Will Chatbots replace Apps?

Will Chatbots replace Apps

Will Chatbots replace Apps? Are Chatbots the new Apps?

The year 2008 when Apple launched the App Store, that brought revolution in digital experience. Everyone wanted a mobile app for their business and soon we had thousands of mobile app and hundreds of mobile app development providers. I have seen many mobile apps built by businesses but quickly taken out because they were not really serving the purpose and customers were not finding any value in having them. Why? Because mobile apps needed the enterprise data, embedded in multiple systems, to be made available in the form of lightweight APIs (something that did not exist), and that too in a certain structure. This in turn, meant that enterprises adopting mobile apps had to then invest significant time and money to prepare the data, build the APIs and then expose it. A costly mistake to jump in without the holistic understanding and more importantly without answering the question “Why my business needs a mobile app?”!

Fast forward a decade and we are seeing the same pattern emerge for new age digital experience – Conversational AI or Chatbots, targeted towards always-on digital native generation who loves more conversational experience than the old-world menu driven websites & mobile apps. Most businesses see this as an opportunity (and rightfully so) and want to capitalize. Wouldn’t it be great to serve your customers in an easy flexible conversational medium that can understand the intention of the customer and respond with contextually relevant answers in natural language? Sure it will be. But before you dive in and start building chatbots for your business, if we have learnt from the past and want to avoid our past mistakes, it is important to assess chatbot advantages and disadvantages and ask a few fundamental questions around your organization’s data readiness.

Here are the top 5 questions that you must strive to answer for your business before attempting to build and launch your chatbot.

Top1: What is the use case I want to solve using chatbots? 

Don’t build chatbots because chatbot technology is the next exciting thing technology world has to offer and everyone is building one. Identify the problem or AI use case where conversational AI agents are the best option and then build one that solves for that specific problem. Or else it will be like putting the cart before the horse and you will end up being disappointed.

Top 2: What data will your chatbot app need and Do you have this data?

Driven by the use-case and customer journey you implement, this might vary from customer service data, orders data, product data, inventory data etc. Make sure you draw out the data map and assess whether this data exists within the organisation or not? It is quite common that you might not have all the data that will be needed and in these scenarios you must figure out alternate sources – either gather data from external public sources or generate your own data. (We will soon share in a different article some of the hacks we have figured out to generate your own data when it does not exist)

Top3: In what format should you provide it to your conversational AI?

Do you need to build APIs in a certain structure that will expose this data or Can you expose your existing data as raw extracts and the conversational app can intelligently make sense out of it? Look at building a full stack chatbot solution where the chatbot is wired to the backend data ingestion and data enrichment algorithms extremely tightly.

Top4: How to make sure that you are able to extract knowledge and insights from your chatbot  app and take these learnings to the other touchpoints?       

I will again recommend you to build a full-stack chatbot solution that comes with its own Analytics capability and getting insights from the context rich conversation data should be as simple as a click of a button. Remember, one of the primary business case chatbots deliver is that if designed properly, these AI applications can deliver unique precision marketing insights that even GA cannot deliver. The reason being that these insights are coming out of conversation data that typically contains a lot more customer context information than the web-click data on which your GA runs.

Top 5: How to break the I-C-E and how do you keep it up-to-date? 

ICE– Intent, Context and Entities are the defining building blocks of NLP (Natural language processing) and here is what each of them means for your business.

  • Intents – Intent in simple terms mean ‘the question behind the question’. It is about predicting the “intention” of your customer from the message she sent to your bot and the bot then responds accordingly. Deeper the intents understanding, richer will be the conversational experience for the end user. Shallow conversational apps will decipher 3-5 top-level intents while Vertical AI chatbots will be able to interpret and predict 50-100 intents that are specific to your business and your product category. Remember, in the world of conversational apps and virtual assistants, the most important aspect differentiating one solution from the other is the capability to predict the user’s Intent. Go for vertical AI apps that understands your domain ontology deeply and there-by the accuracy and depth of their intents prediction model will be far higher than the commonly available horizontal chatbot apps.
  • Context – This refers to the context in which your customer has sent the message. Is she trying to discover products when looking at a category or is she asking a question on the material when looking at a specific product? Finer the context, more engaging the conversational experience.
  • Entities – these are domain specific significant terms such as denier, pregnancy suitable, veins and scars, weight, plus size, etc which has a meaning in relation to your business and products. Wider the entities more relevant will be the conversational app response capabilities.

