Skip to content
June 11, 2024

More AI Use Cases for Associations: Beyond Content Generation 

On a recent episode of The Member Engagement Show, I was joined by Higher Logic’s Chief Technology Officer, Conor Sibley. Conor has spent almost three decades building influential companies where he designed and managed internet-scale platforms. He has in-depth knowledge of emerging technologies and is part of the team developing Higher Logic’s own AI toolset. In this episode, we chatted about AI – beyond the hype (and criticism) – to explore the varied ways it can be used by associations in realistic and practical ways.


Supporting Segmentation and Personalization with AI

“Everyone ends up thinking that AI means Chat GPT, right? And Chat GPT represents a generative kind of AI. But there are tons of other different types of AI that associations can bring to bear,” Conor points out.

For example, many people know about Large Language Models (LLMs), but fewer people have heard of Multimodal Large Language Models (MLLMs). LLMs work with text, while MLLMs will add audio, photos, graphics, and video. So you could use AI not just to write copy, but to create images and other media formats.

Meanwhile, AI has been used for a long time for many enterprise use cases. Think about how, when you log into a platform like Netflix or Amazon, those platforms know a ton about you, and because of that, they can offer highly personalized experiences. Conor expressed that technology vendors like Higher Logic should similarly be helping associations provide highly informed, highly tailored, and personalized experiences to their members.

To do that, you need information about your members – what they like, who they are, what kind of content they interact with – so you can segment them in refine your communication. Once you have member data to inform your strategies and segmentation, you can employ generative AI within a marketing automation platform to craft your marketing copy and content efficiently.

AI for Brainstorming and Context

You can also have conversations with AI chatbots to brainstorm ideas or collect research on topics you’re exploring or writing about. You no longer have to start from a blank page – you can edit and optimize an outline or draft that AI creates for you.

Tell the AI chatbot what you’re trying to accomplish, who the content is for, and what format the content should take, and AI can return thought-starting responses. AI’s ability to come up with ideas and draft content around those ideas, combined with the tools in your marketing automation system that help you logistically segment your audience and share that content, improve your ability to publish targeted, relevant content more quickly and at scale.

That expedited flow of content can then lead to better engagement.

Using AI to Navigate and Summarize Information

Another area where AI can be immensely helpful for associations is its ability to parse massive amounts of information. Conor reminds listeners that this is likely an area where humans need to be in the loop to fact-check – but AI can give those humans superpowers.

“One thing we have to watch out for is what’s called hallucinations – when the AI model generates incorrect or misleading information. Not only that, answers to questions given to an AI model aren’t deterministic. In other words, you can get different answers each time you ask the same question of the AI. So associations are probably going to want to keep some control over the conversation. AI use cases still require a human in the loop,” said Conor.

But with this in mind, imagine an association’s advocacy team is dealing with questions about a massive new policy paper. Members might want summaries or might be asking the association questions about specific points in the policy. At a basic level, AI is fully capable of taking the text and summarizing it. On a more advanced level, you could use an AI technique called retrieval-augmented generation (RAG) to enable an AI model to answer questions based on information from the policy paper.

RAG optimizes the outputs of large language models by allowing them to reference specific, authoritative knowledge sources outside of their general training data. It’s like having the AI cross-reference a source that you know to be reliable before it generates a response.

Here’s how it works:

  • Retrieval Component:
    • The chatbot first retrieves relevant information from external sources (e.g., databases, websites, documents, or APIs).
    • It uses this retrieved content as context to enhance its responses.
    • For example (from the scenario above) if a user asks about a new policy paper, the chatbot can fetch information from that policy paper.
  • Generative Component:
    • After retrieving relevant information, the chatbot generates a natural language response.
    • It can use large language models (like GPT) to create context-aware answers.
    • The generative component ensures that responses are in natural language.

Overall, RAG creates more informed, accurate, and contextually relevant chatbot interactions because it can leverage specific data beyond what’s included in general AI models.

So going back to the policy paper example, you could create a chatbot that references the policy paper so that you could ask questions and get quick answers. And staff could vet those answers before replying to members. You’re keeping humans in the loop, but letting AI help those humans do the job faster.

Elevating Your Internal Knowledge Base with AI

AI can also be used to address departmental silos.

Smaller organizations often innovate and grow faster because there are fewer people, which means less separation of information. While there are many benefits as an organization grows and takes on more staff, one challenge is that in large organizations, you often have to wait for answers – different departments have different specialized knowledge and responsibilities.

