Chatbots for WhatsApp

AI Chatbots for WhatsApp

As the world is witnessing an unprecedented disruption due to Covid-19 outbreak, as consumers increasingly turn to digital, and as your agents and staff are forced to work from home, having a fast and accurate shopping assistance service via a conversational interface is becoming as important as having a solid ecommerce website. Businesses, across sectors, are racing to add AI-powered chatbot that can transact with consumers on the website and on other digital communication channels. (If you are wondering why chatbots, you can read the articles “why AI chatbots for your retail business” and “how AI chatbots will transform the retail industry in 2020” to understand more). Since WhatsApp is the most popular and widely adopted messaging channel in modern world, the obvious choice therefore will be to launch your chatbot on WhatsApp. But wait – So far WhatsApp did not allow businesses to launch chatbots on their platform and only in the recent past have they given access to their business APIs for automated messaging. Things work a bit differently in WhatsApp (compared to other messaging channels such as facebook) and in this article I will explain what you as a business need to know regarding launching your own chatbot over WhatsApp.

Firstly, Why WhatsApp?

WhatsApp has emerged as the go-to messaging app for over 1.6 billion users around the world. Users find it simple, fast, reliable and convenient. Users can securely share texts, photos, videos, documents in private or in groups. For AI chatbots, a platform like WhatsApp makes a huge difference because most of the users will already have WhatsApp installed on their devices and they prefer it the most to communicate with friends and family. For interacting with their favourite brands, they do not need to switch to a different interface like a website or a new mobile app. Moreover, given the heavy usage of WhatsApp on a daily basis, it’s more likely that the user will open and read the notification messages sent over WhatsApp.

Getting started with WhatsApp Business?

On WhatsApp, your business account is tied to one registered phone number with which you first create a profile of your company or business. This allows whatsapp users to know whom they are messaging to and are receiving messages from. You can then include this number in your marketing programs to get more customers to connect with you.

For small businesses where owners themselves (or with the help of a small team) resolve a small number of customer queries manually, they can install WhatsApp Business app, set it up with a number and can get going almost instantly. For medium to large-sized businesses where customer messaging has to be automated at scale, they require more than just an app. These businesses need to integrate with WhatsApp business APIs to receive, respond and push automated messages via chatbots over WhatsApp.

Getting access to WhatsApp business API for chatbots

WhatsApp’s Business APIs are not open to the public in general but only to a bunch of carefully selected companies globally who are known as WhatsApp API providers. And unless you’re a large multinational firm or an enterprise scale company, it’s very difficult to get direct access. Some examples of WhatsApp API providers are companies like GupShup, Twilio etc and for you to get access you will need to work through these providers.

So what are steps involved for launching your whatsapp AI chatbot?

  1. Pick one of the API providers first
  2. Develop and test your whatsapp ai chatbot in a sandbox environment provided by your API provider.
  3. Submit details of your company (phone number, email, company details, usecase, Facebook Business Manager ID etc) to the API provider for their approval.
  4. Once approved, you are ready for production deployment and your chatbot will be able to receive user messages sent to your WhatsApp business number, respond to these messages as per the chatbot design and also notify your customer with push notification messages.
  5. Once deployed, API providers will charge you on a per-message basis which typically varies from one provider to the other. Note that this is the base charge for API calls only and on top of it, you will incur the cost of development of the AI chatbot and that of keeping it running.

Limitations of WhatsApp chatbot

WhatsApp was built for messaging and WhatsApp wants to keep it that way. They do not want businesses to exploit this channel for marketing and annoy the users. Hence the API integration is quite tightly controlled and usage have multiple friction points that you should be aware of.

  1. Approval process is tightly controlled and is rigorous:

    WhatsApp and its API providers are very cautious in approving business API integration. Before approving, they will review the phone number, authenticity of your business, use-cases your business is going to address via the chatbot and only upon sufficient validation will they approve usage of the APIs.

  2. Enforces strict guidelines for automated messages:

    To prevent businesses from spamming user’s WhatsApp by automated marketing messages or ads, WhatsApp enforces some guidelines for automated messaging very strictly. These are called opt-in messages and template messages.

