Most nonprofit associations struggle with maintaining up-to-date and meaningful member profiles. Profiles typically contain only basic contact and billing information, leaving organizations with a limited understanding of their members. To address this issue, our team decided to create a proof of concept (PoC) aimed at enriching member profiles using large language models with just a few data elements as the starting point. The results were fascinating and promising, as they offered great results at a very low cost.
Our primary motivation was realizing that most associations have shallow and outdated member profiles. We wanted to see if we could create more detailed and useful profiles that included areas of interest. By improving the member's profiles, we hoped to make interactions with members more personalized and engaging as well as providing an overall perspective about the common areas of interests of members as whole.
The PoC focused on addressing the key issue of outdated member profiles but without taking the traditional path of lengthily and costly member surveys. Our goal was to enrich the profiles of every member in the organization based on information already available online - so that the membership, communications and events teams could put it to good use.
This new approach is not only much more cost effective but it also has the potential to work for a larger percentage of the member base when compared with those that decide to answer the survey.
We leveraged LLMs in several ways, combining them with internet searches to gather profile information available online. The process involved using AI to summarize findings and validate their pertinence. Additionally, the LLM was used to summarize the areas of interest for each member, providing a more comprehensive and useful profile. See examples below.
To ensure privacy and security was maintained, we used only publicly available data to produce the AI enriched profiles - using data that is available through a search engine such as Google. Our AI agent, which used the LLM through its API, ensured that no data was retained by the LLM nor used training purposes.
The PoC yielded impressive results. In just 10 seconds per profile, using only three input parameters—name, organization, and email—the LLM was able to generate up-to-date profiles that were far superior to the the ones most associations would have. As a bonus, we often obtained the members' LinkedIn URL, which is a useful data point and also facilitated our validation processes to ensure that the correct member was retrieved.
Name: Heather Murray
Organization: Objeto Inc
Email: heather@objeto.ca
Summary:
Heather Murray is a founding partner at Objeto Inc, an IT consulting firm for nonprofits. With over 15 years of experience in business analysis, stakeholder management, and project management, Heather has been instrumental in driving business process improvements and design at Objeto. Her professional journey includes a significant role as an Account Manager for Commercial and Corporate Segments at BMO Capital Markets before co-founding Objeto in 2016. Heather is based in the Greater Toronto Area, Canada.
Areas of Interest:
- Business Analysis
- Stakeholder Managemen
- Project Management
- Business Process Improvement
- IT Consulting for Nonprofits
LinkedIn Profile: https://www.linkedin.com/in/hmurrayto/
One potential challenge with this approach is that not all members have data available online, which means the process won't yield the same results for everyone. However, we expect that this approach will work very well for professional associations where members have profiles online and/or their own websites. Further, it should be noted that online surveys will be far from providing full coverage. Finally, we recommend that this processes should include human validation as it may not always yield the correct results, particularly where it may be difficult to find the exact member with the given data points.
Even though we leveraged slightly more complex technology than what's generally available, our PoC also demonstrated that anyone can achieve decent results using the same prompts we used leveraging the latest version of ChatGPT - as long as it enables Internet access. The larger benefit lies however, in automating this process for all members, enabling organizations to gain a comprehensive understanding of their overall membership. Before some of these technologies were available it would be much harder to go from a black box of membership to a segmentation by areas of interest in just a couple of days. Having up-to-date areas of interest for your members allows you to drive targeted content, better event planning, and more pertinent course offerings - all from 3 initial data points and publicly available information.
We evaluated the success of the PoC by testing it with members we knew, including friends and ourselves. We took it a step further after generating the profile. We asked the LLM to suggest a recent article based on the areas of interest and to an draft email offering it as reading suggestion. The results were so compelling that our test uses were eager to read the suggested content right away! Try a reduced version of our prompt with instructions below.
Based on the success of the PoC, we plan to develop a service for our nonprofit customers that can run this agent in the background, generating alternate profiles for their members. This service has the potential to significantly enhance organizations' ability to understand their members without lengthy surveys and to engage them in more personalized and interesting ways.
This PoC represents a significant step forward in using AI to improve member profiles for nonprofit associations. By enriching member data, organizations can gain a deeper understanding of their membership, enabling more effective and personalized engagement. As we move forward, we look forward to helping our nonprofit clients make the most of this technology to support their missions and build stronger member relationships.
You can take the prompt below as is, and paste it into ChatGPT. Just modify the values that are between square brackets, which are: some context about your association and the 3 data points pertinent to the member. Start by trying our prompt, and then feel free to modify it to see if you can get even better results.
👉 Replace [Technology] by your association area of focus and the 3 member data points (name, organization and email):
You will access the internet to find the right information and complete the task below:
--Task:
As content specialist from a non - profit association specialized in [Technology].Follow ALL the steps below and don't forget to write the email at the end.
1. Create a summary profile for the member below by browsing the internet.Find what their main areas of interests are, write a small summary, capture a bulleted list of their interests.Do not create your own information.
2. Using the internet and based on the information you found, look for a recent article that matches their interests and provide a summary of the article.
3. Test the URL of the article to make sure it is accessible and ensure it has no paywall or is a page not found.
4. Write an email to the person below with the article you found and a brief explanation of why you think they would find it interesting.
- Avoid sales tone and focus on providing value.No Buzzwords.No Jargon.No Fluff.Use a straight to the point style.
- Avoid general statements such as 'I came across your impressive work'
- Explain why the information is valuable to them.
- Consider the organization they work for and their role.
- Avoid articles directly about the organization they work for.
- Avoid articles where the member is the author.
--Member Information:
Name: [Heather Murray]
Organization: [Objeto Inc]
Email: [heather@objeto.ca]
--Example of Member Summary:
Jane Doe is a highly skilled therapist who has been supporting individuals, couples and families in Toronto and surrounding areas for years.Registered provider with social services, and many insurance companies.Whether the concerns stem from communication, stress, grief, trauma, mental health, or addiction, Jane uses a variety of counselling techniques to help create the changes you are seeking.In person, secure video or phone sessions are available.
Her areas of interest include:
- Family Counselling
- Relationship Counselling
- Individual Counselling
- Emotionally Focused Therapy
Her linkedin profile: https://www.linkedin.com/in/jennifer/
--Example of a good email:
Hi Mario,
I hope you are keeping well.Based on your interests in Healthcare and Technology, I thought you might find this article on the impact of trauma on children's mental health interesting.
It provides a unique perspective on the topic and I believe it could be valuable to you.
Consultants get ChatGPT - 4
https://www.theneurondaily.com/p/consultants-get-chatgpt4
Here's a thought experiment: ChatGPT turns everyone into a good writer—good writers remain good writers, and bad writers can use ChatGPT to become good writers.Essentially, AI sets a baseline for IQ, bringing those at the bottom closer to the top performers. Three examples across three industries:
Last April, out of 5, 000 customer support agents, junior workers were 35 % faster than usual, but top performers saw little improvement.In September, the bottom half of BCG consultants saw a 43 % boost in work quality using ChatGPT - 4, compared to just 17 % for the top performers.In November, another study discovered that "the lowest-skilled [law school] participants saw the largest improvements" in performance.
Let us know what you think.
Warmly, Member Services Team