Success or Failure of your Conversational AI solution is fundamentally driven by the data preparedness of your organisation and answering the above questions upfront will help you to avoid the same mistakes we witnessed during the Mobile App era.

I look forward to your feedback, thoughts, brickbats – just reach me at dinesh.sharma@asksid.ai and let’s start a conversation.


Traditional Online Chatbot or Domain Specific Conversational AI? What can create a delightful customer experience online?

Want to create a delightful customer experience online? With Traditional Online Chatbot or Domain Specific Conversational AI Platform?

Differentiating customer experience starts from your website. If web-shop is the superstore for your brand – allowing prospective customers to walk-in, check out and purchase your products, wouldn’t you need your best salesperson to greet your customers, answer their product questions and assist in making a purchase, and in turn, delighting them? Precisely what Prajwal discussed in his last post “Why your brand should move beyond Contact Us?”

Just the concept of automating and scaling up one-to-one conversations using technology appealed to lots of brands and they tried to leverage Online Chatbots. This was in 2015. Naturally, most brands were thrilled when they implemented an Online Chatbot expecting that the experience will drive prospective users to make a purchase and save cost by handling customer service issues and FAQs.

But, despite the best of their intentions most of these traditional online chatbots (I call them chatbot 1.0) failed to deliver the user experience that was as seamless, delightful and efficient as they envisioned them to be.

Reasons why online chatbot 1.0 failed, while a vertical focussed conversational AI delivers the experience

1. The traditional online chatbots are not intelligent enough

Most traditional online chatbots work based on rule-based or decision-tree logic. This means the experience is limited to the thoughtfulness of the customer journey designed by the brand, and to the list of possible FAQs that the brand can come up with. Machine intelligence is driven by the data on which the chatbot is trained and often brands struggle to furnish enough relevant data needed for effective training. (refer what our CTO has to say regarding “relevant data” in our Curious Much Conversation talk series here)

The other aspect is the self-learning capability of the chatbot and most often traditional online chatbots miss this critical part. Wouldn’t it be naïve to expect your sales associate to increase his skills and win deals, handle customer queries without learning on what you got to offer?

“Intelligence is the ability to adapt to change.” Stephen Hawking

An employee becomes an asset when he or she masters your products or services through continuous learning. Likewise, Domain-specific conversational AI matures into an indispensable asset for businesses with its continuous learning capability. Learning for a Vertical AI Bot is not limited from its conversations alone with customers, but by enriching the product details from various other sources; managing future conversations efficiently. Capturing vertical/product specific Intents and customer contexts truly help the AI to mature into a Vertical AI.

2. Online Chatbot (1.0) fails to understand the context and intent.

To make the conversation more appealing for customers, management of context is of paramount importance. Pure-play chatbots can’t handle contextual information for an extended period. Losing track of conversations context means ending in frustrating conversation and disgruntled customers.

A Vertical focused conversational AI can have far richer conversations because of its ability to predict the intent of the users, apply the context, and recognize relevant entities that are meaningful to the brand’s domain and product category.

A common marketing narrative from traditional online chatbot players is that they are Vertical AI and that they can identify “intents” accurately. Peel through the layers and you will see the Intents library with 5-10 unique intents that are generic across industries. The true measure of a vertical conversational AI is the depth of unique intents relevant to the specific domain that it can identify and predict.

3. Switching over to human expertise when the technology fails – LIVE CHAT handoff

To achieve seamless conversational experience, being transparent is key. The shopper needs to know upfront that they are talking to a machine and an option to switch to an agent when in need exists. Additionally, a true domain-specific conversational AI should also be able to accurately predict when is the right time to switch to a human agent and this is where the self-learning capability of the AI makes a difference.

Domain-specific conversational AI can turn conversations into a pleasant one by handing over the conversation at the right moment to a human agent. Not just stop there, but also allow the individual agent to replay the conversation, make him aware of what the shopper did right before connecting to him and resume the conversation from exactly the place where AI switched over – all this in a seamless experience for the shopper.

In summary, domain-specific conversational AI platforms that understand your specific product segment or industry and comes pre-loaded with its own intelligence (intents library, entities library, etc) can help you jumpstart your journey towards creating a delightful customer experience.

Curious Much? Let me know your thoughts at curiousmuch@asksid.ai and let’s start a conversation.

Author: Joseph Mathew (Joseph.Mathew@asksid.ai)