Building SuperText: WhatsApp based Conversational Commerce

How did a hacker news article lead to building a WhatsApp hack that grew into an enterprise AI company.
January 9, 2021


How did a hacker news article lead to building a WhatsApp hack that grew into an enterprise AI company.

Supertext was founded as a startup with a simple promise: We bring you anything if you’re willing to pay for it. Yep, anything. Anything that world well within the legal framework, but anything at all.

What can go wrong?
Turns out, quite a lot.

The year was 2015, the city was Bangalore and the buzz was about hyperlocal. Almost everyone who had a business now had an app. Every app had a delivery option and we wanted to be the layer that tied all of this in the offline world, together. My journey at SuperText lasted for about 2 years from conceiving the idea to being part of the founding team to watch it come to life. Looking back, the company had 2 distinct acts on what it did.

Act I: Building a WhatsApp hack, concierge for over 80,000 customers

When I tried using the ecosystem of apps in the hyperlocal category as a college undergrad, I was surprised at the number of them in Bangalore alone — over 82 to be precise. Surely, the city did not need 14 laundry applications, or 20+ food delivery applications. It seemed like the Uber for X model had gone to an all-time fervor in the city, and customers were not at the center of most of the decisions to commission a native application to be built.

The original idea was to create a service using text messaging instead of going through several steps on an app to order food, a driver or other things people might want. The problem? text messaging was near irrelevance in India, and the ecosystem was around WhatsApp as a messaging application. The problem with WhatsApp? It’s a closed ecosystem. No developer access back in 2015 at all.

Engineering a solution was not only necessary but also meant the very survival of the idea and the company. We worked on a number of approaches to engineer a hack or a roundabout method to use a REST API to communicate using a single number, to thousands of customers. All this done via a CLI and eventually a GUI dashboard with multiple operators on customer support.

We created a python library that allows you to login and use the WhatsApp service and provides you with all capabilities of an official WhatsApp client, allowing you to create a full-fledged custom WhatsApp client.

We eventually figured a way to make a multi-operator, single-number WhatsApp based dashboard work. It was mostly grit, spit and duct-tape holding this or barely holding this together, but it worked. The damn ship had holes, but it still was above water. We now had a system that can accept orders, communicate, send invoices and interact with our customers.

Then came the hard part of launching the business and fulfilling orders.

The Launch

When you think about it, it seems a little ridiculous that we are so lazy we’d rather have someone else deal with an on-demand food, grocery, car, whatever delivery app than deal with the service directly. But companies from YC that year and some more seem to show us that a lot of people like ordering whatever they want and are willing to pay to get other people to take care of the hassles.

Why step out when you can get everything delivered home with the tap of a button? The Supertext experience is built on top of WhatsApp, so you don't need to download yet another app. Supertext gets you Food, Grocery, Medicine, Pet Supplies, Fruits & Vegetables, Meat & Fish, Health & Wellness, Gifts and Send Packages from one end of the city to the other. From your local kirana stores to your favorite brands, grocery shopping to your forgotten charger, we are always on the move for you.

We decided to take the theme of “Be Lazy” and make that the company motto and tag line. To go with it, we needed an approachable brand.

Why?

We want to change the way people get things done, so that you can spend more time doing things you love.  We sensed the need for a platform that serves as a single form of communication for ordering/discovery/availing services and customer services. Once the idea was verified, we chose to launch ourselves as a chat-based service. The reason we chose chat-based service over other forms of communication is because the modern-day customer is well-acquainted and addicted to chatting, convenience, multi-tasking capabilities and the ease to share images location via chat.

What?

Supertext was aimed to be your shortcut to get things done on-demand, a one-point portal that uses a chat-interface to fulfill all your daily tasks. A thin layer that aggregates demand and then routes it to the most appropriate service in the backend for fulfillment.

How?

We prioritized the lowest possible amount of friction to the whole experience. That meant there were no apps to download. Supertext was a Whatsapp only service on 8884455506. You tell us what you want, choose from options, pay and get the product/service delivered anywhere all over chat.

Did people use it?

Big time. The launch was supposed to be a mellow beta to see how things pan out. A 2 week sprint of cobbled together ideas and 1 delivery guy. An experiment by all means, that went off the charts by the end of the first week.

