Insights from Using Chatbots in NGOs: 28,063 Unique Engagements in 4 Months

About four months ago, I designed #chatbots powered by OpenAI API with advanced prompt engineering to assist the NGO community in writing/editing proposals and reports (primarily for #BHA and #ECHO) as well as support managers with aspects of project cycle management and diversity & inclusion in HR and management practices (link to the announcement of these chatbots is here).

While I haven't collected data on specific user inputs, the platforms (LinkedIn and the chatbot hosting website) have collected general information, such as the number of users who engaged with the chatbots, as well as their organization, sector, and role. Using this data, I conducted the analysis presented below by integrating the Chatbots Users Overview with the LinkedIn Engagement Matrix. Please note that the methodology of this analysis, which is not in-depth, is outlined by the end of this overview.

My objective of this overview is to highlight the chatbots growing use within our sector and to emphasize the need for establishing ethical guidelines and data protection measures in response to their evolution

Engagement data shows that since July 10, 2023, a total of 28,063 users have interacted with the #AI chatbots I have deployed, noting that this figure does not include interactions with other chatbots like ChatGPT or Claude AI.

Breaking down the engagement by roles, the Project/Grants Managers led the way with 589 engagements, followed by Program Managers at 477 engagements. This indicates a keen interest from those in charge of managing and implementing projects and grants. Additionally, Founders, including those from local NGOs, logged 309 engagements, while Co-Founders including those from local NGOs recorded 140 engagements. Directors, which include Country Directors, Executives, or CEOs, also showed notable interaction with 140 engagements. It's important to note that the engagement from Founders, Co-Founders, and Directors may include both direct use of the chatbots and their sharing with others.

Furthermore, the 'Others' category, which includes all other roles, accounts for a substantial 26,408 engagements. This suggests that the chatbots are widely appealing and used across various roles, such as managers from support or program support functions, volunteers, assistants, officers, coordinators, specialists, advisors, and more.

When looking at the engagement by sectors, the non-profit organizations were the most engaged, accounting for around 8,391 unique engagements, about 29.9% of the interactions. Following closely was the international affairs sector, at 9.4%, which likely includes donor agencies, UN Agencies, and government aid agencies. Other sectors with notable engagement included civic and social organizations at 4.0%, government administration at 3.8%, and business consulting and services at 3.0%. The 'Others' category, which represents a diverse array of users not specified in the top categories, made up a substantial 49.9% of the engagement. This indicates a broad interest and application of chatbots across various fields.

Regarding organizational engagement, the Danish Refugee Council / Dansk Flygtningehj?lp (#DRC) was at the forefront with 1.5% of the total interactions, equivalent to approximately 421 users. The International Rescue Committee (#IRC) and the International Committee of the Red Cross (#ICRC) were close contenders, each securing 1.3% of the total interactions, which is about 365 users for each organization. UNHCR, the UN Refugee Agency and 联合国儿童基金会 followed with 1.2% and 1.1% respectively. However, the 'Others' category accounted for a significant 93.6% of the total interactions. This category includes a variety of organizations, such as the Norwegian Refugee Council , International Medical Corps , Project HOPE , CARE , as well as other international and local NGOs. I couldnt analyze the exact percentages for #NRC, #IMC, #Project_Hope or #CARE engagement, but LinkedIn highlighted that individuals working in these NGOs were among the first to open, share, and likely use these chatbots.

This data suggests that the "Others" also presents a vast array of organizations, which may range from smaller NGOs and grassroots entities to other international organizations or voluntary groups exploring these AI chatbots.

It is important to note that the engagement scores by organization, sector, and role don't necessarily indicate that these organizations or individuals are the most frequent users of the chatbots; rather, it may reflect that the chatbots announcement and access to it might have reached them more extensively than others.

It is my assumption that the engagement scores have been influenced by my direct and indirect contacts, as the majority of my network consists of individuals and organizations within these roles and organizations.

While conducting this analysis, it's important to clarify that no data was collected regarding the specific inputs users entered into the chatbots or the exact purposes for which they were used. However, to provide you with a general overview that highlights key aspects of how users engaged with these chatbots, I utilized another AI tool to generate a brief summary and a 'word cloud.' This overview and word cloud represent the keywords from user feedback, offering insights into the chatbots' most valuable features and use cases.

The summary shows that the top usage of chatbots centers around editing proposals and reports. This includes tasks such as enhancing narrative quality and managing the character count in proposal or report narrative paragraphs. Additionally, chatbots are frequently used in preparing for key organizational meetings, like kickoff meetings for grants and Go-No-Go discussions. They play a role in drafting essential elements for proposal planning and design processes. Another use is in composing advocacy letters and fundraising approaches, as well as inquiring about BHA and ECHO rules and regulations, thereby aiding in day-to-day compliance. Users also highlight the utility of chatbots as a brainstorming tool in project design, helping in shaping indicators, proposing various methods for activity evaluation, and assisting with budget narratives. Furthermore, some people mentioned using them as a diversity lens in reviewing job descriptions, job announcements, or TORs, and suggesting team-building activities.

This analysis, while not in-depth and having its limitations, serves as an interesting starting point for conversations about leveraging this technology and its ethical considerations.

28,063 unique individuals using, testing, or sharing these tools within four months is a significant and important statistic that deserves our attention and reflection.

I recommend starting a conversation within your team or organization about exploring AI tools while considering data protection, ethical use, and balancing potential benefits and risks to ensure responsible implementation.

Ali Al Mokdad

Analyzed and Co-Written by ChatGPT and Microsoft Copilot.

Here is an outline of the steps taken in data analysis, rather than an exhaustive description of the entire process but if you're interested, feel free to send me a message, and I'll be happy to share my approach.

Please note that the data analysis does not have in-depth or advanced analytical methods; it is mainly leveraging/prompting different AI tools to analyze data that is already there on the platforms:

1. Review the data from the engagement matrix in Creator Mode on LinkedIn, which includes the automatic overview provided by LinkedIn.
2. Review the Chatbots' hosting platform data, focusing on users, members, and followers.
3. Use #Microsoft #Copilot to compare and link results from 1 and 2.
4. Utilize Zapier GPT plugin to scan and review notes, LinkedIn messages, and comments and highlight the relevant ones/user feedback. Use 谷歌 Lab's 'Help Me Write' function to scan and collect emails with questions or feedback.
5. Utilize OpenAI - ChatGPT code interpretation functionality to analyze all the data collected from steps 1 to 4 and structure the analysis into different categories.