Delivering clarity through conversational AI.

Our Approach

We provided specialized AI chatbot development services, designing and launching a branded chatbot tailored specifically to Goodwill’s needs. This chatbot serves as a centralized source of truth, powered by a conversational interface leveraging advanced large language models. To keep information fresh and accurate, we built a custom scraper that pulls content weekly from 154+ Goodwill websites, feeding directly into the chatbot.

The bot opens by asking users for their zip code, identifying the nearest store within a 50-mile radius. From there, users could ask questions like “Do you pick up furniture?” or “How do I apply for a job?” and receive real-time, location-specific answers. The chatbot was placed as a button-triggered pop-up on Goodwill.org for the duration of the pilot.

Try the chatbot out for yourself

Lessons Learned

While visibility was intentionally limited to control testing, early results showed the bot could reduce customer service load, meet key user needs, and deliver consistently accurate information. Feedback and sentiment analysis showed strong user interest and satisfaction.

Location matters—literally.
Users want hyper-local information, and zip-code targeting worked well to surface the nearest relevant Goodwill location. But with gaps in coverage (e.g., zip codes near Denver returning no results), we saw how important it is to expand the underlying database for full coverage.

Top placement drives engagement.
When the chatbot was visible above the homepage fold, engagement significantly increased. Future placements in top navigation or as a persistent icon would improve discovery and usage.

Donations dominate.
One in three queries were donation-related, showing a clear need to streamline how this information is surfaced online.

Monthly insights are a goldmine.
Categorizing prompts gave Goodwill meaningful insights into what users actually care about—and created a roadmap for updating web content and improving branch-level service.

Impact Highlights

Over the course of the 60-day pilot, the chatbot didn’t just function—it delivered. Users engaged at a steady pace, with over 2,200 prompts submitted across a wide range of topics. Donation-related inquiries made up the lion’s share at 34%, revealing just how important it is for users to understand what items are accepted, where to drop them off, and whether pickup services are available.

Job-related questions followed closely at 20%, showing the public’s ongoing interest in Goodwill’s workforce development programs. Surprisingly, just 8% of inquiries focused on store hours—suggesting that donation and employment support are truly where users need clarity most.

Behind the scenes, the chatbot quietly kept everything fresh. Our weekly scraping tool ensured the latest hours, policies, and service details were always available—no more bouncing between outdated pages or calling the store. And because the bot responded based on the user’s zip code, every answer was hyper-local and relevant.

This automation wasn’t just a win for users—it was a relief for staff. By offloading the most repetitive and time-consuming questions, the chatbot helped Goodwill teams focus on the work that matters most. The bot provided instant, 24/7 access to reliable information, without relying on customer service agents.

But perhaps most valuable was the insight the chatbot uncovered. We categorized every prompt submitted and shared monthly summaries with local branches—offering a clear window into what people are asking, what’s working, and where content or operations could be improved. These insights helped inform everything from donation FAQs to job application instructions.

And this is just the beginning. The chatbot was built with scale in mind, and plans are already in place to expand beyond the initial 154 sites. By increasing its visibility in site navigation and continuing to share insights across the organization, Goodwill is well-positioned to meet users where they are—and answer the questions they’re already asking.