By the end of 6 months, we had over 80,000 unique users in Bangalore alone. All of whom had an association with just one unique phone number they texted to, right there between their groups and their other DMs, on WhatsApp. We made some noise, some process, burnt some money and learnt a thing or two about user experience and unit economics.

For all you who were/are in Bangalore — you must be wondering why this sounds eerily similar to another service you probably are using. You are right, and yes, we did predate Dunzo by months before they launched, similarly, on Whatsapp. Now, back to the journey.

What did people use SuperText even for?

The sheer breadth of tasks that were coming to Supertext took us by surprise, but not before taking our knees out whilst running 24 hour shifts to fulfill orders. The reviews started pouring in. We didn’t have an app so these went right to our Facebook page, and hundreds of people began to thank us for that bag which got delivered or those keys that made it from the office or the late-night pizza that saved the hunger pangs.

To make this happen, we not only had to spread ourselves horizontally, but cater to over 45 categories of unique information or product-based requests that came in, on a daily basis. They were:

  1. Food: Items from restaurants, fast food junctions, roadside vendors
  2. Beverages: Juice parlours, Alcohol etc.
  3. Alcohol: Bought from vendors also outsourced
  4. Vegetables and Fruits: All vegetables and fruits whether pick up and drop from vendors or outsourced to other vendors.
  5. Meat: All meat, pork, beef etc.
  6. Dairy + FMCG + Provisions: Packaged Food, Shampoos, Aata, Maida and stuff you may get in Provision stores, Foodworld, etc.
  7. Baked goods: Cakes, cupcakes, baked cookies etc.
  8. Flowers: All flowers, bouquet orders
  9. Balloons: Helium, Hydrogen and Normal Air Balloons.
  10. Real Estate: Housing (Buying, Selling, Rentals) and properties
  11. Gadgets Rentals: DVD, Laptop, DSLR, Violins etc.
  12. Transport Rentals: Bike, Car, TT, Tempo, rentals
  13. Govt. Documentation: PAN Card, Passport, Khata Deed
  14. Home Servicing: Plumbers, Carpenters, Electricians, Home Cleaners, Maids
  15. E-commerce: Amazon, Flipkart, Snapdeal purchase. Items ordered online, delivered to office in 2–3 days and then delivered to customer
  16. Pick up and Drop: All pick up and drop orders where SuperText moves stuff
  17. Table Reservations: Reservations
  18. Travel Bookings: Flights, Railways, Airways
  19. Cabs: Ola, Uber, TaxiForSure, Auto, etc. for intra city travel
  20. Show Bookings: Movies, Shows, Concerts, Drama, Festivals etc.
  21. Medicine/ Pharmaceuticals: Over the counter + prescription items
  22. Stationery, Books, CDs: All office supplies, Personal Stationery,
  23. Gifts and Greetings: Fancy Gift Items, Showpieces, Greetings
  24. Event Planning: Birthday planning, Party Planning etc.
  25. Clothing: Undergarments, Tees, Shirts etc.
  26. City based queries: Closest ATM, etc.
  27. Offline Establishments: All shops, restaurants, stores etc. (Product bought and delivered)
  28. E Commerce Portals: Amazon, Flipkart, Snapdeal, Paytm (Delay involved in sourcing the product and delivering it)
  29. Booking Portals: BookMyShow, MakeMyTrip, etc. (Instant Booking Services, No man movement)
  30. Aggregator Service Providers: Shippr, UrbanClap,
  31. Individual Service Providers: Individual Drivers, Maids, Flat Agents etc. providing service to the customers
  32. Food: Items from restaurants, fast food junctions, roadside vendors
  33. Beverages: Juice parlours, Alcohol etc.
  34. Alcohol: Bought from vendors also outsourced
  35. Vegetables and Fruits
  36. FMCG + Provisions
  37. Baked goods: Cakes, cupcakes, baked cookies etc.
  38. Flowers: All flowers, bouquet orders
  39. Balloons: Helium, Hydrogen and Normal Air Balloons.
  40. Real Estate: Housing (Buying, Selling, Rentals) and properties
  41. Gadgets Rentals: DVD, Laptop, DSLR, Violins etc.
  42. Transport Rentals: Bike, Car, TT, Tempo, rentals
  43. Govt. Documentation: PAN Card, Passport, Khata Deed
  44. Pick up and Drop: All pick up and drop orders
  45. Table Reservations: Reservations

Phew.

