AI for Allyship and Inclusion: Supporting CNCF Communities with Context-Aware AI

Using GenAI to draft community emails or presentation assets should save organizers time. However, default models often create a manual review loop where we can’t miss those nuances. Standard LLMs frequently fall back on biases: generating meetup images of only white men, missing gender-neutral nuances in non-English languages, or adopting an exclusionary tone that alienates underrepresented groups. To prevent this constant rework, we need to embed allyship skills and context directly into our AI tools. This talk explores how to design and build context-aware AI assistants tailored for community operations. We will look at how to inject the right cultural context, linguistic rules, and guardrails so your AI tools can catch exclusionary language and evaluate event assets for representation bias out of the box. Attendees will walk away with actionable blueprints to teach their own AI tools these vital allyship skills, shifting the operational burden away from manual human review.

Transcript

1. KCD UK · 19 OCTOBER 2026 AI for Allyship and Inclusion Supporting CNCF Communities with Context-Aware AI Imma Valls Staff Developer Advocate, Grafana Labs
2. What if the invitation is the problem? “…an exclusive community of top-tier engineers…” AI Socratic community description, generated with AI 01
3. GRAFANACON · K6 WORKSHOP AI didn't challenge the bias. It amplified it. Generated community illustrations Default prompt Developers: young men Operators: young women Everyone: white Explicit prompt 02
4. CLOUD NATIVE BARCELONA · EVENT DESCRIPTION AI won't ask about the Code of Conduct. We have to. Default prompt Prompt with inclusion guidance No Code of Conduct, not flagged Code of Conduct included 03
5. A COMMUNITY ALLYSHIP FRAMEWORK Early prototype What if we gave AI our community values? 01 02 03 04 Inclusive communication Gender-aware multilingual writing Representative imagery Community safeguards Check roles, age, gender, skin tone Code of Conduct, reporting, accessibility Welcoming, not elitist or intimidating English and Spanish, each on its own terms Fewer repeated corrections. Earlier flags. Not unbiased AI. Inspired in part by Jorge Ordovás's DEI Companion github.com/joobid/DEI-companion 04
6. BEYOND CONTENT From generating content to evaluating events ILLUSTRATIVE EXAMPLE Event review · public information Example community event Code of Conduct linked Reporting and enforcement explained Inspired by publicly available CHAOSS guidance on DEI event badging CHAOSS DEI Event Badging Not an official CHAOSS assessment, endorsement or badge Next: test with DevOpsDays Barcelona and KCD Edinburgh Not real results Accessibility information Scholarships or participation support Visible inclusion measures Verified Needs clarification Not found Verified Needs clarification Reports what is publicly visible. Not a badge or a verdict. chaoss.community/diversity-and-inclusion-badging 05
7. THE EMERGING ARCHITECTURE Make good intentions repeatable AI-assisted Input Community values and guidance CREATE REVIEW Text, images, announcements, assets Public event info against explicit criteria Aim: Claude, ChatGPT, Gemini Starting in Claude Always Human judgment and action Still evolving. Doesn't solve bias or replace human oversight. 06
8. ALLYSHIP IS SOMETHING WE PRACTICE AI won't make our communities inclusive. People will. Explore, adapt, contribute 01 02 03 Review what AI creates, especially wording and imagery Give AI explicit community context and inclusion guidance Ask what might be missing from the events and spaces we create github.com/immavalls/ai-for-allyship Also: my inclusive speaker rider eyeveebee.dev/creating-my-inclusivespeaking-rider 07
9. Thank you Slides bit.ly/ai-for-allyship-2026 08