In niche non-profit grant-making, informed decisions are essential. Relying on accurate data helps align granting strategies with the organisation mission, vision, and goals, as well as addressing community funding gaps.
We recently undertook a project to analyse a large amount of free-text survey data to guide granting direction. This case study explores how we leveraged GenAI, specifically advanced language models (LLMs) like OpenAI‘s ChatGPT and Anthropic‘s Claude, to achieve our goals.
The Challenge
The organisation faced a significant challenge: an extensive collection of free-text (and quantitative) survey data that was difficult to process. Registered charities were surveyed to understand their recent experiences with funding, service gaps and their future needs. The goals of the analysis were to extract meaningful insights, identify potential non-profit service providers aligned with the mission, and provide valuable input to the granting committee to ensure that funding aligned with the strategic goals. The key question was: how could we efficiently process this unstructured data to produce reliable, actionable insights?
The Approach
To overcome this challenge, we employed a multi-faceted approach combining GenAI-driven analysis with essential human oversight. Here’s how we tackled the project:
- Data Preparation: We simplified, combined, added some structure and reviewed the data, ensuring it was anonymised and free of Personally Identifiable Information (PII).
- Prompt Design & Output Review: The initial prompts quickly became too complex, so we divided the project into six focused tasks, many with multiple prompts. This systematic approach allowed us to address different aspects of the data and merge the results later for a consolidated view. We maintained consistency by using a task template for prompt design, detailing the role, input data, task, output format, examples and additional instructions.
- Multi-LLM Approach: By using multiple LLMs and merging their outputs, we enriched the analysis and drew on the strengths of multiple AI models. ChatGPT custom GPTs and Claude Projects also helped us organise the work effectively, making the process more manageable and providing context for the prompts.
- LLM-Driven Search: We used ChatGPT for geographically restricted searches to identify potential non-profit service providers who could address identified service gaps. This also led to a small side project to develop a small utility that utilised the Google Custom Search API, then feeding the results to the ChatGPT API for summarisation (read this #LinkedIn article for my experience in developing this: Software Development with AI Companions ).
The Outcomes
This approach saved significant time and provided valuable insights.
- Reduced effort by 50% compared to traditional consulting approaches. In one portion of the process, the Internet search and analysis, it is not clear that the work could have been completed without GenAI.
- The combination of LLMs and human expertise uncovered innovative insights that informed the grant-making strategy, generated new project ideas, and identified service providers that might have otherwise been overlooked.
- The agency, being all volunteer run, felt that adding GenAI to the process was like adding a temporary staff member to the team.
Conclusion
This case study demonstrates the power of GenAI to augment human skills and knowledge. By integrating GenAI into the data analysis process, we derived valuable insights that directly informed the granting strategy, achieving over 80% of our goal (with the remaining 20% needing a human touch to complete). The iterative nature of the process, combined with human oversight, ensured accurate and actionable results.
Take the step of integrating GenAI into your own processes and unlock the potential for even greater success in your initiatives.
Additional Reading
Canada’s Nonprofit Workforce – ‘The share of all nonprofit job postings demanding artificial intelligence skills increased by 32.7% in the second half of 2023 compared to the first 6 months.‘
#AI #GanAI #DataAnalysis #NonProfit #GrantMaking #LLM #Innovation #CaseStudy #StrategicInsights