Nuggets.ai

Inspiration

Our idea was initially inspired by the personal experience of some of our team members who had been in roles which required them to provide customer service and understood and empathised with the pain points of customer service crews. Later, upon considering TreeHacks problem statement #13, we wanted to challenge ourselves to leverage artificial intelligence to aid customer service crews and improve overall user experience.

Today, customer service crews are swamped with requests that could be solved with a small nudge. The overwhelmed state of customer service has caused businesses to lose more than $60B per annum just in the US alone. Yet, most issues that customer service crews solve do not actually require them. In fact, 3 in 4 customers want to resolve product and service issues on their own instead of contacting service crews, but only 9% of them are able to do so completely via self service. From FAQ-pages to customer service chatbots, companies create more channels to reduce the burden on customer service crews. But in reality this creates more challenge for the customer to arrive at a self resolution. Providing customers with instantaneous nudges at the point of a digital roadblock would greatly reduce drop-off and alleviate customer frustrations, elevating the entire product experience for businesses.

What it does

Nuggets.ai predicts the root cause of customers’ digital roadblocks on SaaS platforms based on their navigation behavior, and provides instantaneous prompts to unblock them. A secondary feature of Nuggets.ai is aggregating data to provide teams with crucial analytics on common error touchpoints to supercharge the product experience

How we built it

We utilised react for the frontend of our product which is hosted on Vercel platform. The backend is coded on express which runs a python script to call the OpenAI API to generate word embeddings. We also used the AWS RDS Postgres database which stores metadata extracted from parsing of the HTML pages.

Challenges we ran into

One challenge we faced was the lack of data to train the AI model. We had to scale back on the features which we wanted to include inside our MVP as a result. Our tech stack was also technically complex and involved integrating multiple components such as the frontend, backend, database and AI model with deployment on the cloud.

Another crucial challenge was having to learn largely on the spot. Since 3 of us in the team have never been for a hackathon and 2 of us did not have prior coding experience, we had to think of ways to collaborate such that we could still leverage on all the skills of the team members. We spent long periods of time teaching and learning Python such that we could meaningfully contribute to the building of Nuggets.ai.

Accomplishments that we're proud of

We are proud of the vision behind Nuggets.ai -- in particular, it solves a niche pain point that may not be apparent to people who have not been in customer service. Nuggets.ai is the type of product that consumers and customer service crews do not know that they need yet but will be addicted to the way it revolutionises the way their products are built. Nuggets.ai also has a strong product story which is one that several consumers will probably relate to.

Furthermore, Nuggets.ai leverages on Open AI's API to generate instantaneous prompts to guide the customer through digital roadblocks. We achieve this through the semantic understanding of the user's intent, and matching it with the most appropriate corrective action. In creating Nuggets.ai, the most complex and what we are most proud of is the integration of the whole system (including the front end, back end, database and python script) since python was used to perform the OpenAI API but it was difficult to link. Another aspect was running the Python script with the Javascript code.

What we learned

Most of our team was unfamiliar with the tech stack, and it was the first time we did a hackathon. We learnt about the importance of writing good code with testing, which enabled us to debug faster, as well as deployment of our infrastructure on the cloud. Given the time constraints, all of us were exposed to languages and frameworks that we were unfamiliar with at some stage, and coped admirably with that uncertainty. Additionally, 2 of our team members were not technically inclined and had no prior coding experience, and so it was a steep learning curve for them.

The most tangible learning which we took away was the importance of collaboration and good communication across the hackathon duration. It was difficult to build an idea from scratch in just 36 hours, and our determination to complete our MVP and consistent checkpointing allowed us to achieve our objectives.

Furthermore, as we large potential in Nuggets.ai, this is definitely just the start of the journey for us. While the end product for TreeHacks need not be perfect, we realised the importance of taking a step back to consider the bigger picture, taking time to consider our business model, competition and holistic business viability.

What's next for Nuggets.ai

We believe in Nuggets.ai as a product and the value, both in the economic sense and social sense, that it can bring to the world. Hence, we are committed to continuing to pursue Nuggets.ai as a business, the first step being validation of the venture. We hope that in the long term, Nuggets.ai not only helps to lubricate user experience but achieves a greater vision of unlocking the hidden potential of the internet to serve the purposes they were first built for and reach people they were intended for. While we do anticipate to face roadblocks ourselves, such as in building a regulation compliant plug-and-play solution that is compatible with all websites, we believe that these very challenges is what make Nuggets.ai worth building.