EXCLUSIVE Q&A: GoodwillFinds develops generative AI chatbot
GoodwillFinds, a growing consortium of 18 individual non-profit Goodwill organizations, is streamlining online customer service with artificial intelligence.
Chain Store Age recently spoke with Matt Kaness, CEO of GoodwillFinds, about how the initiative is using a generative AI model to automate key aspects of its member Goodwill organizations’ digital customer experience.
[READ MORE: Exclusive: Goodwill e-commerce initiative builds online marketplace]
Goodwill operates on a federated model, meaning it's an affiliation of 154 individual, independent nonprofits at the state and local levels.
Why did GoodwillFinds decide to launch an AI chatbot?
As a non-profit, it is incumbent upon GoodwillFinds' leadership to create exceptional customer experiences at the lowest cost possible, ensuring that a maximum amount of the proceeds from the sale of goods on the site go to fulfill the mission of the Goodwills that sell with us.
We also recognized that many of the inquiries customers had could be resolved by exposing key pieces of information from multiple internal systems in a secure, dialogue-driven manner. This frees up our internal customer experience team to handle the most complex and value-added inquiries.
How did you develop and deploy Gem?
Gem is proprietary to GoodwillFinds. Our CTO and internal team developed it using OpenAI’s capabilities via API integrations. It was a team effort with the GoodwillFinds customer experience staff helping to train the model over several weeks of development.
[Editor’s Note: OpenAI is the developer of ChatGPT, a generative AI model which interacts with users in a conversational style that mimics human interaction and uses machine learning to continually refine and improve its responses.]
Once we reached an acceptable level of consumer interaction in testing, we released the model on the site and continue to curate the data and level of empathy the model uses as we monitor its interactions with our customers.
How does Gem work?
Gem’s primary function at this time is to allow customers a quick, easy-to-use portal for self-service related questions they may have about orders they placed, items listed for sale on the website, shipping and return policies, and operational processes.
We briefed OpenAI’s large language model with key references like our FAQs, as well as real-time updates of transactional history and product listings. When a customer navigates to the "contact us" page on the website, they are prompted to engage with Gem in a conversational way to have their questions or issues resolved.
When Gem needs to gain access to private data to fully service a customer inquiry, it will require the customer to login so we can ensure privacy is maintained. If Gem is unable to directly handle the customer’s questions to their satisfaction, the bot opens a case with our customer experience team and an associate takes over the management of the interaction.
What have the results been so far?
Since launch, Gem has successfully handled 68% of all customer inquiries, with the remaining 32% being escalated to our customer experience team for direct engagement with a live customer care associate.
Are there any future plans for Gem you can discuss?
We want to continue to enhance the customer service capabilities of Gem so that more complex cases like items being damaged in shipment are handled via automation.
More broadly, we expect to develop capabilities for Gem to engage with customers during their website browsing experience, assisting with product recommendations and curation of items in the vast catalog on our site at any given time to items most interesting to the individual browsing the site.