    • An opt-in is when your customers agree to receive messages over WhatsApp from your business by providing you with their mobile phone number through a different channel such as your website for example. It is your responsibility to store customer opt-ins and to ensure each customer that you choose to contact through WhatsApp Business has agreed to receive messages in the first place.
    • Template messages are messages of a specific format that your business can send to users proactively and not as a reply to the user’s message. Any message that your chatbot is sending proactively after 24 hours since the user’s last message is qualified as a template message. Template messages use placeholder values that can be replaced with dynamic content. The template message needs to be approved by WhatsApp before you start sending them. Any proactive message whose format is not approved as a template message by WhatsApp will be rejected.
  3. Limited UI design elements in AI bots:

    WhatsApp has limited UI design elements for showing messages and accepting user inputs in the chatbot conversations. Unlike other platforms such as Facebook Messenger, it does not support rich UI elements like cards, carousels, buttons, etc. Only texts, links, images, videos, and documents are supported. This will limit your ai bot experience significantly especially if you compare facebook messenger or webchat.

AI Usecases for Whatsapp chatbots.

Despite the above limitations, your AI chatbot on WhatsApp can solve use-cases that are impactful enough for your business and your customers. The most vital being Customer Service Automation. With the help of AI chatbot you can answer FAQs instantly, provide information on order tracking & ongoing promotions, guide users to the nearest stores and also accept customer service requests automatically. Taking a step further, your whatsapp chatbot can switch to live chat with your human agents only when necessary allowing them to handle 3X-4X more cases than they would normally do via phone or over instant chat service. WhatsApp chatbots can also help users discover the right products and cross-sell or up-sell based on AI.


There is no doubt that AI chatbots have become the need of the hour for businesses with eCommerce. At present, it is the most effective way to automate and scale customer support, bring more leads, engage customers in a timely manner and ultimately lead them down the sales funnel. WhatsApp channel is going to take the chatbots to the next level in terms of user adoption and engagement. If your business is already having a chatbot on channels like website, Facebook Messenger then WhatsApp is the next thing you should be aiming at.

Can AI Exist Without Machine Learning?

Can Artificial Intelligence Exist Without Machine Learning?

We have already talked about artificial intelligence and machine learning in our previous blog. We have seen how machine learning is just a subset of AI and that it helps AI accomplish its goals. Now let’s discuss if AI can exist without ML.

Surely not, right? If there’s no ML, how can AI work? ML is a tool used by AI, surely. But wait.. Not so fast !!

You see, the term artificial intelligence existed long before Machine Learning was commonly used. We are talking about the 1950’s and 60’s here. There were systems which claimed to use AI to work. They are known as GOFAI systems but they didn’t use ML. So what’s going on here? What’s the catch?

What is GOFAI?

Before machine learning, AI generally used a form of AI called GOFAI. GOFAI stands for Good Old Fashioned AI. It is also referred to as Symbolic AI. 

The core belief of GOFAI is that any complex system can be defined using human readable symbols and rules. The rules would define the processes that take place within the system. Some examples of Symbolic systems are,

  1. Algebra – Here we use x,y,z to represent variables and +,-,x,/ to represent operations between variables. So an equation like x = 3y + z is termed as a rule.
  2. Chess – The pieces like king, queen, bishop are symbols and the moves are the operations between the pieces.
  3. Programming languages – many languages use certain keywords like “for, while, class, object” etc. So they would be symbols. Statements like “for x in list: x = x + 2″ would be called rules.

These are obviously not AI systems. But they serve as examples of Symbolic systems. The idea behind Symbolic AI systems is that we can take any complex real world interaction and encode it in this manner.

GOFAI also deals with a class of AI problems called Strong AI

What is Strong AI?

AI problems which fall under Strong AI usually deal with complex real world scenarios. But the best way to understand Strong AI is to look at what weak AI means.

Weak AI problems are narrow in nature. They deal with a specific scenario and nothing else. Most modern machine learning and deep learning tasks are Weak AI problems. It tries to solve simple problems like image classification, text classification etc.

Consider the case where we have to identify accidents in traffic cam feed. The solution would be to build a system which takes the video stream as an input and look for specific signs of accidents in real time. It needs to be able to identify a car crash and the like. This is considered to be a Weak AI problem as it solves a narrow task. This AI would not be capable of doing other tasks.

A Strong AI system is to model more complex systems such as a general artificial intelligence which can act, feel and do as a human would.

GOFAI was seen to be trying to solve these problems. Although unsuccessfully.