So if a member reaches out with a question, you might not personally know the answer. And sometimes by the time you get an answer, that member is frustrated.

Some of the associations I’ve worked for or heard about address this challenge by maintaining internal FAQ documents or knowledge bases – but even these can present a challenge. As they become larger, information gets overlooked because staff are scrolling through a multi-page document.

Imagine pairing that internal knowledge base with an AI engine to create a chatbot based on your own answers to FAQs. Organizational knowledge is now quickly available across the organization – instead of tracking down a busy colleague (who might be tied up in a meeting) or searching internal documents when a member asks “What’s the member discount for this event?” you could ask your internal chatbot and get an answer.

There are incredible productivity gains possible from allowing people to self-service their way to finding the answers they need. And association members are infinitely happier when they get quick answers.

With testing to ensure your confidence in the responses these tools can generate, you could even set up a similar public AI chatbot in your online community for members to interact with directly. A member could then ask a question of your community AI chatbot and get an answer based on the information stored in your community, rather than having to search through threads. And the easier it is to get answers, the more members will return to your community, rely on your association, and engage.

AI is a Game-Changer for Translation and Multilingual Experiences

Another exciting AI use case Conor explores in this episode is the power of AI for creating seamless multilingual experiences.

Trying not to get too in the weeds of how it works: AI, particularly advancements in deep learning, is significantly enhancing semantic embeddings and, consequently, our ability to both search and translate content.

What is semantic embedding? Semantic embedding is a technique in natural language processing (NLP) where words, phrases, or entire texts are represented as vectors in a numerical space. Picture dots on a graph that are either closer to or farther from each other and to the X and Y axes to represent where they stand in relation to each other. These vectors capture the semantic meaning of the text by encoding information about context, syntactic structure, and relationships between words. The primary goal of semantic embedding is to translate textual information into a format that machines can efficiently process and understand.

Why does semantic embedding improve translation? Being able to improve a system’s ability to understand the subtleties of the source language, including synonyms, idioms, cultural references, and polysemous words (words with multiple meanings) leads to translations that are not just word-for-word but convey the intended meaning more accurately (and more quickly).

So, in Conor’s example, previously things like search engines categorized content based on keywords. If you were to search for “cat,” they might return articles about cats, but not recognize that articles about “felines” or “kittens” are also relevant. With semantic embedding, systems better understand the meaning of words, rather than just making direct matches back to specific keywords.

So, if you’re a French speaker searching for “le chat” (french for cat), you’re currently only getting results in French. But with search engines and other systems imbued with semantic meaning, it could soon be possible to get back results for “cat,” “gato” (Spanish), “gatto” (Italian), “katze” (German), etc. And, better yet, organizations could use tools on their website that automatically translate their content.

This is a game-changer for organizations serving international audiences – or where laws or regulations (like in many parts of Canada) require organizations to provide their website and marketing materials in more than one language). In a couple of years, we’re likely to see associations being able to have content appear in any desired language – sometimes in even different versions of the same language such as French and Spanish. All your content would be delivered to interested audiences, regardless of the source’s and searcher’s language.

The Power of AI Without Putting Data at Risk

Another topic Conor covered on this episode of the podcast was Higher Logic’s newly available AI features:

  • AI Assistant: Your new virtual team member! Draft emails quickly and easily, all based on your prompts. Say goodbye to manual work and hello to efficiency!
  • Bulk Upload: Simplify content management by effortlessly uploading and organizing your multiple community content entries.
  • AI Suggested Tags: Let AI suggest relevant tags for better discoverability and engagement in your community.
  • Smart Newsletter: Leverage AI-driven newsletters with content based on subscriber actions.

Though many associations have policies that don’t allow the use of content coming from Chat GPT or Claude or others out of concern for those public AI systems “scraping” the organization’s data for use in training their models, Higher Logic’s tools offer a private enclave.

Higher Logic’s AI tools do not get trained on your data. Any of the information we help generate and any of the content you input isn’t leaching out of your environment. You’re not at risk of that content showing up inside a future model and your data isn’t being co-opted. Our focus is on tools that save you time, safely, all contained inside of your Higher Logic platform.

Interested in learning more about Higher Logic's new AI features, or raising your hand to be an early adopter?

Kelly Whelan

Kelly Whelan is the Content Marketing Manager for Higher Logic. In this role, she develops content to support association professionals and advise them on member engagement and communication strategy. She also hosts Higher Logic’s podcast, The Member Engagement Show. She has ~10 years of experience working in marketing for associations and nonprofits.