Notice AI

Our web URL crawler will ingest subfolders for your websites
After a quick training job, your chatbot is ready to go!
Easily add PDFs and text files to supplement your knowledge base
Code we produce to embed your chatbot in your website
How we help companies scale their support with AI

Direct link to Youtube Video: https://youtu.be/bHvZCKPJmIc

Inspiration

We believe that support for products and services is fundamentally broken. Neither the provider nor the consumer have a great experience at scale. We wanted to bring the magic of ChatGPT to any product and make it easy for non-technical people to train their chatbots on any knowledge, and update them and manage bots at scale.

What it does

NoticeAI is a product that helps take your company’s support & internal documentation and with <1m of setup we can deliver a Generative AI that will converse with your customers in a highly interactive and accurate way to resolve their support needs.

How we built it

Hosted in Azure, we leverage OpenAI APIs to train models on arbitrary data. We also use LangChain to improve the interactivity with the chat agent.

Challenges we ran into

When we talked to customers (we've had over a dozen customer meetings), we encountered some objections to the security of our SaaS deployment. To address this, we plan to move to an Enterprise-ready architecture where we deploy training jobs and store training data solely on the customer's VNET. We can also store the trained model in their VNET. This will offer a more secure deployment model for enterprise customers.

Accomplishments that we're proud of

The fact that non-technical people can log in and create bots that are knowledgable about any data you put into it --we succeeded in providing a no-code way to train ChatGPT-like models for anyone who wants to do it.

What's next for Notice AI

We are building out our enterprise deployment model and talking to customers. We've actually had more than a dozen customer conversations with small and large businesses and have been incorporating their feedback into our product. We are also wrapping up talks to bring in 2-3 design partners via paid POCs. We also plan to provide ticket escalation paths to allow the user to file a ticket for humans in case the LLM cannot address the current issue.

Disclosure: The core product was built prior to the hackathon. During the hackathon we improved the overall chat interactivity and fixed bugs, as well as prepared a demo chatbot with information about the hackathon. We also created the pitch deck, and a target architecture diagram for Enterprise customers during the hackathon.