MiaRec Midyear Update with CX Today

CX Today’s Charlie Mitchell sits down with MiaRec CEO, Gennadiy Bezko, as he proudly showcases our latest innovations: AI Prompt Designer, LLM-based sentiment analysis, and robust multi-language support. These groundbreaking advancements are just the beginning of MiaRec's commitment to revolutionizing contact center AI.

Join us as we dive into what’s resonating with customers, unveil our ambitious roadmap, and explore the cutting-edge trends shaping the future of customer experience. Powered by state-of-the-art Generative AI technology, MiaRec empowers contact centers to transcend traditional boundaries, effortlessly uncovering insights and driving unparalleled performance.

Watch the full 20min interview and/or read the transcript below!

Charlie: Hello and welcome to CX Today, my name is Charlie and I'm bringing you the latest in our CX update video series. I'm delighted to be joined by Gennadiy Bezko, CEO of MiaRec. Gennadiy, it's great that you are joining us today. How are you doing?

Gennadiy: Great to see you, Charlie. I’m doing very well, thank you.

Charlie: It’s great to have you on and I'm really excited to be talking about everything MiaRec, a company that we've covered a lot and is doing some really cool things in the world of Conversational AI and Generative AI. To start this conversation, it'd be great to look back over the past twelve months and hear your perspective on what kind of products and services are resonating with your customers.

Gennadiy: That’s a good question. It has been a very, very productive year for us. In the last twelve months, we did a lot of AI-based product releases that our customers are excited about. Just to name a couple, we created an automated call summary using AI and upgraded our sentiment scoring to use a large language model (LLM).

AI Prompt Designer

We have another exciting new feature called AI Prompt Designer. When you use AI, you need to create a good prompt. If you use a LLM or Generative AI, the better the prompt you create, the better the results you get from AI. It takes some practice to figure out the best prompt for you. We found a way to provide our customers with a tool that helps them experiment easily and intuitively. People see a lot of benefits in how they can use AI right now and how they can understand these things.

AI-Powered Auto QA

Auto QA is the most successful feature in our product because it saves a lot of time for contact center supervisors. What is Auto QA? Auto QA is automatic quality assurance agent evaluation. We evaluate or score every call. Traditionally, you would score 3-5% of calls ad you don't know what the other 97% of calls are. There may be problems or low performing agents. So, with Auto QA you can score every single call. You can use AI to transcribe every call, then tell it to look at the transcription and evaluate every call using your score criteria. Examples:

• Did the agent introduce themselves properly? 
• Did the agent verify customer information like phone number or account number? 
• Did the agent resolve the issue? 
• Was the customer happy at the very end of the call?

All of this can be automated using Auto QA, and that’s a huge success. 

Customers see a lot of value in this, and we have a lot of good feedback on that product. Everybody talks about Generative AI, LLM, ChatGPT, that's very powerful and very useful and it opens a lot of opportunities for different things. Auto QA existed before all of that. Traditionally, Auto QA was done using keyword search. You just look at certain keywords or key phrases, and if you find those key phrases, like "thank you for calling," there is a greeting. "What is your name?" They asked for a name. That was possible before, but LLM gets us to the next level. With LLM, we can look at the whole transcript, the whole conversation, and understand the meaning of this conversation. Instead of looking at the specific keywords that you need to pre-configure ahead of time, you can explain to AI and that's it. You don't need to tell it how they need to greet, what words they need to speak to represent the greeting part, etc, you just explain in plain English. 

Now, when you configure the scorecard, you don't have to be an engineer. That is another huge benefit.  What we see with our customers after the initial onboarding process when we help them to configure the initial scorecard, is that they get the idea how it works very easily, and they customize the scorecard on their own. That’s the beauty of LLM. You speak in plain English to a large language model, and it understands you. You don't need to talk in engineering terms on what words they need to use or what expressions they need to put in that query. That’s a huge benefit and huge advantage of LLM in Auto QA.

Charlie: I think there's lots of really great points there to reflect back on. The first part is really interesting in terms of how you can change the prompts within a solution like call summarization. We hear a lot about optimizing the large language model behind such use cases, but the prompts make just as big, if not more, of a difference to the actual output that you have at the end. I think the fact that you're allowing contact centers to optimize that for their specific operations is a big step forward for the summarization use case.

Also, in terms of Auto QA, one thing that really stands out to me is if you use the MiaRec solution, you can also pair it with topic analysis so you can see an agent's performance across various intents and you can use that information to personalize coaching and up-level the whole agent performance. I'm sure we could dive much deeper on all of that. 

Looking back over the past twelve months, it's interesting to look forward into the next twelve months too. I don't know if there's anything that you can share as to your roadmap for the year ahead and the features and benefits that you'll bring to your customers.

MiaRec Roadmap

Gennadiy: Regarding the roadmap, we have a lot of exciting ideas about what to do with AI. The whole industry right now is exploring all opportunities on how they can apply AI technology to business use cases. Technology is evolving so fast. What was not possible just two years before is possible right now, and what is not possible right now probably will be within two years. It's a really exciting time.

