Chatbot Research

Part
01
of two
Part
01

Chatbots Deep Dive: Part 1

A summary of the most engaging chatbots is best broken down on an industry-by-industry basis. Some of the most engaging and innovative chatbots include those used by the financial, human services, and fashion and style industries. We will outline three case studies of chatbots in each industry, with examples of what makes them so engaging and further detail.

There are approximately 10-20 non-coded chatbot platforms on the market; although no definitive figure is available, we are able to confidently surmise this ballpark figure.

There are implications for chatbots as a result of the Facebook privacy scandal, with major chatbot platform Chatfuel announcing their apps will not reconnect to Facebook pages, and Facebook announcing they will not allow new chatbots on their platform.

EXAMPLES OF ENGAGING CHATBOTS

FINANCIAL

This chatbot was developed by Capital One Financial and currently used by the company is called Eno. This chatbot is leveraged by artificial intelligence, and communicates with Capital One Financial customers via text messages. Although it is currently only available for a small pilot of customers, it is one of the most engaging chatbots on the market. Eno provides its users with updates, allows them access to information immediately, and can even pay their bills upon request.

Eno has access to customers' account information, to give them information on their balances, update them on their recent transactions, allow them to see their credit card due dates, and credit card limit. Unlike other many other chatbots on the market, Eno only uses text messages, and does not require the user to install another application. This makes it particularly useful for older customers, who may not use a smartphone, and other non-smartphone users.

HUMAN SERVICES

Woebot is a chatbot aimed at recognizing the signs of depression and offering a bespoke therapy service to its users. Woebot is used primarily through Facebook Messenger, and is an artificially intelligent chatbot. It is trained in cognitive-behavioral therapy (CBT), which is one of the most widely known methods of treating depression.

Woebot tracks a user's mood through the use of keywords, and is able to show users how their moods change. The chatbot can do this whilst maintaining privacy and finding patterns that might be more difficult for an average user of the chatbot to analyze.

Woebot is then able to talk about the user's mental health and wellness from brief daily conversations. It asks questions such as what’s going on in the user’s life, and how they are feeling that day. Depending on the results of these questions, the chatbot is then able to send useful videos and other therapy tools depending on the user's mood and needs at the moment. It is aimed at promoting better mental health in users who may be uncomfortable in discussing these issues with another person.

FASHION AND STYLE SERVICES

The Dutch-based fashion retailer H&M currently use a chatbot that is able to market clothing styles directly to its users. The chatbot used by H&M can be linked to a user's instant messaging account on apps such as Kik. It then analyzes the potential customer's photos currently stored on their device or in cloud storage, and can pick an outfit for a user based upon these photos.

Users are able to update the chatbot with styles and preferences for the chatbot to more closely pick out appropriate outfit for the user based on their tastes.

This is an innovative method of using a chatbot to directly market multiple products to a customer and potentially generate sales.

CHATBOT PLATFORMS AND THEIR CHARACTERISTICS

There are chatbot platforms available for clients who have the capacity to develop and write code, and those who do not. We will make the distinction between non-coding and coding chatbot platforms.

NON-CODING

The most widely used non-coding chatbot platforms include Chatfuel, Botsify, Flow Xo, and KITT.AI.

Non-coding chatbots allow a user or company to easily build a chatbot without necessarily being an advanced computer user, or have any experience of using code.

Chatfuel is an extremely popular chatbot platform used across instant messaging applications such as Telegram and Facebook Messenger. It is estimated that over 46,000 chatbots have been created using the platform, with clients including Adidas, MTV, British Airways and Volkswagen. Chatfuel is free with basic features, or Chatfuel Pro can be purchased for $30 USD per month. Chatfuel is simple enough for beginners or small businesses to construct a chatbot of their own.

Botsify is another widely used chatbot platform, used across instant messaging applications such as Facebook Messenger, WhatsApp, and Instagram. It is estimated that over 40,000 chatbots have been created chatbots using the platform, with clients including Apple, Shazam, Grin, and Travelex. Botsify is free, or price plans start at $10 USD per month for a basic package, up to $50 USD per month for a business package. Botsify uses templates that the client can customize, with 20 templates in the free version, and more available with each price plan.