Expert Systems

There are many examples of GOFAI systems out there. They are also called expert systems which solve problems in a specific domain. These systems use Symbolic AI to create symbols and collect data from domain experts to form production rules. Examples being

  1. Phone IVR systems – These systems are created by collecting data about what users generally call about. Using this information, an IVR flow chart is created which guides the user through its various functions. These systems encode production rules as flow charts and use symbols on the phone keypad which it can recognise (e.g. 1, 2, 3)
  2. Automated Medical Diagnosis – Systems are capable of identifying specific conditions using patient data, reports, scans, etc. Some of these systems rely on a vast amount of data from domain experts. The final GOFAI system is used to check this knowledge base to find a match. This type of system uses a knowledge base and an inference system.
  3. Simple chat bots – Chatbots fulfilling a simple flow of customer service, food ordering, and ticketing, etc also use a form of GOFAI. They are quite similar to Phone IVR systems, just with a better interface. They follow a flow chart, guiding you through button clicks to reach a specific goal.

Is GOFAI really AI?

So we have seen how AI can exist without ML. Such systems use the concept of GOFAI and symbols. But GOFAI has its detractors. Many do not consider it to be AI at all. There are many arguments against GOFAI and whether or not Strong AI is achievable.

One of the more interesting arguments against Strong AI is John Searle’s Chinese Room Argument in his paper “Minds, Brains, and Programs”. In it Searle asks us to consider that he is sitting inside a closed room. He accepts small pieces of paper from a slot box in the door. His task is to translate from English sentences to Mandarin sentences. He would then use phrase books within the room to help him finish the translation.

He compares Symbolic AI systems to this chinese room experiment. The person within does not understand Mandarin, but is still capable of completing the task. But this does not mean that he truly understands the language at all.


Although GOFAI and Symbolic AI has its opposers, it is true that these systems are all around us. They do not use machine learning, instead they use symbols and rules but still many claim that they use AI. Detractors of GOFAI, on the other hand, argues that systems which uses symbolic AI are not really intelligent and therefore cannot be treated as AI systems.

Curious Much? Drop us a note at or visit our website and let’s start a conversation!

Top 4 reasons Why AI Projects fail?

Top 4 reasons Why AI Projects fail?

For those who are new to AI technology, Artificial Intelligence is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans. The term “artificial intelligence” is often used to describe machines that mimic “cognitive” functions that humans associate with the human mind, such as “learning” and “problem-solving”.  To understand more on AI read this blog article from Tejas

Despite increased interest and adoption of artificial intelligence (AI) in the enterprise, 85% of AI projects ultimately fail to deliver on their intended promises to business, according to a recent industry report. In this blog post, I have put together the top 4 reasons for failures

1. Organizing your AI team

Building your own AI team is not at all easy. Unless you have a budget the size of Google’s, building an in-house data science team is costly and time-consuming. There is a shortage of data science experts everywhere, and the experienced ones are often involved in engaging projects with big projects and companies. So, in this constrained environment, what is the best way forward? If you are starting fresh then trying to build a big team upfront is absolutely unnecessary. At the same time, starting out with college graduates and training them to become experts can be a costly mistake. In the quest of saving dollars, you are compromising on the quality of the deliverable. Experienced folks, usually from engineering background, with deep understanding of the business and the data are the right choice to start with. In we started without an AI expert in the team. We put our best and experienced folks to take a dig at our AI expeditions. We laid out an implementation plan with well-defined and small milestones. One needs to get into a hypothesis and experimentation mind-set. Validate ideas quicker, eliminate or accept based on experimentation outcome and move ahead. In this process, we learnt from our mistakes and we improved at every step and came back strong. The first set of AI engineers then mentored the fresh and lesser experienced AI folks and helped them in productionizing their AI experiments. Today we have a sizeable AI experts team who can go all guns given a problem ahead of them.

2. Lack of proper Data Strategy

The importance of data strategy is often underestimated. In the past, data was just a by-product of processes and business activities. It’s different now. Data is the most valuable resource that allows companies to gain competitive advantage and come up with new ways of improving their operations. In most cases when dealing with AI projects, collection of data is a challenge. There’s not enough of it and if you are somehow able to get hold of the data then there can be issues with labelling, training data, etc. Data doesn’t just have to exist, it has to be labelled. If data isn’t properly organized, humans have to devote their time to the tedious task of labelling it. Data labelling is troublesome and time consuming, yet somehow many companies just don’t think about it at all. Because an AI system can only be as good as the data it’s fed with, you can’t have any tangible results if there’s no meaningful data behind it. Here at, we have tools developed to automate data collection, validate data quality and easily manage it. You have to treat your data like gold and ensure it is not corrupted by any means and over the last 3 years, we have tried to carefully automate a large part of this process.

