We are starting down our journey of creating a chatbot. There are many different architectures. Are you using ServiceNow, Microsoft Azure A.I. APIs, https://digitalhumans.com (Uneeq), Cleeva, etc.? If you have a chatbot, what benefit have you seen? I have heard some call centers have seen a 30% reduction in tickets. The idea is to index content, support text, talk (nice to have), an avatar (nice to have), enable content curation (maybe upvoting?), enable transition via API to Genesys, and trigger digital workers / RPA via a SN API call to open a request / ticket.
Sort by:
It definitely reduces 30 - 50% tickets. In addition, it’s scalable, 24*7 support, tells more about the customer.
Microsoft Azure AI Services provides a range of the top AI models available out of the box to target as well as the flexibility and speed required to handle a large amount of Chatbot based queries. It also provides us with a platform that allows us to easily “spin up” additional AI model deployments to be able to test different model’s responses. The Azure AI service provides API endpoint with an array of security features including API key based access.
With regards to storing our contextual data we have tested and used both Azure Blobstorage and file storage via containers. This area is still being further accessed to see which will be most cost effective as well as efficient for managing large amounts of context documents and data.
The Azure search services has been used for indexing the stored contextual documents and data. This allows for a range of search types including semantic, keyword as well as hybrid searches. Microsoft Azure AI services provides Azure OpenAI studio or Azure AI studio (beta) to assist in putting together the above in a coordinated and coherent manner. Both provide a “Playground” area to test including testing Chatbot responses and an out of the box method of deploying a python based web app to Azure app service for hosting an interface to the Chatbot.
In a separate case a custom API has been developed with the purpose of being able to query the Azure AI service and Search index API endpoints. This allows this custom API to act as an Azure API gateway that all Chatbot calls pass through. This is still a test case but the thinking behind this to be able to record queries and responses as well as usage that passes to the Azure AI service thereby allowing analytics to be applied assisting in revealing what kinds of questions are being asked so as to introduce constant improvements to the system. The API is easily updated to support multiple use cases and point to multiple Azure AI service model deployments each with their own contextual data. A test .NET web app was also created to interface with this API allowing for testing scenarios with little to no effort on the Azure side of the system
Seriously consider a chat bot that comes with a system already in use such as ServiceNow chat bot for what ServiceNow offers or SAP's chatbot for SAP's functions. Creating one from scratch or creating one that is covers many systems or areas is going to be much more difficult.