more valuable

Swag

Grows more valuable

Community Newsletter

Grows more valuable

Boot Camp

Grows more valuable

Deep dive: Reference Documentation reframed

The Reference Documentation pattern exists to give developers the complete technical information they need to use a platform correctly. The implementation assumes a human reader: navigating a hierarchy, reading progressively, resolving ambiguity through context and prior knowledge.

In the Agentic Relations context, Reference Documentation must serve two readers simultaneously. The human reader still navigates, still uses context, still exercises judgment. The agent reader retrieves — it searches for specific sections as context for generation tasks. It cannot navigate ambiguity; it amplifies it into bugs.

The practical implication: every Reference Documentation page should be audited against a standard set of agent queries — "how do I authenticate," "what do I do with a 429," "what is the correct structure for X." If the page doesn't answer those questions directly and unambiguously, agents will generate incorrect integrations using it as context, regardless of how clearly it reads to a human.

Deep dive: Ambassador → Agent Champion

The Ambassador earns trust through authentic community presence. The relationship is the mechanism. Human developers trust Ambassadors because they know them, have seen them speak honestly, have observed them using the platform in real contexts.

AI agents don't operate on trust. They operate on pattern matching against whatever context they have. The "Ambassador" for the agent ecosystem is a practitioner who maintains the corpus of configurations, recipes, and schemas that agents draw on. I call this practitioner the Agent Champion.

The Agent Champion's currency is not reputation. It is a demonstrated success rate: the percentage of AI tool interactions with the platform that produce working integrations because the recipes the Agent Champion maintains are accurate. This is measurable in a way that Ambassador impact never was.

Read the full Agent Champion role description →

Deep dive: Hackathon → Agentic Hackathon

The traditional Hackathon evaluates what developers build when given creative freedom, tools, and time pressure. The implicit assumption is that build output is primarily a product of developer skill and creativity.

When AI coding agents are the primary builders, that assumption breaks. What gets built in an Agentic Hackathon is primarily a product of how well the team directs its AI tools — and how well the platform's agent-consumable infrastructure supports that direction.

  1. Direction as a first-class role. Each team has a designated Director who writes the prompts, evaluates agent output, and decides what to accept, reject, or redirect.
  2. Robustness over completeness. Submissions are stress-tested against error conditions and edge cases, not just demoed.
  3. Recipe output is required. Teams submit the prompt recipe library they developed during the build.
  4. Architectural decision logs. Teams document which AI decisions they accepted or overrode, and why.
  5. Platform monitors, not just mentors. Agent Champions observe team-AI interactions throughout, cataloging where the platform's agent-consumable infrastructure is failing.