WaveAccess presents ValueAI, a new AI-fueled text analytics platform

Sorting incoming messages has never been as important as in the Internet era. The sooner an important user request gets into the right hands, the sooner it will be answered. And the probability is higher of a satisfied client.

For the routine request classification you always need someone to read all these messages and to decide who in the company is able to answer the request. This causes a lot of manual work, and sometimes prevents leading specialists from answering the requests timely. Solutions with AI automate this process: they ‘read’ messages, classify them, and allocate every request to the right specialist or department many times faster than human specialists are able to.

However, the development of an AI from scratch is often time consuming: it requires a year or longer and a full team of specialists. Almost 86% of Machine Learning projects fail due to different reasons. Most often the reason is data quality: entangled, unprepared, or incomplete data rarely leads to the creation of a relevant model, and this issue is often relatively hard to detect beforehand. However, here at WaveAccess we have different solutions for this data quality problem and we apply them whenever developing new AI models.

ValueAI is an automatic and easy-to-install customer request processing engine with a ready AI in it. We created ValueAI so our customers may immediately leap into utilizing a relevant model without years of costly preparations. For your business ValueAI can:

• Automatically distribute incoming user requests
• Reduce response time threefold
• Reduce the amount of manual work
• Optimize leading specialists’ workload
• Reduce human errors
• Improve the SLA compliance

As a result you get more successfully closed deals, service quality improved, and customer loyalty increased — all of these without the risk of failing the dev project.

How to get started

To get started with ValueAI, you will need a labeled dataset in the form of a .csv or .xls spreadsheet. ValueAI will use this dataset to find patterns in data and predict target values — urgency, category, chance of conversion, and so on. The larger the dataset is, the better. To protect your data, ValueAI can be installed into your closed circuit without online access.

Before the training starts, the built-in tool will help you curate your data and improve its quality — for instance, if your data has too many examples for some category of tickets, and lacks examples in another one. You’ll see which model accuracy you can reach using this data in percentage.

Feel free to start the training immediately, no coding skills or profound knowledge in Data Science is needed. After the training you’ll receive a ready AI model that can sort new data based on the training. This model can be already integrated with your corporate software like a CRM, Outlook, or with another system that collects incoming requests. You also can increase the accuracy of the model by experimenting with more data. Feel free to choose the best performing model for the integration.

As a result, you’ll be able to benefit from the ready-to-use sorter utilizing Artificial Intelligence in one month, minimizing the risks of AI development.

Potential uses and benefits of request processing automation

ValueAI is a multi-purpose AI. It targets large request flow and long response time issues. Potentially, it can even be trained to like certain posts — if your business processes have a requirement for this. The companies that process their request flows manually will benefit the most.

• Customer support
  

  Challenge. SAAS and gamedev companies deal with hundreds and thousands of tickets per day. Some of these tickets are with strict SLA, and some of them don’t matter. In this incoming flow the most important tickets may be lost for hours, which may cause reputational risks.

  One of our clients, a logistics service provider, had a whole department dedicated to reading incoming emails and creation of tickets. The emails were read one per one as they arrived, so that a premium-client's emails sometimes left unanswered while a specialist was processing less urgent messages.

  Solution. The AI based request processing module was integrated into the client's CRM. This helped grade the complexity of requests received by email based on topic and message body. Now support tickets are closed 2 times faster even if the load increases.

• Ecommerce
  

  Challenge. The more customer requests that are processed, the more deals will be successfully closed. Sometimes the success of a deal depends on the speed of the answer, but there’s no hint if a prospect is going to answer the phone or not, that’s why a manager will process the requests one by one, without ranging them. So some prospects will have to wait for an answer for too long. In some segments conversion growth leads to visible sales growth.

  Our client, a business-class ticket aggregator, faced a challenge of low conversion. When managers called back to leads, some of them never answered the phone, especially if the call back was delayed. Managers were in need of a solution that would focus their attention on the requests from hot leads and also would sort out insignificant messages.

  Solution. Conversion can be predicted based on lead metadata. Leads with highest odds to convert are allocated to leading specialists. We helped our client to gather data for analysis, and train the model. The solution predicts if a lead is ready to answer the phone. Average conversion increased by 1,5% within 6 months, resulting in 17% sales growth with 3 times faster order processing.

• Sensitive data holders — financial institutions, medical companies, etc.
  

  Challenge. Banks, financial services and healthcare companies deal with a flow of requests containing personal and other sensitive data. This data should be encrypted, and never be transferred to third parties. Manual sorting of the flow is time consuming and sometimes leads to errors. One of the companies needed to process requests to a US based telemedicine service in order to redirect them to appropriate care specialists.

  Solution. We have led the company’s CEO through a discovery session revealing how to use ValueAI in her business processes, and later we installed the product into the company’s server. The client’s team trained several AI models. After the required prediction rate was achieved, our specialist visited them on-site and helped to integrate the model with their website’s CRM. The prediction rate is up to 78%.

Your company’s AI without programming. How to start

ValueAI is a ready-to-use Artificial Intelligence, which can be used ‘as is’ for a high amount of tasks. We invite you to book a free demo with the ValueAI crew and discover infinite opportunities of the platform.

You will learn:

• How to prepare your data before learning, how to upload it, and how to use the Warnings report to improve the quality of your dataset
• How to start the training without programming skills, and how to use all the resulting Reports
• How to improve the prediction rate of your model, using different approaches, and how to choose the best models
• How to go on with your best performing model: for example, how to integrate it into your corporate software

Furthermore, we can

• Free you from excess work, and train your models based on your data, doing all the experimenting until the desired prediction rate is reached
• Customize the product so it better works for your specific industry or business goals
• Train your employees to increase the prediction rate, read analytical reports, and experiment with models

Why ValueAI

ValueAI enables WaveAccess to speed up our machine learning development projects and offer our clients reasonable prices and terms. We use the platform when we need to check a hypothesis, compare models’ prediction rates depending on different approaches, and choose the best technology to work with.

For users,

• No Data Scientists required. Reduce the project’s budget compared to the development of an ML-based project from scratch.

• Off-the-shelf AI engine. The first results with the tool can be achieved after an hour, with the customized ready-to-use model achieved after a month.

• Easy to use. Improve the results of your training as easily as you’ve trained your first model.

• Cloud or on-premise deployment. ValueAI will successfully perform its tasks while being installed on your offline server.

***

ValueAI is a user-friendly low-coding tool. Feel free to book a demo with the ValueAI crew, and see how it can facilitate your request processing.