CODING

The most widely used coding chatbot platforms are IBM Watson, Microsoft Bot Framework, Wit.ai, Semantic Machines, Reply.ai and ManyChat.

These chatbots tend to be more complex, using artificial intelligence in order to handle more complex conversations with users.
IBM Watson is a popular coded chatbot platform, used across instant messaging applications such as Facebook Messenger, WhatsApp, and Instagram. The estimated number of created chatbots using the platform is not available, but clients include Staples, H&R Block, and Autodesk. A chatbot coded on the IBM Watson platform "can serve different verticals including customer engagement, education, financial services" and more.

Microsoft Bot Framework is another popular coded chatbot platform, used across a wide variety of applications such as Microsoft's website and apps, and Skype, Slack and Facebook Messenger. The estimated number of created chatbots using the platform is not available, but clients include UPS. A chatbot coded on with Microsoft Bot Framework "is globally accessible" through a cloud-based service "across 141 countries, and bots can communicate in multiple languages".

APPROXIMATE NUMBER OF NON-CODING CHATBOT PLATFORMS

As a ballpark figure, there are between 10 and 20 non-coding chatbot platforms currently available. Although we were unable to find an exact figure, across numerous sources we were able to total the amount of non-coding chatbot platforms mentioned to arrive at this ballpark figure.

IMPLICATIONS OF THE RECENT FACEBOOK SCANDAL

We can conclude that the recent Facebook scandal will have ongoing implications for chatbots, as the scandal has already affected app developers.

Chatbot developers such as the bot-building service Chatfuel have already announced their intention to change their policies and working practices in the wake of Facebook's privacy scandal. Chatfuel announced in a private email that it will not reconnect chatbots to Facebook pages while Facebook pauses for an app review.

As mentioned above, Chatfuel-based chatbots number some 46,000 apps, so this announcement already has wide-reaching implications across the industry.

Facebook has also announced that it is currently not allowing new chatbots to be added to their platform, so the implications are wide-ranging.

CONCLUSION

The most engaging and successful chatbots on the market are available across many industries, with the examples cited here financial, human services and fashion. These specific examples include Capital One Financial's Eno, Woebot, and the chatbot used by H&M.

We can conclude that there are approximately 10-20 non-coded chatbot platforms on the market, and chatbot platforms are either coded or non-coded versions, depending on a client's wishes, budget and usage.

As a result of the Facebook scandal, the chatbot platform Chatfuel announced that it will not reconnect chatbots to Facebook pages while Facebook reviews their apps. Facebook also announced it will not allow new chatbots to connect to their platform.
Part
02
of two
Part
02

Chatbots Deep Dive: Part 2

Recent technology advancements in artificial intelligence (AI) have allowed chatbots to become more sophisticated and useful in their roles. Chatbots can now be configured to assist customers 24-7 entertain readers, and market the right products at the right time. Examples of companies that build custom chatbot solutions for businesses are Contus, BotsCrew, Root Info Solutions, Unibot, QBurst, OCDLAB, Ideas2IT, Engati, Peerbits, and Aivo. The average cost to build a custom chatbot ranges from $6,000 to $12,240. More details on 10 chatbot-building companies, the chatbot building process, and the typical timeline and costs for building a chatbot are outlined in the sections below.