Maybe our systems needed an upgrade.

Customers began to rave about us

The biggest validation for why Supertext was working as an idea was in our Facebook and WhatsApp groups. We had an experience rating average of 4.8 on 5, for over 80,000 unique orders.


SuperCentral entered the chat

We realized that the early WhatsApp workaround we had engineered had been successful but now began to be a big bottleneck in terms of speed. We got back to the drawing board to come up with V2. This time, the powerball feature would be automation. Automatic replies, order summaries, delivery estimates and more. ‘Bot everything’ was almost the only phrase spoken at our terrace office.

At it’s core, Supertext is a human augmented intelligence system called SuperCentral that uses customer profiling tools, hyper-local search integration and pricing algorithms to find specific options for multiple verticals.It also enables speedy interaction and fulfilment enabled by the use of AI algorithms, Optimised search and a Centralised full-stack vendor collaboration platform; has allowed Supertext to deliver on happy experiences and delight users with easy decision fulfilment.

We worked hard on solving problems such as consistency in pricing estimates and hundreds of other aspects that define a good experience for the customer, that went into a ‘Promise Engine’

Whilst building SuperCentral we encountered 3 problems:

  • WhatsApp was increasingly more hostile to third party clients and non-app based communication workarounds.
  • We were losing money each order we delivered. Whilst we had industry leading unit economics, it was still not enough to make a profit, all things considered.
  • WhatsApp took notice on the significant volumes of messages on our number and just decided to permanently ban us, and overnight, we got squeezed out.

No matter what we did, this might not end up scaling the way we wanted it to, and building yet another app seemed purpose defeating from where we saw it. However, we had made some progress on the chat automation, intelligence and chatbot side of our tools and now began to wonder if there was better use for it outside a delivery business tied to the hip. In short, we wanted to go away from being a hyperlocal delivery metalayer to a software only firm.

What can go wrong?
Turns out, plenty.

Act II: Building a Chatbot Company

Our networked world has changed commerce forever. A level of online service that was once a luxury now feels like an inalienable right — 24/7, instant gratification is now the minimum we expect. Businesses are racing to keep up with consumer demands and technological innovations. The next phase of digitisation often casts its shadow upon us as we are still perfecting the last. Chat Bots back in 2016 promised a swifter, smarter online experience.

Our new virtual assistants were supposed to be ever-ready, able to listen to our questions and respond intelligently. They will answer our queries, aid our searches and anticipate our needs, learning all the time to refine and improve the experience on offer. The promise from developers is that Chat Bots will revolutionise online customer experience. However, many businesses may ask; are Chat Bots an innovation and investment too far? Do we really need virtual assistants to guide our customers online? In terms of information and access, people have never had it so good — why do we need to offer even more? Do consumers even want AI virtual assistants? What additional benefits could they possibly provide? Are they just a novelty item I can do without?

AI Chat Bots go one step further, providing simple but intelligent automated responses and / or services within a messenger platform. This allows you to ask a Chat Bot simple questions and have it deliver a helpful response — sharing opening times, delivering search results based on recognised phrases (e.g. “show me the red shoes you have in stock”) or carrying out simple administrative tasks (e.g. making a reservation at a restaurant).

Early attempts at Chat Bots are relatively simple, but then came the opening of the floodgated with Facebook and the Messenger API. Independent developers could now create their own bots, accessible to its huge global user base. Conversational Commerce was going to be the next big thing, people said. At Supertext we decided to punt on this movement of the world and build a tool around the needs of customers when they are creating their AI chatbot.

Supertext built conversational experiences for brands and businesses using Artificial Intelligence. Our tool enabled enterprises to deploy intelligent chatbots to several platforms quickly and easily.

