Man interacting with chatbot kiosk at airport
Man checking into flight with chatbot kiosk at airport PHOTO: Shutterstock

Chatbots are real and becoming mainstream. Smart speaker sales are through the roof, and Gartner is citing that marketers will spend over 250 million dollars on conversational technologies such as virtual agents and chatbots deployed on websites and social networks by the end of 2017. To assist brands with building their first enterprise chatbots, CMSWire has spoken to an array of AI and chatbot building experts from various industries in order to compile the five things you should consider doing when building a chatbot, and five pitfalls you should avoid.

5 Dos of Enterprise Chatbot Building

1. Pick The Most Suitable Chatbot Platform

First things first, you need to identify where your chatbot will live, both from a technical standpoint and in terms of where your customers will interact with your chatbot. While most of most chatbot building platforms provide easy publishing process to the various messengers such as Facebook, WhatsApp, Telegram and Snapchat, it’s worth taking your time to study the market and choose the platform that works best for your project. “There are numerous platforms that can be used to develop chatbots... It’s advisable to pick platforms that can publish to various other messengers platforms rather than picking a messenger platform because it would enable your bot to reach out to a lot more channels.” says Pradeep Kumar, VP of Technology Innovations at Udupi, India-based Robosoft Technologies.

Kumar went on to explain that businesses also have to consider what kind of features their bot will require, “It’s better to experiment with various platforms and pick the right one that suits your requirements based on ease of developing the bot, NLU (Natural language understanding), rich messages support, integration with other systems such as CRM, CMS and custom backend services,” he said.

2. Leverage Artificial Intelligence and Deep Learning

Nobody wants to receive responses that are purely scripted based on the question, they want the bot to show flexibility and be dynamic. Deep learning in particular is key, so the chatbot’s algorithm is being educated to respond correctly over time. According to Farzin Shahidi, CEO of NextPlane Inc., deep learning and AI will continue to fuel the enterprise chatbot industry, creating a much more immersive environment, and lead to the advancement of conversational interfaces. "The ideal chatbot should be able to use machine learning to power intelligent conversational interactions," says Shahidi.

3. Aim For Human-like Conversation

Our experts agree that a strong chatbot should be able to understand smalltalk and common slang phrases that are part of the lexicon. “If possible, design the interfaces around small talk related to your [industry] since the chances of the user asking about [industry] specific small talk is high,” Kumar said. For example, if you’re selling tea online, users may ask your chatbot about what biscuits they recommend for a particular type of tea — and that’s the kind of question that separates the wheat from the chaff.

To take your chatbot to the next level in this regard, Kumar advises brands to look into their existing chat logs to find common questions and phrases being posed by users. “Convert extracts of those logs into specific intent so that next time users ask similar questions your bot will be able to understand and map it to a proper intent,” he says.

4. Build Trust

Whether you’re being open about your chatbot being a chatbot or not, you want your customers and clients to trust the conversation they’re having. “It’s essential for people to feel comfortable — to gain a sense of trust and empathy from the very beginning of your relationship. In that regard, trust is a prerequisite for everything to follow, and building trust must be the first goal for your AI-powered service,” claimed Snehal Shinde, Co-founder of Sunnyvale, CA.-based Mezi

Heightening this sense of trust could be as simple as throwing up an initial message from your bot that outlines the security measures in place to protect the conversation, as well as the option for the user to delete their data and chat logs at any time — a feature that brands will want to implement as GDPR edges closer.

5. Add To The Customer Experience

As impressive as your chatbot appears to be remember it's about substance not necessarily the "wow" factor. The chatbot needs to improve the overall customer experience. “If customers are using your AI to order groceries, your chatbot should be able to anticipate that they are buying the ingredients for a salad, and so they may pop up with a salad dressing suggestion, or remind them that they forgot to order the cucumbers they ordered last time,” says Shinde.

In other words, your AI needs to do more than just take an order. It needs to be smart enough that it adds to the customer experience as opposed to just completing a transaction, he says. "Not all messages should be an up-sell or cross-sell, suggestions should be focused on adding to the experience, such as 'Do you have plans this weekend that you need to shop for?'," says Shinde.

5 Don’ts of Enterprise Chatbot Building

Whether your chatbot is already built, or still in the planning stages, here are five traits your chatbot should be without.

1. Don’t Make Your Chatbot Too Human

Previously, we touched on how chatbots should engage in human-like conversation. But according to Pradeep Kumar, there’s a line in the sand. "[You want to] keep the conversation human-like, but avoid being too kind or pretentious," he said. It needs to stay on point and be relevant.

After all, the client wants an answer to his/her query and the objective should be focused on being able to offer a response. They don't want the bot to ask endless questions about how their day is going, how they're feeling or how their weekend was. 

2. Don’t Forget Backend Integrations

Chatbot developers often get sidetracked by how the frontend will work, and how they can "wow" the end user with conversational interfaces and smart emoji usage. But in reality, it’s the backend integrations that help furnish the brain of your chatbot. “While the frontend is the 'sexy' thing to talk about, back-end integrations with systems like the company CMS, and CRM are far more important,” said Rajasekar Nagarajan, Senior Chatbot Developer at Silicon Valley-based Ideas2IT. And that’s because your chatbot needs to fetch information from those systems regularly in order to identify users, personalize conversations, respond with correct information and link to the right pages or blog posts.

3. Don't Overload Users With Information

As much as you may want to showcase the intelligence and speed of your chatbot, you don’t want to overwhelm your customers or clients with information — no matter how many emojis you squeeze into the message. Kumar advises that chatbot developers should, “keep the responses to user queries specific and ensure that the response satisfies the user's response. Remember that the user is talking to your bot so that it can give you specific responses.”

Whitney Fishman Zember, Managing Partner, Innovation and Consumer Technology at WaveMaker also weighed in, echoing Kumar. “The biggest consideration is the medium where most consumers will be interacting with your chatbot — and that’s a smartphone. [With limited screen space], your customers don’t want every response to be a novel. Consumers are usually multitasking and want their questions and needs answered quickly and efficiently.”

4. Don't Just Build a Reporting App

This links back to ensuring that your chatbot enhances the customer experience. “It’s important for your AI to anticipate a customer’s needs and offer useful help without having to be asked. Otherwise, you risk being seen — and used — as nothing more than a glorified search engine,” said Shinde. He also shared that to avoid falling into the search engine trap, the chatbots need to stay with the customer throughout their purchase journey.

Nagarajan cautions readers to really think through the needs of their customers, “Facebook Messenger bots are on a tear. However, over 90 percent of them are simply reporting apps wrapped into a messenger. Don't follow this format. Really think about how you can at least make the bot appear intelligent initially, by anticipating certain common conversations.”

5. Manage Expectations

Right off the bat, you want to set expectations at the beginning of each conversation users so they know they are talking with a chatbot and not a human. Now, this one is indeed a matter of taste and strategy, as some brands consciously don’t tell their customers that their livechat systems are populated by chatbots — but if your chatbot isn't leaning on truly cutting edge technology, you may want to set the record straight so nobody expectations don't get set too high. “Expectations are critical,” said Tom Wilson, Technical Advisor at Charleston, SC.-based Citibot. “Don’t start off trying to make your chatbot be a cure-all for everything!  Focus on simple solutions; test, continue to improve and then add functionality,” he told CMSWire.

What’s your go-to chatbot building hack?