1 of 27

Can Al strengthen our communities, or will it hollow them out?

Al Meets Community

Moritz Arendt - Head of Product

2 of 27

Moritz Arendt

  • Head of Product at Open Social.
  • 10+ years of experience with Drupal.
  • My background in Psychology and UX
  • I worked the last 1 ½ years to bring transparent, privacy-conscious AI into our Community product

3 of 27

We are an EU-based tech company with offices in Amsterdam and Enschede, The Netherlands.

Our mission is to support and empower �expert communities worldwide in their digital transformation journey.

It is our vision that the web is place rooted in pro-privacy, anti-monopoly, and open collaboration principles. We create trusted digital platforms where ownership and data sovereignty are paramount.

�Trusted by the world’s most ambitious organizations:

Open Social is here to serve mission-driven organizations around the world since 2010

4 of 27

Our products will work together for your

Community Collaboration Platform

5 of 27

The goal is to augment human connection, not replace it.

Built to assist content creators.

Built to enhance the community member experience.

Creator

Gaia Al

Members

Built to support platform managers.

Manager

Open Gaia Al is a framework of Al assistants and agents Social's built into your Community Collaboration Platform.

Our Approach: Al as a Community Co-Pilot

6 of 27

Global challenges with online communities

Content & Moderation

User Navigation & Onboarding

Community Knowledge Extraction

7 of 27

SparkBlue by the numbers

7

40.000+

expert users across 170+ countries

100+

sub-communities and networks

20.000+

contributions to discussions and consultations

4.000+

documents, white papers etc. in many different languages

8 of 27

Dangers of AI for community

While AI is equipped to help it can damage the Human relationships that communities are based on

We encountered different conflicts during implementation

9 of 27

Community managers are drowning in content.

Al offers automated moderation to save hundreds of hours.

But can an AI’s judgment build the same trust� as a seasoned moderator?

The First Tension: Efficiency vs. Trust

10 of 27

Automation must empower, not replace. The best tools give

moderators superpowers while keeping them firmly in control.

Automatically scores the relevance of comments (0-100) to assist moderators in identifying spam, profanity, or low-value content.

Feature: Gaia Relevancy Score

This prodoctxikes users sive going togingihs stcn iml-uhortlravin?irde:show Isavt you

teok to promotee are semitranium and-moreenlection.

60

Relevancy score

1days ago

Denny Callier

Reply

1

9

How are thevsel prolorit who people directly lest using sut a comment?

20

Relevancy score

1 days ago

Doriane Romano

Reply

1

6

This is on to relevance of comments (0-100) to assist modera low-value content.

Feoboncit

Site Admin

S

Add comment

Gaia Relevancy Score

80

Relevancy score

The Tool

Finding the Signal in the Noise

11 of 27

A chatbot can provide instant answers, recommend content and guide users. �But when does helpful assistance cross the line and take away a user's agency to explore, discover and make their own connections?

The Second Tension:

Assistance vs. Agency

12 of 27

An effective Al must be trained and prompted not just on data, but should be specialized and understand

unique culture, content and vocabulary of the community it serves.

1.

Teaching an Al to Speak the Language

of a Community

The Gaia AI Chatbot will allow your users to ask questions on the platform regarding content, help them with onboarding or explain how the community works.

13 of 27

Al can help members write posts and generate summaries. �

But how do we assist members without making the platform feel �robotic, sterile or inauthentic?

The Third Tension:

Automation vs. Authenticity

14 of 27

Automation is useless if the output doesn't meet the user's need for professional, trustworthy depth.

Automatically compile digests of discussions and posts.

Assist creators by generating drafts, summaries, and adjusting tone within the editor.

Feature: Gaia Content Summary & Editor Assistant.

The Feature

The Reality of Al-Generated Content

15 of 27

Gaia AI Launching features

Content Summaries

Gaia Relevancy Scores

Gaia Editor Assistant

Gaia Chatbot

16 of 27

Embrace the Feedback

Loop.

Al Must Learn the Language

of the Community.

Build Trust Through

Transparency.

Design for Co-Pilots, Not

Autopilots.

Our Playbook for Human-Centered Community Al

17 of 27

Co-Creation: We build with our clients and the community, not just for them.

Open Architecture: We can build with flexible, non-proprietary tools, preventing vendor lock-in.

Shared Values: Our culture is built on collaboration and transparency.

Drupal is uniquely positioned to lead in ethical, human-centered Al.

The Future of Al is Open & Collaborative

18 of 27

We are committed to build with Drupal AI

The Drupal community is moving fast and we will be part of it with

With that let’s build AI that puts people, privacy, and open collaboration first.

19 of 27

Roadmap 2026

Gaia AI

20 of 27

Making Gaia AI actionable

User impact score

AI Spam prevention

Summary output

Gaia Search

21 of 27

Gaia Search

For community members, finding the right answer in a large community with PDFs and other documents is slow and noisy. Gaia Search will combine RAG-based semantic search with agentic, multi-step search to surface relevant content items and summaries results.

RAG lifts quality by matching on meaning instead of keywords, while agentic search can clarify intent, chain multiple lookups, and synthesize a final answer. The impact: fewer dead-end searches, faster time-to-answer, and less moderator intervention needed in “where can I find X?” type questions.

22 of 27

Gaia Smart Filters

In large resource libraries, filters quickly become unmanageable. In the contexts like IATA’s SIH, where users face over 70 different filter options.

Gaia Smart Filters will act as an intelligent layer on top of the existing taxonomy: instead of forcing users to manually click through dozens of fields, Gaia interprets a natural-language query and automatically applies the right combination of filters.

Long term the agent can propose clarifying questions (“Are you interested in guidance, case studies, or regulations?”) and offer one-click ‘filter recipes’ tuned to IATA/SIH use cases (e.g. “Regulatory updates by region,” “Practical implementation guides,” “Best practices by airport size”).

For community managers, Smart Filters could generate insights into which combinations are most used or missing, helping refine the 72-filter model over time.

23 of 27

Gaia User Impact Score

For community managers, it’s hard to see who truly adds value. User Impact Score aggregates Gaia’s relevancy scores per user (and longer term) per topic to show who delivers high-quality input.

This makes it easy to identify subject-matter experts, ambassadors, and hidden contributors that deserve recognition or involvement in campaigns and co-creation. Over time, User Impact Score becomes a health signal for topics and working groups, informing Engagement Studio targeting and Lumina insights.

24 of 27

Gaia Spam Prevention

As communities scale, spam and low-quality (often AI-generated) noise can quickly drown out real discussions. AI Spam Prevention uses Gaia’s Relevancy Scoring to automatically flag content that looks like spam, abusive language, repetitive AI text, or off-topic noise.

Community teams can define automated workflows such as “auto-hide below threshold,” “send to moderation queue,” or “notify user with a warning,” dramatically reducing manual triage. The result is a cleaner platforms by default, without requiring a proportional increase in moderation capacity.

25 of 27

Gaia Summary Outputs

Community managers spend a lot of time manually curating content for newsletters and dashboards. Gaia Summary Output automatically generates summaries of key content, event and resources. In this next step we want to exposes them as:

  • Email-ready snippets for Engagement Studio campaigns (e.g. “This week in Climate Action Working Group”), and�
  • Configurable dashboard blocks that highlight top conversations, decisions, or contributors.

Summaries are driven by Relevancy and natural language prompts, so email campaigns and dashboards focus on the most meaningful activity.

26 of 27

27 of 27

Thank you