Most existing chatbots back then and now are brittle and fail to provide answers for customer queries that stray from the intended purpose. They do not comprehend natural language and context. At Supertext, we employed 3 layers of neural networks between the customer, enterprise and the chatbot to provide for the most contextual answers to even daunting queries. Our bot training mechanisms are easy to deploy and iterate upon. This ensures that the bot continuously improves with the data generated through conversations


Summary:

Supertext enables brands /businesses to engage with their consumers in human like conversations through artificially intelligent chatbots, across various interfaces (e.g. facebook messenger, whatsapp, native app) for complex tasks in customer support, brand assistance and/or sales generation.

Supertext builds tools to make conversational commerce intelligent, automated and personalized. Our SDKs allow your business to come onboard messaging products. Products include Software Development Kits, mobile messaging products and service level support tools.

At Supertext our products encompassing smart, versatile Chat Bots typically fall into three key service areas:

Commerce Bots

Chat Bots that allow customers to browse, select and buy products without ever leaving the messenger window.

Customer Service Bots

Intelligent, responsive and always-on customer service Chat Bots can live on brand Facebook pages to efficiently respond to users’ queries and complaints with relevant, useful information at any time of day.

Content Bots

Chat Bots that allow users to create personalised news feeds, bringing them timely information and alerts on topics of their own choosing.

Supertext enables brands /businesses to engage with their consumers in human like conversations through artificially intelligent chatbots, across various interfaces (e.g. facebook messenger, whatsapp, native app) for complex tasks in customer support, brand assistance and/or sales generation


Problem:

Today, consumers expect instant resolution of their instant personalized needs or questions. As brands are not being able to instantly address a customer’s needs, brands are losing more and more customers. Over 87% of consumers switch brands and do not return for at least 2+ years due to low resolution time and bad customer service experiences. To fulfill the increasing consumer expectations, brands are required to increase spending on customer support executives, training and infrastructure to retain existing consumers while adding more manpower/ infrastructure to increase sales fulfillment of personalized needs.


Solution

In order to address increasing customer expectations in a more efficient and effective manner, Supertext is building tailored chatbots made for specific business needs that can be deployed for multiple consumer touch points across various interfaces or channels. Our chatbot can provide an end-to-end solution for mobile engagement by eliminating any communication breaks and saving manpower, training and infrastructure costs.

No matter how clever the technology, Chat Bots will only find favour with customers if they improve our current online experience. We asked our research panel to assess the state of play and found that — despite the many innovations in online service — there is significant scope for improvement. Online customers reported the wide range of frustrations they experienced in a single month. Most common (45%) is a lack of basic information contained on everyday commercial websites (retailers, utilities, banks, local government services etc). Close behind was the inability to ask simple questions (40%). A third of the sample (33%) said that, even when the option to ask questions existed, the tools they had used in the last month were of poor quality and didn’t provide a timely response. These, and the other complaints detailed below, indicate the need for online service to improve. Businesses offering the best customer experience will be at considerable advantage in converting browsers into buyers and earning repeat business.

Consumers regard Chat Bots as the key-holders to speed, unlocking immediacy and convenience in online services. This mirrors where current frustrations with online services are most acute — where poor application hampers the speedy resolution of a task or issue. These factors highlight where developers need to concentrate their efforts, ensuring Chat Bots bring about positive change. The ‘personality’ of the Chat Bot barely registered with consumers –AI needs to be smart and effective first — witty and personable second. The graphic (right) shows 40% would use Chat Bots to help track down the right human assistant — possibly logging a request to speak directly. Just under one in five consumers (18%) admit to actively avoiding calling businesses to elude the frustration of automated menus. In those under 24, this rises to a shocking 61%.

We pinned our tool on 3 core features:


Conversational Design: Chatbots are driven by free flowing human like conversations as opposed to rule based/ guided flow chatbots

Speed: Framework that allows for rapid prototyping for unique use cases. Deployment & training of chatbot within 2 weeks.

Modular infrastructure: Deployment across multiple interfaces and for different uses cases such as Customer support, sales generation or assistance.