COMPANIES THAT BUILD CHATBOTS

1. Contus (http://www.contus.com/chatbot-solution.php) — Contus Bot builds chatbots with pre-built templates, multilingual capabilities, multiple language conversation ability and natural language interpretation. 2. BotsCrew (http://botscrew.com/) — This company develops chatbots that include voice assistants and work on any platform including FB Messenger, Slack, SMS, Cisco Spark, Kik, and Skype. 3. Root Info Solutions (https://www.rootinfosol.com/chatbot-development-company) — Root Info Solutions develops chatbots for many platforms including SMS, Web, APIs, IBM Watson, Microsoft Bot Framework Amazon Alexa, and Apple’s Siri. 4. Unibot (https://unibot.ai/) — This developer creates chatbots for instant messaging that automate interaction in a variety of business sectors including online banking, payment systems, ticketing systems, web portals, gaming, and government programs. 5. QBurst (https://www.qburst.com/solutions/chatbots/) — Qburst creates self-learning chatbots that intelligently receive feedback, complete tasks, and answer complaints so that small businesses can create high-value interactions with customers. A specialty of this company is creating chatbots with “moods” that can express emotional responses. 6. OCDLab (https://ocdlab.co/bot-developers/) — This company creates chatbots suitable for various fields including news, customer support, entertainment, e-commerce or corporate internal communication. Their bots integrate with enterprise solutions and various public platforms. 7. Ideas2IT (https://www.ideas2it.com/chatbot-development-services/) — This company crafts chatbots for multiple industries such as ecommerce, healthcare, retail, internal company HR, conversational banking services, sales, and marketing. 8. Engati (http://www.engati.com/) — Engati is a platform for chatbot building that allows its business users to build, analyze, and create their own customized bots in just a few minutes. Users can create chatbot workflows with no programming. 9. PeerBits (https://www.peerbits.com/chatbot-development.html) — Peerbits is an international app development company that offers chatbot development designed to simplify user tasks. 10. Aivo (https://aivo.co/en/) — Aivo develops chatbot solutions with the ability to automatically interact with customers across multiple channels. Their focus is on helping larger companies improve customer experiences by gathering insights from individual interactions.

HOW CHATBOTS ARE BUILT

According to Entrepreneur, successful chatbots tend to act very human in order to help users, and offer multiple benefits to companies such as custom leads, personalized ads, social media monetization, saved time, and increased sales from 24-7 support. Several steps are involved in the process of creating a chatbot from scratch.

Step 1. Determine the bot's purpose. The goal of a chatbot is to offer a service that users will want to use over and over again. No single chatbot can address all of a company’s needs. Therefore, it is important for companies to decide which specific problem they want their chatbot to address efficiently. Step 2: Decide which messaging app the chatbot will exist on. Some chatbots that live on popular platforms such as Facebook Messenger or communication apps like Slack. However, there are many other alternatives such Kik, Viber, and Skype. Step 3: Choose a developer platform to build the chatbot with.
There are many app tools for developers to build chatbots. Step 4: Craft your chatbot’s brand personality A chatbot’s communication personality should match its brand. For example, a financial company’s chatbot may have a serious tone, while an entertainment company may have a more upbeat tone. Step 5: Design the chatbot’s conversation flow Every chatbot building platform offers various options to build meaningful conversations. Several of these platforms even provide drag-and-drop interfaces, multiple choice answers, or if/then statements to build out the conversation logic and style. Step 6: Connect the newly designed chatbot to its messaging app
At this step, the chatbot is finished. It is then the time to connect the app to its working environment to facilitate live testing. Step 7: Complete beta testing with a group of users Because chatbots are highly interactive and customer-facing, it is important that conversation flows be tested for multiple use-cases with different people. After the chatbot has been totally de-bugged, it can be released and promoted to the public.

CHATBOT DEVELOPMENT COSTS AND TIMELINE

The complexity of a particular chatbot project and the number of teams it is designed to serve affects the overall cost and timeline. According to software development company RubyGarage, the average cost of creating a chatbot from scratch ranges from $6,000 to $12,240. The timeline involved with having a company build a bot for the client is listed below. 1. Creating the chat integration (40-56 hours of development) 2. Development of the communication interface (160-216 hours total of development). This step involves two subparts: • Creation of the language user interpreter (40-56 hours) • Natural language user interface development (120-160 hours) 3. Business logic development and connection with relevant APIs and mobile apps or from scratch (120-192 hours)

CONCLUSION

Many companies offer custom chatbot building services including Contus, BotsCrew, Root Info Solutions, Unibot, QBurst, OCDLAB, Ideas2IT, Engati, Peerbits, and Aivo. The process of building a chatbot involves determining its purpose, choosing a platform, building a custom conversation flow, and completing extensive testing. Overall it takes software developers anywhere from 480-680 hours of project work to create a custom chatbot.
Sources
Sources