Using AI Tools to Scrape, Categorize, and Communicate Cutting Edge Research on Women’s Empowerment

In the June cohort of the XD131 Applied AI course, participants were asked to choose a final project that would draw from guest experts, featured tools, and practical work sessions to apply AI in a novel way to a personal interest. This post features a brief Q&A with Sabreen Alikhan and then the final project produced for the course. 

Q&A with Sabreen Alikhan

Q: Could you share a bit about yourself and why you took a course on Applied AI?

I serve on the leadership team at Women for Women International, where I'm spearheading strategic initiatives to thoughtfully integrate AI and emerging technologies into our mission-driven work. Our work directly supports long-term transformative change for vulnerable populations in conflict-affected settings, and there is tremendous opportunity for us to leverage AI for greater impact — from improving program delivery and data-driven decision making, to enhancing organizational productivity and resource efficiency. While there is great potential for what AI can bring to our work, it also requires us to be extremely thoughtful to protect the vulnerable communities we serve. 

What drew me to the Applied AI course was the chance to engage with other organizations that have navigated this landscape. As someone who guides our organization's innovation strategy (and enthusiastically follows the social impact innovation space), I was keen to hear practical insights into change management, stakeholder engagement, and the operational realities of AI adoption in complex settings. The course's focus on applied learning and diverse organizational perspectives aligned well with my remit to advance digital transformation with both ambition and responsibility.

Q: In preparing for the final project, were there any tools, experts, or experiences that were particularly helpful?

The tools we were shown that related to knowledge management were the most insightful for me. In particular, AI solutions developed by Hilton Foundation, Urban Institute, and Mercy Corps helpfully articulated the evolution from use case concept to AI governance frameworks to working prototypes delivering value across different organizational functions. And while I'm familiar with the possibilities of generative AI for knowledge management, I found uniquely useful the under-the-hood insights from experts who got their hands dirty with the realities of change management and implementation within their respective organizations. Mercy Corps’ "minimum ethical product" chatbots trained on internal document libraries showed how to avoid building a solution in search of a problem, namely through proactive IT engagement, achievable, well-governed steps, and user-centered design. 

These practical examples of powerful generative tools reinforced for me the enormous potential to accelerate our learning cycles and make our in-house evidence exponentially more accessible across all teams.

Q: Why did you choose this final project option? What were you hoping to gain from the project?

I lead my organization’s global applied research, so I live and breathe data. I've long been fascinated by the power of data scraping to rapidly accumulate real-time information from across the web, but the coding barrier (e.g., needing Python or other programming languages) tends to make experimentation more cumbersome than I'd like. No-code scraping tools like Browse.ai finally remove that friction, making what used to be a technical specialty now accessible to anyone who wants to supercharge their research process. I wanted to test just how quickly and easily I could tackle a practical task: extracting and transforming research summaries from a single online source.

I was eager to use this as a concrete example to help my teams envision the productivity gains within our reach. Tools like this represent an opportunity to automate the kinds of routine information-gathering tasks that can consume significant staff time (not to mention the tasks we don’t undertake because they would be prohibitively time-consuming). By demonstrating how quickly I could systematize a small research extraction and processing task, I wanted to spark conversations about where else we might apply these approaches to free up our team's capacity for the more complex, strategic thinking that truly advances our mission.

The final project:

I'm constantly seeking and consuming relevant development economics research to share with my teams, so I decided to experiment with no-code web scraping using Browse.ai combined with GPT for Sheets to make the process more efficient.

I created a workflow that automatically extracts data from VoxDev.org, a development economics research portal I follow. Using Browse.ai, I built two robots: one to scrape search results for relevant topics (handling pagination across 300+ results), and another to capture the full research summaries from each article (“deep scraping”). I integrated this workflow with Google Sheets so the extracted data (i.e., article names, authors, URLs, and complete research descriptions) flows directly into a structured spreadsheet.

From there, I used GPT for Sheets to further process the data, experimenting with transformations from tweet-length summaries to classifying research by relevance to WfWI’s work. While I hit token limits before completing everything I wanted to try (noting the free versions got me very far!), the proof of concept worked well.

The real potential is in building off of the initial data pull: scheduling automatic updates, adding in other research sources, feeding the compiled research into tools like NotebookLM for deeper interrogation alongside WfWI research, and even automating quarterly research newsletters for colleagues. It's a practical way to transform sporadic web browsing into a comprehensive, systematic, highly interactive knowledge management system.

The applications extend well beyond my team's immediate needs; from streamlining partner identification and feeding real-time data into our strategy and watchlists to tracking rapidly changing sectoral news and automatically consolidating the latest technical frameworks, I look forward to where my teams can take these possibilities next.