What is our primary use case?
My main use case for Kore.ai is developing and deploying enterprise-grade AI agents and chatbots, which includes designing conversational flow, setting up intent testing, and evaluating LLMs that can integrate to automate customer interaction and streamline internal support workflows.
One specific example involved developing a demo virtual assistant designed to optimize internal support workflows and customer interaction testing. A key part of our workflow was evaluating how well different LLMs integrated with the platform and also rigorous intent testing implemented in that.
What is most valuable?
I think the best features about Kore.ai are how easy it is for a developer to use. For example, the drag-and-drop dialog builder is exceptional. Also, NLU and intent testing are also good.
When comparing it to other software, I think it is easy for a developer to build the agents, which helps significantly reduce time-to-market while keeping the architecture clean.
In terms of integration and flexibility, Kore.ai provides a significant advantage that makes it highly adaptable for complex enterprise environments. For example, API and back-end integration, authentication handling, and data mapping, etc. Also, multi-channel deployment flexibility is a feature as it is an omni-channel agent.
Kore.ai helps in operational efficiency and faster time-to-market, which is development velocity. Although it has had its cons, such as the platform being buggy and support not being that great.
What needs improvement?
One thing I have learned is the unpredictability of LLMs and how buggy Kore.ai is as a software. We had a lot of issues with Kore.ai, and now we are trying to shift to other software.
Kore.ai has many cons, especially in that the platform needs to focus on stability as it is buggy. It is hard to find mistakes, as it is difficult for debugging, and it also slows down during heavy training.
Better guides would help, as the manuals explain what buttons do, but they lack good examples. It would help to have simple step-by-step guides showing how to write custom code or fix errors. Additionally, it would be better to have good support since the platform has serious technical glitches, and the customer support takes too long to fix them. The first reply usually offers basic tips instead of quickly sending the problem to senior engineers. Furthermore, I think the platform is expensive and is geared towards massive corporations. I think it needs to have cheaper basic pricing options so smaller teams can test out quick ideas without spending too much money.
For how long have I used the solution?
I have been using Kore.ai for around one year.
What do I think about the stability of the solution?
Currently, I do not think Kore.ai is stable because previously it was.
What do I think about the scalability of the solution?
Kore.ai has one of the best scalabilities in terms of handling massive growth in both user traffic and conversational complexity.
How are customer service and support?
I think the customer support needs to improve, as it is inadequate right now. They do not resolve issues quickly and they do not forward it to senior engineers. Rather, basic support is provided.
What other advice do I have?
I think there is a reduction in fallback rates. By fine-tuning hybrid NLU, the bot's ability to correctly understand user intent has increased significantly. I think it led to an approximate thirty to forty percent reduction in unhandled fallbacks.
I think people should focus on hybrid NLUs and not just use any large LLMs for everything. You can use a standard visual builder for important transactions to keep them accurate. Save the LLM features for unexpected questions or conversational fallbacks, and also prepare for the learning curve, as the platform is easy to learn for basic setups but hard for advanced coding.
My overall review rating for Kore.ai is six out of ten.
Which deployment model are you using for this solution?
Public Cloud
If public cloud, private cloud, or hybrid cloud, which cloud provider do you use?