What makes a good Smart Virtual Assistant?

May 2nd 2020 in Conversational
Photo by Priscilla Du Preez on Unsplash

As the conventional wisdom goes, a chatbot is a piece of software that can simulate human conversations in both voice and text. Here is a Google trends chart to show the popularity of the term chatbot over the past few years.

The inflated expectations of the chatbot technology has led to its own downfall, with so much of not so good user experience and adoption of the chat widget that now occupies the bottom right corner of almost every website on the planet. Dilbert can sure vouch for it.


Though there are many chatbot implementations across the universe, many IT and Business leaders can vouch for not all of them having a deeper impact on customer experience and bottom-line goals. Very few implementations can engage with their audiences at scale, in a highly intuitive and personalized manner and hence having a lasting impact.

As the dust settles down with the hype cycle, most of the chatbot implementations are getting over what we call as the 'bot hangover' and trying to leverage the true power of language interface, powered by real AI solutions to achieve real business value gains.

How do you go about creating an engaging experience that delivers great value to the firm and the customers? Before you operationalize your Smart Assistant solution, here is a checklist of asks that you should have for a chatbot success story.

Don't have a BOT, because everyone else does. Don't start your chatbot project because your competition or the next door neighbour has one. While there are dime a dozen chatbot vendors who can get a widget added to your website's bottom right corner in no time, seldom they have a lasting impact. The excitement of having a new chat widget soon dies down with the usage and adoption nose diving.

With all the promise around Conversational AI, the business need should be your primary driver and not the technology.

A bot is not a miniaturized website. One of the common mistake made in many bot implementations is trying to replicate the functionalities of a web application inside a chat bot, with most of the UI elements of a GUI, just that they are in a smaller window.

While the underlying technology around Natural Language Processing and AI have been highly commoditized, there is a huge gap in terms of applying them to solve the right problems. Don't try replicate all your website features inside a chatbot. Your customers would be much better off getting their job done using the GUI layer, than typing them out in a small chat widget.

Don't try to make your chatbot a master of all. No human is and neither can your diminutive chatbot. Instead, make your chatbot good in the LU skills of your primary domain. Be it Travel, Healthcare, Insurance or Banking, train your Smart Assistant to have a very sound LU on the Domain Specific Vocabulary. Let it say 'I know only this and I am good at it.'

Any machine learning application matures with age. Include a well designed training workflow process to continuously retrain your models and let it mature with age and errors it makes.

Does it carry humility? "A Great man is always willing to be little." And, this very much applies to your Smart Assistant as well. Your Smart Assistant may be trained to do a lot, but there is so much it may not know yet. Always design your Smart Assistant to be transparent about its limitations and have a very good exit path for any questions it cannot answer or tasks that it cannot perform.

Don't let your Smart Assistant to be the 'know it all'. No one ever is. Instead keep it simple and your Smart Assistant can perfectly be effective if it has great content and is fun to talk to. Your Smart Assistant may start Simple and be upfront with its 'I don't know'. But, let it make mistakes, learn and evolve over a period of time.

Does it have a Personality? Having a well defined personality is critical for your Smart Assistant to have an engaging experience with your users. Remember, the Smart Assistant represents your brand image and it needs to reflect that in the way it interacts with your customers. All the internal training you may do for your human cohorts would as well be applicable to your Smart Assistant as well.

Design your conversations to be friendly, engaging and with a little bit of wit. Follow the typical social dynamics principles that apply to our human interactions. Let it be trustworthy and reliable on the information it provides. The last thing you want is for your Smart Assistant to mislead your users.

How good are its social skills? Does it start with a greeting? Does it apologize when it makes a mistake? Is it polite? All the typical social skills that apply to human would as well apply to a Smart Assistant as well. Have your Smart Assistant trained with a very good corpus of social interactions beyond the typical business interactions and tasks that it is trained to perform.

Do you know your chatbot users? One of the key aspect of having a SA is to take your products and services to channels beyond the website and a mobile app. With messaging as the No.1 digital activity on the planet, one of the significant merit of having a Smart Assistant is to ride on the mobile moments of the messaging apps and having your business services on them.

If your users are primarily on Whatsapp, you may want to have your SA built on WA. Same goes for any other collaboration channel like Slack, Teams or Google Chat. In addition, it is also important to know which customer segment would prefer a Voice based access vs a Text based access. Don't add a Voice interface because it is cool to have.

In summary,

  • Have clarity on why you would need a Conversational Interface for your application.
  • Pick the right use case that would give you the biggest advantage in a short period of time.
  • Know your audience and their preferred channel.
  • Don't let your Smart Assistant act super smart. Keep it simple.
  • Let it have a persoonality that reflects your brand.
  • Humility and Social Dynamics as much apply to your Smart Assistant.
  • Have a clear roadmap on how you want to measure usage, ROI and mature your Conversational AI application.

Natural User Interface.

Until now, we have always had to adapt to the limits of technology and conform the way we work with the computers to a set of arbitrary conventions and procedures.

With NUI, (Natural User Interface), computing devices would adapt to our needs & preferences and for the first time, humans would use technology that is most comfortable and natural to us.

Bill Gates