Every association leader I talk to has “leverage AI” somewhere on their strategic plan.
Boards are asking about it.
Members expect it.
The pressure to do something with AI is real.
But here’s the problem: most associations are implementing AI in ways that won’t move the needle on member engagement. They’re checking a box without creating real value.
The associations that will win in the next five years aren’t the ones that adopted AI first. They’re the ones that adopted it strategically — using AI to solve genuine member problems rather than just adding technology for technology’s sake.
The American Geophysical Union (AGU) offers a compelling example of what serious AI implementation actually looks like — and the results speak for themselves.
Key Takeaways
- Most associations use basic demographic segmentation and outdated survey data—checking a box without creating real member value.
- Early career professionals were 20% more likely to interact with AI-driven recommendations, and members discovered previously unknown peer connections they never would have found through traditional search.
- Start with a real member problem, use behavioral data over declared data, personalize to individuals (not segments), and ensure quality data infrastructure.
- rasa.io has built AI personalization for associations for over a decade, with 220+ associations using proven infrastructure with demonstrated results.
The Problem with Surface-Level AI
Before we look at what works, let’s be clear about what doesn’t.
Basic demographic segmentation that gets called “personalization.” Sending different emails to members based on their job title or membership tier isn’t personalization. It’s segmentation. And in a world where companies like Netflix employ algorithms that know exactly what each viewer wants to watch next, segmentation feels increasingly outdated.
Using AI simply to generate text. If you’re using AI to write email copy without feeding it any data about what your members actually care about, you’re just automating mediocrity. The emails might be grammatically correct, but they won’t be relevant or interesting.
Relying on survey data that decays the moment you collect it. Member interests change constantly. The preferences someone indicated in an onboarding survey two years ago — or even two months ago — may bear little resemblance to what they care about today. Static data produces static results.
Manual tagging and taxonomy systems. Having staff manually categorize content and match it to member interest codes is time-intensive, inconsistent, and fundamentally limited. It can’t scale, and it can’t capture the nuance of what individual members actually find valuable.
These approaches aren’t wrong, exactly.
They are just not enough.
They represent AI as a feature rather than AI as a transformation.
What AGU Did Differently
The American Geophysical Union faced a challenge familiar to many associations: they had an enormous wealth of content — hundreds of thousands of peer-reviewed articles and scientific abstracts — but members couldn’t easily surface content that actually mattered to them.
“People were saying, ‘We know the information is there, we just can’t find it,'” explains Thad Lurie, AGU’s Senior Vice President of Digital and Technology.
AGU’s membership skews young — more than 50% are students and early career professionals. These members don’t have mature professional networks. They’re actively looking for connections to peers doing similar research. And they have high expectations for personalized digital experiences because that’s what they encounter everywhere else in their lives.
Rather than implementing a surface-level AI solution, AGU built something fundamentally different using rasa.io’s AI personalization engine.
The approach moves beyond the traditional taxonomy checkbox model. Instead of comparing one member’s self-reported interest codes against another’s, the AI analyzes the full text of every abstract and article a member has published or engaged with — and matches it against the entirety of content and people in AGU’s ecosystem.
“Now what we’re doing, instead of comparing your checkboxes to their checkboxes, we’re comparing the entirety of every abstract and article you’ve ever published with AGU against the entirety of the text for each abstract or each other person’s entirety of what they’ve published with AGU,” Lurie explains.
This creates what Lurie calls a “fingerprint” for each piece of content and each member — a unique signature based on actual behavior and genuine interest rather than self-reported preferences that may be outdated or incomplete.
The Results
The impact has been significant. Early career professionals — the demographic AGU most wanted to engage — were 20% more likely to interact with AI-driven recommendations than with traditional communications.
But the more telling result was qualitative. When researchers saw connections to peers they didn’t know existed, “their excitement was palpable,” according to Lurie. The feedback started pouring in almost immediately. The AI was surfacing collaborators and content that members would never have discovered through traditional search or self-directed browsing.
The initiative earned Lurie the Association Trailblazer Award from CEO Update and digitalNow’s Innovative Leadership Award — recognition that this approach represents something genuinely new in how associations can serve their members.
“I think what we’re doing is trailblazing,” Lurie says. “And I would love to put this in the hands of other organizations as quickly as we can.”
Personalized Email Engagement Using AI
With the rasa.io’s Campaign tool, create email campaigns
tailored to each individual.
What This Means for Your Association
AGU’s success points to several principles that apply broadly:
Start with a real member problem, not a technology mandate. AGU didn’t implement AI because their board told them to. They implemented it because members literally couldn’t find the information they needed. The technology served a genuine purpose.
Behavioral data beats declared data. What members actually engage with tells you far more than what they say they’re interested in. AI that learns from behavior — opens, clicks, time spent, content consumed — builds an increasingly accurate picture of each individual over time.
True personalization means segments of one. The goal isn’t to sort your members into five or ten buckets. It’s to treat each member as an individual with unique interests, needs, and preferences. That requires AI that operates at the individual level, not the cohort level.
The data infrastructure matters. As Lurie puts it: “You must have the data. AI is only as effective as the quality and depth of your existing datasets.” Associations sitting on rich content libraries and years of member interaction data have a significant advantage — if they can connect that data to AI that knows how to use it. Tools like AI data platforms can help unify and vectorize your association’s data to feed personalization engines.
This isn’t experimental anymore. rasa.io has been building AI personalization technology for associations for over a decade. AGU’s award-winning implementation runs on proven infrastructure used by more than 220 associations. This is production-ready technology with demonstrated results.
The Strategic Question
The real question for association leaders isn’t whether to use AI. It’s whether to use AI in ways that create genuine member value — or to implement surface-level solutions that check a box without changing anything meaningful.
AGU chose the former. The awards, the engagement metrics, and most importantly, the member response all validate that choice.
What will your association choose?
*Ready to see what AI-powered personalization could look like for your members? Learn more about rasa.io Campaigns.









