Main content

Don't mix up chatbots and conversational AI. There's a big difference, says Pypestream CEO

Barb Mosher Zinck Profile picture for user barb.mosher May 27, 2020
Got chatbot fatigue? You're not alone. But all bots aren't created equal. Beyond bots are "conversational AI," connecting the dots with your back end systems. Pypestream makes the case.


Richard Smullen is a serial entrepreneur with a deep focus on early-stage technology. In 2005 he created an incubator/VC company SouthWinston LLC that focused on advanced technologies.

One of the companies funded by the investment firm is Pypestream, a conversational AI platform designed to give customers full control of their experience with a brand. But if you think this is another chatbot/virtual assistant, Smullen says you're wrong. 

Chatbot vs. conversational AI

It's important to point out how much the conversational AI industry is growing. In Scott Brinker's 2020 marketing technology landscape supergraphic, the area of technology with the most solutions is social and relationships, and it's the third fastest-growing. Within that category of solutions, conversational marketing and chat have seen a growth of over 70% over the last year. 

There's a big difference between a chatbot and a genuinely conversational experience, said Smullen. First, there are two types of chatbots, rules-driven, and AI-driven. Rules-based chatbots follow a predefined workflow, while AI-driven chatbots leverage NLP (natural language processing) and machine learning to understand what the user is asking or looking for. This second one is more conversational, and I suspect there would be many who would argue that it is true conversational AI.

The difference between chatbots (or virtual assistants) and a conversational experience lies in the integration of back-end systems to provide information to users. Smullen explained that when you take systems like your ERP, support ticketing, or banking system and ingest them with AI, you can then have direct transactional conversations with customers - without needing a person on the business side.

Chat is simply words going back and forth, he said. Conversational AI opens the door to all kinds of messaging and engagement opportunities. Smullen gave Uber as an example as a conversational AI and said he wants to help every company turn into an Uber.

Creating a conversational UI experience

So what does he mean? Conversational interfaces are immersive, transactional messaging experiences. Getting the balance on your credit card, booking a vacation, submitting an insurance claim, or checking the call usage balance on your mobile phone are all examples of experiences you can build with conversational AI as opposed to talking to someone. These are things you can do for yourself faster than if you had to go through a call center agent or a live chat.

Conversational experiences work by connecting backend systems, even legacy solutions, with a conversational AI, automatically surfacing information and actions to the user. Pypestream builds on these experiences with a range of interface features such as carousels (like a listing of hotel rooms), maps, surveys, list pickers, gamification, and more. You can also upload files, and there is an integration with DocuSign for signatures. 

To ensure the type of experience that makes a customer feel like their needs are understood, it's critical to understand the intent, tone, and sentiment of the customer (from what the user types, down to the kind of emojis they use). 

To help companies get started, Smullen said Pypestream has a professional services team that looks for the high activity use cases in a company where there is an opportunity to automate. For example, some companies want to get rid of their call centers or don't want to invest in expensive call center technology and instead provide an on-demand version of themselves where the customer can serve themselves.

The call center is only one example of where conversational interfaces are delivering improved customer experiences. Pypestream has developed use case templates for healthcare (e.g., medication monitoring, appointment setting and notifications, and health tracking), travel, DTC (direct to consumer) solutions, and more. 

These aren't one and done automations either. The company also continually monitors and analyzes how users move through an automation, adjusting and improving the experience in real-time.

Getting started with conversational AI

I asked Smullen how a company moves toward this type of customer experience, and he offered this advice:

  1. Become digital-first. Conversational experiences require complete buy-in from all areas in the company. You need to break factions up and get everyone to accept and buy into an experience that supports everyone's needs in a single instance (marketing, customer service, support, etc.). You need a holistic view of the customer lifecycle. Smullen said you could start with the call center, but you will need support from others to add in additional use cases.
  2. Understand your KPIs. Why are you implementing this solution? Is it to reduce costs? To improve customer satisfaction? When you know why you want to create an experience, you can design it appropriately, including making all the right integrations in the back end.
  3. Have a team ready to embark. Many of Pypestream's customers tried to start internally at first with chatbots, and what they learned is that they can't build these experiences themselves. 

My take

Talking with Smullen gave me an interesting perspective on the chatbot/conversational AI market. It's not that I didn't already understand that simple chatbots do not provide true conversational AI; I did. But what I didn't consider is how advanced conversational AI experiences can be and that I may already be having some now.

But here's the challenge for Pypestream and other solutions - they still look like chatbots. And depending on how they are built and promoted by a brand, they still make people believe they are talking to a chatbot or virtual assistant. I mentioned this to Smullen, and he acknowledged that it is a challenge they face, including trying to convince customers not to call their conversational UIs ‘virtual assistants.' Smullen said chatbots and virtual assistants give the impression of talking to a person; conversational UIs are not akin to a person; they are a self-service application that represents the company.

Basic chatbots might be a good place to start, but you'll probably find out sooner rather than later that they don't provide the value that a true conversational UI does. 

A grey colored placeholder image