Markets included building experiences that solve for intricate business use cases in verticals such as Banking, Travel, Product discovery and Insurance assistance. Banking — we engaged and worked with one of Singapore’s largest banks to enable intelligent personal financial assistants as a service for their customers. This involves a significant understanding of the banking processes, checks and measures in place even on a pilot program strata Travel — we built a next generation travel experience that allows for seamless discovery, bookings and trip organizing between you, your friends and the place you are travelling to Insurance — we deployed a personal insurance recommendation bot for one of Singapore’s largest insurance companies

It’s interesting that the industries exhibiting the most frustrating customer experiences online are the least likely to improve them. Public sector bodies in particular are notoriously slow to provide accessible online services — and when they do, they’re often inadequate, confusing and riddled with poor UX. It’s one thing if bots can make that a thing of the past, it’s another whether those ranking highest on this list will be prepared to adopt and invest early enough. However, while modest budgets can be a challenge for the demands of digital innovation, the investment in the long term will mean huge cost savings. The ROI of digital automation investment now means it accommodates leaner resource for the future.


With such high expectations, what might prevent Chat Bots from delivering on their promise? While some people would prefer alternative channels (e.g. opting for a real person over AI (43%) or would stick to the normal website (32%)), the success of Chat Bots will mainly hinge on the quality or service and experience. Top of mind for consumers is ensuring that the AI is sufficiently advanced to understand the questions put to them (55%). No small feat considering the potential combinations of queries, colloquialisms and meanings that exist in everyday language. Also high on the list of concerns is reliability (41%), making sure Chat Bots follow through on instructions, rather than leaving you without the booking or purchase requested.

As an industry, developers must prove that Bots not only serve a useful function, but a quality one as well. Bots will become a new product offering and, like all products, they need to be tested, refined and adequately supported when taken to market. By inspiring confidence in Chat Bots, developers can ensure their continued adoption and success, trusted by brands, services and users alike.

Existing online customer services cause huge frustration, especially the lack of up-front, basic information (45% of people had experienced the problem in the last month) and poor (or no) options to ask simple questions (40%).

  • Utility companies provide the most frustrating online services, followed by local authorities and tradespeople.
  • Telephone calls no longer work as a back-up channel when frustrations arise. Almost one in five consumers (18%) actively avoid calling businesses. For under 24s, it’s a shocking 61%.
  • After learning a little about what they do, consumers think Chat Bots would significantly improve online services, providing better 24-hour service and quick answers to simple questions.
  • People would use Chat Bots to get speedy answers (46% people would use Chat Bots this way), and to track down the right human assistant for their needs (40%) — might this be the death knell for automated phone menus?
  • Overall Chat Bots were the second most preferred channel of communication, only beaten by face-to-face interaction. Chat Bots scored so well because of convenience and immediacy.
  • Age is not a barrier for Chat Bots, baby boomers (55–77) ranked their benefits even higher than Millennials (18–34).
  • Chat Bots were preferred over online chat, telephone, social media, apps, web forms or email for answering quick queries. Is this the digital silver bullet?
  • What could possibly go wrong? Well, the main barrier to Chat Bot domination is whether expectation can match experience. 55% said they would be put off if Chat Bots didn’t fully understand their questions.

This was a compelling proposition, if it played out. So, we decided to build a bot company.

At Supertext we then began to build conversational experiences for brands, businesses and enterprises. Using some crafty Artificial Intelligence tools we have built, we can ensure your business is supercharged with Chat, at a rapid pace. AI is eating the world. Is your business on-board the future bandwagon? “Is it clever enough?” You bet.

Seamless integration
Your organization is a melting pot for the future. This requires complex tools and technologies to be used. Supertext ensures that the intelligence of the Supertext platform is integrated right along you and your team’s existing workflow with minimal or no rework.

Brand and Chatbot personalization
We “get” brands. How they see themselves, how they want to be seen, how their audience sees them. By using audience insights and analysing emerging trends, we develop their tone of voice, visual language and how they speak to their customers.

Content marketing through Chatbots
We make content for humans, not robots. When brands provide content that’s funny, informative or valuable, people hold them in higher esteem. By integrating content strategies that are actually meaningful, we earn an audience’s approval to interact with them more. Tech and dev Interactivity and innovation. They exist at the heart of almost everything we do. And it’s our in-house development team. From data visualisation and microsite builds to interactive Chatbots and unique, innovative tech in the field of AI, we tackle it all.