We have been providing AI-based analytics using different technologies for many years. Will LLM find new opportunities and new ways to apply AI. For a sentiment score, we had a few generations of sentiment scores:

1. First one was keyword based. We create a list of good words and a list of bad words and calculate how many good or bad words were in a conversation and you guess what would be the total score for the conversation.
2. Second generation was a very useful simple language model. This was used to create a score for short phrases or a sentence, one sentence at a time. They say, “thank you for calling,” that’s good. Or they say, “I have a problem,” then this is bad. There are hundreds of different sentences, and you score every sentence to get an average score for calls.
3. We just recently upgraded to LLM, and now our sentiment score looks at the whole conversation as one piece. Instead of paying attention to individual negative phrases, it actually looks at the result of the call. A customer may be unhappy at the very beginning, and they’ll be complaining about some problems for the whole conversation until the very last minute when the problem is resolved. When you resolve the problem at the end, you don't want to flag this call as negative. You don't want to create this noise for your users or alarm people for no reason. With LLM, you can reduce this noise and make a more accurate sentiment score.

So now about the roadmap. We did this before, in the roadmap we applied the same approach, migrate to LLM in our other products. We have topic analysis where we assign different topics for every call such as, order cancellation or asking for a refund. Our product supports assigning different topics that have been pre-configured to every call. You can create different kinds of reports on how many calls you have in different docs. Historically, we were using keyword search.  We were looking for some specific keywords or phrases to assign a topic. It works very well, but it doesn't work for very complex use cases. With LLM, again, we can look at the whole conversation and we don't focus on certain keywords, we focus on the outcome. For example, if there was an order cancellation and it actually was cancelled, this is called cancellation. If it's not cancelled because the agent was able to save the customer, then we assign a different topic for that call. That's what we are working on right now and will be releasing soon.

Another thing that is very exciting that we are working on is omnichannel support. Historically, MiaRec was focusing on telephone conversations. We were analyzing telephone conversations for sentiment summary, Auto QA, all of these exciting features. It’s interesting that to do all of these AI analytics, we actually transcribe every voice conversation into text. Then we work with text. We don't work with voice anymore. We always work with text as a source of analytics. Now, we basically transform it into text-based analytics, and we can easily add chat analytics, email analytics, support and tickets analytics. This is something that we are working on right now, omnichannel support, because many customers have multiple ways to communicate with them, not only voice. Many companies open chat to their customers, they handle support tickets, email and so on. We are excited to release omnichannel support in future.

Charlie: Lots of really great stuff in there. Going back to the start of your answer, I think the sentiment analysis point was quite key because we're seeing more and more contact centers take sentiment and share that insight with other parts of the business. So marketing is using the sentiment scores from the contact center to inform their proactive outreach, and customer success and customer retention using those scores to try and save customers. As you mentioned, they only summarize the whole call, not the important part.

I think that such evolutions are very big, especially for a lot of contact centers that leverage MiaRec. I know they might have a work system like Five9 integrate with MiaRec and then pass that information to the Salesforce CRM. I think it's a benefit that not only improves agent performance and training, but it also has big permutations for the whole of the CX operation. That really stood out for me, as well as the LLM-based topics and more. It’ll be interesting to hear your thoughts also on one tech trend for the next twelve months that's really going to impact your customers. Could share with us what that will be?

Rising Tech Trends

Gennadiy: In my opinion, one big group of trends in the contact center industry is using AI assist. AI assist is one of the exciting products that you can implement right now, and it works. With AI assist, you can empower your agents and make them more knowledgeable because AI will use the knowledge base and give them some hints on how to handle call. They can be more productive during the call. AI assist I think this is a huge trend. It's low hanging fruit and it's very useful for companies.

More future trend, in my opinion, is AI bots. Right now we have AI, and we use AI successfully to measure human performance, like with Auto QA. When we use the AI system to listen to every conversation, agent to customer, human to human, we listen to every conversation and score every conversation. We use AI to measure if this conversation was good or bad. Now, what if we go to the next level and instead of using AI to measure humans, what if we use AI to replace humans? If we use AI to actually communicate to contact center. We have AI bots, and this is a big trend. I think this is the future.

At this time technology is not quite mature enough, so we can only handle simple use cases. But even for simple use cases and simple call scenarios, we can have a lot of benefits. Customers can have a lot of benefits because they can reduce their workforce. They can release the workforce from simple tasks and use them for more important tasks. Simple tasks will be handled by AI bots, and more important critical tasks will be handled by humans. This is the future, in my opinion, using AI to replace humans in contact center. We are not there yet but look at OpenAI. They recently released a new model, ChatGPT for omnichannel. How impressive what that model does. It can talk to you, it can speak to you, it can answer you, it can understand you, and it can search information for you, and you communicate just in voice. It's really exciting, and this is the future. I think what is not possible right now, what still doesn't work quite well, will be possible just in a couple of years.

Charlie: It was a really exciting announcement and I think it does show the potential of what AI can do in the contact center and how important the orchestration piece is for contact centers. If we go back to start in today's conversation, a lot of how contact centers can optimize those capabilities and through the playground, I think MiaRec is demonstrating how it can do so.I'm sure we can talk a lot more about a lot of this, how AI might replace certain human roles within the contact center and more. but I think that's a great place to end today's chat. It’s been a great conversation. Thank you very much for joining me today, Gennadiy.

Gennadiy: Thank you, Charlie.