3. Unrealistic expectations from AI technology

Many organizations are tempted to follow the big players like Amazon, Google, Microsoft, etc. It’s just logical to assume that what they do makes sense. However, companies like Amazon have gathered enough capital to test various, even seemingly crazy solutions with very lofty goals. They can afford it, and experimenting with technology is great. However, smaller companies cannot afford to take so much risk. Observing the giants, learning from their successes and failures – that does bring value. But keep in mind that each business is different and has different needs. Don’t let yourself lose sight of what matters to you. Do some research, get your strategy ready, and then play your game accordingly. Implementing a new piece of technology can be challenging, and that’s something organizations must always remember. AI is not a magical tool – it needs the right use case matching your objectives, the right data, and success criteria, it also has certain limitations. When you identify the AI usecases you think is right for you, learn both the possibilities and the limitations the different AI solutions have to offer. It’s an important part of the process to know what AI can do for your business, realistically, and be able to later track the performance of the deployed models. Please remember – there is no such thing as perfection, even for AI. AI also makes mistakes. No AI algorithms will be 100% error-free. In our case, we have a Question and Answer system which can answer user questions asked through the Chatbot on the client’s ecommerce brand website. There will be times when the AI model is less confident about which question the user is asking and therefore what should be the answer. This is the case when AI is not able to clearly decide from its trained AI algorithms. In such cases, we handle this gracefully by providing the user with Did you mean? – Question suggestions. Small strategies like these will help your organization avoid your AI failing badly but more importantly also allows the AI to self-learn very quickly. Continuing on the same example, today our AI technology is able to answer 80% of all questions asked by website visitors of our client thereby prompting the client to take our solution and launch it even in their retail stores.

4. AI Bias

AI bias, or algorithmic bias, describes systematic and repeatable errors in a computer system that create unfair outcomes. Though the name suggests that AI is at fault, as described above, it really is all about people and how the AI is trained. Chief Decision Scientist at Google, writes: “No technology is free of its creators. Despite our fondest sci-fi wishes, there’s no such thing as ML/AI systems that are truly separate and autonomous…because they start with us. All technology is an echo of the wishes of whoever built it.”.

We faced this situation in with our ecommerce NLP intent model. We saw that more and more user questions and complaints were classified as product discovery. What this means is questions like “How to wash cotton tops?” showed cotton top products, which clearly is not what the user wanted. Upon investigating we got to know that the engineer who tagged it was more biased towards classifying such sentences as product discovery. This created an imbalance in the knowledge-graph which lead to incorrect and biased intents model. To solve this, we had to revisit the labelling and correct it. To avoid such problems, it is advisable that more and more people validate the labelling before it is pushed to the LIVE environment. Again, this does not mean extensive manual effort but smart algorithmic ways to sample the right data set and validate is enough and can do the trick.

One famous example of AI Bias is Microsoft’s AI Chatbot Tay – which could automatically reply to people and engage in “casual and playful conversation” on Twitter. By flooding the Chatbot with a deluge of racist, misogynistic, and anti-Semitic tweets, Twitter users turned Tay – a Chatbot that was described as “a robot parrot with an internet connection” – into a mouthpiece for a terrifying ideology. After a failed effort to clean up Tay’s timeline, Microsoft pulled the plug on their unfortunate AI chatbot experiment.

Conclusion: So, how can you ensure a successful AI implementation?

The good thing is that with so many organizations having already failed at AI, you can learn from their mistakes and avoid making the same ones. It’s a good practice to observe the market, not just in your direct competition, but also in the tech world. This way, you will know what you can realistically expect, what use cases are promising, what limitations you have to take into consideration. Take ‘Fail Fast, Fail Early’ as a motto when it comes to AI. You have to experiment, test hypotheses, go down some blind alleys, learn from them and use these to find success. Having processes and methodologies for agile, rapid experimentation that delivers results quickly and at a relatively low cost is critical for success. Do not expect to draw up the perfect AI strategy for your business at the start and to execute it. Start with the assumption that you will have to evolve and change this as you progress.

If you like this blog or have anything to share, please leave your comments at