Chatbots have stopped being a “nice to have” and have become a “must have” for ecommerce sites. They aren’t, however, a panacea. While the goal is to have the vast majority of questions handled by the bots, there’s always a need to escalate calls. While not a purely artificial intelligence (AI) issue, the need to integrate AI better into the support flow is a key to improving business performance.
Chatbots are critical today on ecommerce platforms. They are also moving past the basics of having an icon in the lower-right corner of a window that visitors can click on if they wish, and then have a simple chat. Companies are doing two things to extend the power of the bots. First, the bots are gaining access to more information about the customer’s visits to the site, present and past, and that helps with recommendation responses. Second, they are becoming more active, gaining the ability to pop-up at key moments in a customer’s journey.
That second part is being enabled, as should be no surprise, through access to more data. Vendors are providing more information about their products, so that chatbots can address more pre-sales questions. In addition, data about the customer’s journey through the site can be analyzed to understand interest in different product categories. Combine that with studies about idle time and other typical visitor behaviors, and systems can find appropriate times to ask questions of the prospect.
“Using a context based approach,” said Shauli Mizrahi, CTO, Co-Founder, Rep AI, “chatbot systems can identify, with good confidence, when a prospect is considering leaving a site. The appropriate question asked of that person can help retain interest and increase sales.” For instance, if the prospect is on the home page, a general question about why she is there, followed by a pointer to the right products, can increase interest and customer retention. Idle time on a product page can be addressed with questions about product interest to see if there’s feature or comparison information that can be provided.
Mr. Mizrahi says that the industry average is a 3.3% conversation rate for visitors using chatbots. The goal is to reach 30% of visitors in order to improve interest and the attractiveness of ecommerce sites.
Pre-sale is only the start of the customer chain, and the area most amenable to bots. While some of the questions asked at that point can’t be answered by an automated system, the main use of escalation is in customer service. When a person comes to a site with a problem, that person is already upset, and quick service matters. Right now, there is often a long time lag to get to a life person, and that can increase frustration and lead to lost future sales.
Companies such as Rep AI are working to address that risk by providing KPIs to service departments to understand wait times and realize when they need added support. With pre-sales, shortening that time can help prevent some prospects from going to competitive sites and buying the products, while service response times being shortened can help retain customers.
That isn’t the limit to the systems. Training loops for AI systems have to get better. One way that is happening is by capturing human responses to the customers. By more rapidly analyzing those answers, the training loop can be shortened. That means human responses can quickly be used to enhance chatbot responses so that more issues can be handles automatically.
What Is Still Missing
Help center use of AI is evolving. I like what’s happening, but there is one critical area that needs to be addressed. In long ago times, the mid-1980s, there was a graduate seminar in which a credit card company gave a presentation. This company didn’t have a set limit and wanted to be able to give more accurate analysis to their help center people to quickly approve or reject outlier purchase attempts. The dashboard was impressive for the time, with clear analysis provided by the expert system as to purchase history and other customer information.
The help center systems I’ve recently seen have more than the chat stream for the customer service representatives to review. There is site pathing information and a bit more. However, there is work that is needed on the user interfaces in order to pull out information faster. In addition, there’s a place for AI to do analysis based on that visitor and similar types of past visitors in order to provide more context for the representative to use in the way of solving the problem and providing product recommendations.
It is good to see what is currently happening to improve the integration of chatbots into the workflow; but, as recent times have shown, ecommerce has taken a leap in increased importance. More people are not only using it but realizing all the options from which to choose. Even with the few large companies dominating much of the market, there is still competition. The smart companies will be looking at how to improve customer service by better integrating artificial intelligence into the service flow, both pre- and post-sales – in more ways than only chatbots.
Source: forbes.com
