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.

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.

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.

I have been using Kore.ai for around one year.

Currently, I do not think Kore.ai is stable because previously it was.

Kore.ai has one of the best scalabilities in terms of handling massive growth in both user traffic and conversational complexity.

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.

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.

My main use cases for Kore.ai include a diversity spanning from banking and the BFSI sector, to the travel market including aviation, and applications on the automobile side. Kore.ai has created sub-products for different industry verticals, which provides good use cases in terms of banking.

A specific example of a use case in banking is where a client needs to perform real-time transactions from one account to another. I can call using Kore.ai, and as a consumer, I can transact an amount of dollars from one account to send to any beneficiary that is already added into my account. On the aviation side, we have done use cases with Riyadh Air, which is a new airline in the Middle East focused entirely on guest experience. Customers can call Riyadh Air help assistance to book a ticket, schedule a trip, or select seats at certain airports.

I want to add the use of AI technology and the ASR and TTS services that we use as part of my main use cases. The performance of the bot becomes more dependent on what kind of external services or external LLM sources are being used. We are currently using Microsoft ASR and TTS services in most of the bots that we have deployed with Kore.ai, and Kore.ai has their inherent native Microsoft speech services enabled as well. Therefore, Kore.ai is more efficient when it comes to Microsoft ASR and TTS speech services. They have their own LLM, but based on our experience, we have used Cloud Anthropic most often and have also used OpenAI, which works very well with Kore.ai.

My favorite feature that Kore.ai offers is their Agent Desktop. If you are integrating Kore.ai with a contact center solution as an integrated solution, Kore.ai also provides a standalone solution. You can perform both types of deployment, and their Agent Desktop, Agent Assist, and Agent Co-pilot features are very exciting in terms of how they can pull in knowledge. They have their own knowledge libraries that can facilitate your agents when calls are routed from an AI agent to a human agent.

The Agent Desktop and Agent Co-pilot become especially useful for my team when you have a large volume of knowledge to traverse through as a human agent. With Agent Co-pilot and Agent Assist inside the platform, the information from the knowledge base becomes easy for you to access. Based on customer intents during calls or chats, Kore.ai Agent Assist can detect the intents quite efficiently and bring out the best knowledge articles from the knowledge libraries to present to you as an agent. The system does most of the work that an agent has to do in finding knowledge and searching for it in real-time. We have improved the average handle time significantly with the use of Agent Co-pilot.

Another exciting feature is the industry vertical-based bots that have already been tried and tested by Kore.ai. I don't believe any other vendor offers this with specific bots for the healthcare industry, aviation, BFSI, automobile, and insurance. They have predefined use cases already plugged in, so you don't have to start from scratch. Predefined templates inside the libraries can be reused and built upon for your bots.

Kore.ai has positively impacted our organization by helping us roll out the platform in one of our Middle Eastern markets first, where Arabic language was a challenge. We addressed those challenges through our own local native Arabic speaking personnel and then moved to the European market, where there is significant language diversity. The more exposure Kore.ai received with us, the same kind of efficiency we achieved when switching from one language to another. We have built a team of 30 plus agents who are conversation designers, AI engineers, AI implementation engineers, and Kore.ai experts on the platform. Organizationally, we have progressed considerably with Kore.ai.

The positive impacts we have seen include expected reductions in average handle time, which is typically around 40 to 50 percent for any BFSI industry use case. In automobile and aviation, the AHT reduction comes at a cost because the calls are longer. We track parameters such as AHT, customer experience, and CSAT. For example, how the bot engages with the customer, carefully takes the intents from the client, and then responds back to them reflects these metrics. We see KPIs related to average handle time and agent reduction playing a significant role as we are the biggest BPO provider.

There are some technological gaps with Kore.ai when it comes to language detection because this problem is common among all conversational AI vendors. They are using external sources for automatic speech recognition and generating text-to-speech services. The speech recognition mechanism remains primary for these vendors, including Kore.ai. We have also observed some limitations in scalability, particularly on Azure, where we have had to scale it on different cloud platforms around the globe. Implementing Kore.ai on Azure microservices might be a challenge compared to what we have seen in AWS, where cloud services are easier to maintain.

From the perspective of post-implementation support, Kore.ai can improve significantly because I have seen it lagging in their industry vertical. Other vendors are quite effective at providing post-sales support, and that is an area where Kore.ai can gain market traction through improvements.

I have been using Kore.ai for more than five years.

Kore.ai is definitely stable. The on-premises version is the most stable, followed by the hybrid cloud model, and the public cloud setup is stable as well.

Kore.ai's scalability has limitations, particularly on Azure cloud, which is not cloud dependent. It is mostly cloud agnostic, and while it scales very well with AWS, there are certain microservices on Azure that need elasticity, indicating gaps from the cloud provider, not from Kore.ai technology itself.

Customer support is where Kore.ai has significant room for improvement. Post-implementation support is a particularly discouraging aspect for me.

Previously, we have partnered with Cognigy and have our own in-house solutions, including Unify apps. While other vendors have their positives and negatives, we prefer Kore.ai due to our strategic partnership with them, making it our go-to solution in the market. We worked with Cognigy previously, which is now acquired by Nice, and we specifically compared Kore.ai with Cognigy.

Definitely, we consider the digital transformation journeys for customers, taking into account that investment costs are typically higher in the first two years for implementing technology, identifying use cases, and mapping them. Once up and running, the benefits of AI come into play. The results we see are agent reductions of 15 to 20 percent in multiple cases, lower telephonic costs due to SIP provisioning, and improved customer experiences with voice bots, chatbots, and reduced call times.

Licensing is worked out on a case-by-case basis with their account management teams based on volumes. Their expert app services, which provide professional support during implementation, are higher in price. We have an in-house team capable of implementing Kore.ai, but post-implementation support, as I reiterated earlier, needs improvement both in terms of cost and delivery.

Teleperformance is an exceptional reseller from Kore's side, and we have a great partnership with them. We are direct vendors and resellers of Kore.ai as a direct vendor. For our public cloud deployments, we use AWS most often. We deploy Kore.ai using multiple configurations, mostly public cloud in AWS Frankfurt, Microsoft Azure in UAE, and we also have one on-premises deployment for one of the leading banks in the Middle East, so it is a blend of all. From the perspective of post-implementation support, Kore.ai can improve significantly because I have seen it lagging in their industry vertical. Other vendors are quite effective at providing post-sales support, and that is an area where Kore.ai can gain market traction through improvements. Another exciting feature is the industry vertical-based bots that have already been tried and tested by Kore.ai. I don't believe any other vendor offers this with specific bots for the healthcare industry, aviation, BFSI, automobile, and insurance. They have predefined use cases already plugged in, so you don't have to start from scratch. Predefined templates inside the libraries can be reused and built upon for your bots. I rate this review overall as a seven.