Inside the Three Tiers of Chatbots: Reactive, Reflective and Proactive; And how to take your bot to the next level

With customer communication coming in via email, phone, video, chat and social media, businesses are being sprayed with a fire hose of conversational data and are quite literally drowning in it as a result.

It’s a problem that will only grow as communication channels multiply—and it’s the core issue Dashbot is trying to solve through chatbots and conversational analytics. In fact, according to research firm Forrester, 60% of businesses said automating their conversation channels was their number one priority over the next three years, so we know brands are looking for a better solution to this deluge.

However, it’s not quite as simple as just launching a chatbot.

We’ve seen three tiers of chatbots out in the wild: the Reactive, the Reflective and the Proactive. Here’s what each looks like—and how to take your bot to the next level:

The Reactive

The least mature chatbots are a first line of defense for a particular use case, such as customer service.

A good example is a retailer whose call centers are overloaded and which doesn’t want to hire more agents. Instead, this retailer may decide to deploy a chatbot for common customer queries like returns.

This retailer then pulls in relevant data about returns from customer support tickets and transcripts from customer service agents to feed the bot. That’s a significant task for any big box retailer, who would have tens if not hundreds of thousands of support tickets from a 60-day timeframe. Someone has to read through all of these messages to figure out the most common questions and the appropriate responses before the bot can be deployed.

This is chatbots at their most basic.

The Reflective

The next tier of chatbot includes brands that fine tune existing bots to enhance performance.

Most companies don’t ever reach this stage—they think launching the bot is the end goal.

That was the case of a telecom company in Brazil, which created a bot to handle basic customer service questions, such as how to add a new channel to an existing TV package, how to get a refund and whether a technician could come out to fix service. Within a year, the bot failure rate increased from 30% to 60% and the company shut it down.

This is a worst-case scenario that results when a business doesn’t bother with analytics to assess whether a bot is performing the way it should or update the bot to make it better.

Instead, let’s look at the American Red Cross, which has a chatbot named Clara, who answers questions about blood donations, such as where to donate and donor eligibility requirements. The American Red Cross can monitor how many people ask about donating blood and then how many of those users actually found a donation center as a result to help determine the bot’s success.

Or, if we return to the more generic retail example, the retailer could keep an eye on what consumers are asking, along with the overall customer satisfaction score, to better monitor performance. If sentiment is poor for a particular question, the brand can investigate why consumers are complaining and resolve the issue before more customers have a negative experience.

This ability to gauge success with data and adjust bots as necessary is a hallmark of reflective bots.

The Proactive

The most mature chatbots don’t merely exist and update—they seek out problems to actively improve and move their businesses forward.

Intuit is a perfect example. The software giant behind tax prep product TurboTax has a data science team that assesses conversational data to seek out the emerging trends customers are asking about, such as a new tax rebate on electric cars in Florida. They can then amend the bot to answer those questions and ensure TurboTax is able to help customers with even the most recent changes.

This nearly real-time update is proof of a proactive bot that doesn’t wait for problems to arise before improving but instead tackles them before they impact the user experience.