Data driven
If something is measurable, we can optimise it. We gauge brand awareness to shape sequential storytelling into effective outreach and campaigns. We constantly refine and improve audience targeting, with multi-variant testing across platforms for your deployments.

K Startup Grand Challenge & Startup Bootcamp

Through the course of building the chat platform, Supertext was a part of the Startupbootcamp program in Singapore and the K Startup Grand Challenge in South Korea. These were transformational experiences in the amount they taught me on company building and the things we ought to be doing more and the things I could do better. To be part of a group of entrepreneurs from across the world had significant outtakes through the months spent in these programs

The JioMart Moment

The $850 billion Indian retail market is the fourth largest in the world and is largely unorganized.

As of 2003, India’s retailing industry was essentially owner manned small shops. In 2010, larger format convenience stores and supermarkets accounted for about 4 percent of the industry, and these were present only in large urban centers. There has been significant growth of e-commerce in India, trying to take advantage of economies of scale and selling all sorts of things online. The market penetration of e-commerce is supposed to be around 7% in 2021, compared to less than 3% in 2017 further accelerated by the risk of infection due to Covid-19.

Scenario

  1. JioMart is a portal from where consumers can buy groceries.
  2. Local Kirana shops will be selling through this application.
  3. Local Kirana shops will provide same-day delivery of goods.
  4. Consumers can ask for a Quotation by various shops, compare them, and further place orders according to preference. This will be done in WhatsApp.
  5. Payment can be made through WhatsApp Pay, JioMoney, or by using a credit/debit card.

Problems solved by JioMart:

  1. JioMart is expected to target the local Kirana shops so the idea is local Kirana shops will be able to sell their goods through JioMart, this will lead to upliftment of India in terms of technology, opportunities from the root level.
  2. Same day delivery- local shops will be able to provide faster delivery as compared to competitors such as BigBasket, Grofers, Amazon Pantry, etc.
  3. Consumers can ask for quotations from various sellers- can compare prices/delivery time/product quality/ other factors.
  4. Using WhatsApp pay as a payment method-WhatsApp has a user base of 250–300 million in India-people will be making transactions through WhatsApp as Pay option can be inbuilt in WhatsApp-no need to download/use any other application for payment method.

So here’s how you can place an order on JioMart

To place order through JioMart, customers are required to add JioMart’s WhatsApp number 88500 08000 to their contacts

— After adding the number, customers are required to send a message to the JioMart number. Type “Hi” and send to the number.

— JioMart will then reply with the shopping link along with a message that reads “Welcome to JioMart WhatsApp Order Booking Service”. The shopping link will be only active for 30 minutes. To generate a new link, a customer will have to send a fresh message.

— Once you open the link, you will be redirected to page that would ask for your personal details like mobile number, area, locality etc.

— After the details are entered and submitted, the new page with the grocery items listed vertically will appear. You can add the items to your cart.

— Once you generate the order, JioMart will send an invoice along with the grocery store’s address and location on Google Maps.

— Post which, the customer will recieve an SMS from the store when the order is ready. They can go and pick up the order from the store and pay there itself.



The Conclusion

Through this all, I often had to remind myself that I was still a full-time undergrad student finishing my Engineering degree. I spent most of my semester months across Singapore, Korea or the terrace offices of Supertext.

As I made my way to out of college, Supertext was a ramen profitable firm with clients mostly from Singapore for building enterprise grade Chatbots. We were on a path to profitability but ultimately it was not the rocketship for 10X growth that I wanted it to be. We had also strayed far from our beginnings as a hyperlocal aggregator to an enterprise chatbot builder. I decided to exit the company and informed my team about it, who handled it elan. Since then, I completed stints at business school but would argue that I am none the wiser.

Supertext was a journey. One that I had plenty of battle scars to show from. To this day, I believe it helped prepare me better for the other ventures and ideas I had tried since then, and for that, I am always in thankful for the learnings.

How about yet another
newsletter?

Your inbox has a few of these newsletters already. Why another one?
I write essays on Extended Reality, Entrepreneurship and eCommerce.
They're called battle cards, and might just come of tactical use when you're building the next big thing.