What Is an AI Chatbot and Does Your Business Really Need One?11 min read04/12/2018
Chatbots have been one of the most popular topics among marketing specialists for the last few years. Many Ecommerce companies including Amazon and eBay have created their own virtual assistants based on artificial intelligence (AI). Chatbots have become a common trend followed by numerous brands in different industries. The question is whether these assistants are that useful as their developers claim.
What are chatbots?
Chatbots are software tools that simulate human communication via text messages or voice. They allow businesses to automate various simple tasks by working according to a predefined algorithm. A typical chatbot can engage users in a dialogue by responding to their requests or helping them solve their problems.
Since the first chatterbot Eliza capable of passing the Turing Test and successfully imitate the way people talk, virtual assistants have significantly changed. With the growing popularity of instant messengers in the 2010s, chatbots gained a new life. Mostly based on messenger platforms like Facebook Messenger, Telegram, Viber, Skype, and Slack, bots can either work as separate mobile applications or be a part of the website environment.
How chatbots work
Chatbots can be divided into simple and advanced or self-learning. They differ by approaches used in order to recognize human requests and tasks they’re aimed at.
Keyword-based chatbots process user commands by recognizing keywords in the user message that then trigger corresponding responds or actions. These bots rely on a list of keywords in their database and suitable answers. In other words, keyword-based chatterbots use a simple “If-Then” algorithm. They can’t recognize a human language.
- Limited in capabilities
- Hardly able to imitate the human conversation
Chatbots based on artificial intelligence use natural language processing (NLP) to “understand” what users ask and what they need. As a part of AI, neural networks enable developers to build complex virtual assistants that rely on advanced algorithms. These algorithms allow chatterbots to learn from each interaction with users and better understand their requests over time. Apple’s Siri and Amazon Alexa voice assistants use the same principle. Furthermore, AI-based chatbots can imitate a specific character or simulate such human feelings as empathy.
- AI-based bots can understand human language
- Become smarter with each user interaction
- Behave naturally within their scope.
The main goal of chatbots is to enable users to send commands to a computer in the natural communicative way instead of using visual interfaces along with buttons or code-based commands such as SQL requests. In other words, the information you used to get by switching between different web pages, can be requested from a chatbot within one chat. Below, we will consider common chatbot use cases.
When you choose a hotel, you need to manually configure a wide range of options like a departure date, price, location, etc. With Booking.com’s AI-based chatbot, you can book an apartment by talking to a bot and answering its questions. The Booking Assistant can respond to common questions related to the client’s current reservations. If the bot can’t help the customer, it redirects the issue to a suitable employee. This solution helps the company deal with 30% of incoming queries in an automated way.
In 2016, The Washington Post launched their own chatbot for Facebook Messenger. The simple keyword-based virtual assistant can deliver top stories and the latest news on the topic set by a user. For example, to get the top news on the automotive industry, you should type in “automotive” in the chat interface. In addition, the bot can provide election results in a specific city if a user enters its zip code.
The Fintech industry also started turning towards the conversational commerce. In March 2018, Bank of America presented its AI-based bot. The virtual assistant called Erica can provide users with their bank account balance, schedule a payment, list recent transactions, and even give some tips regarding user financial behavior. In addition to text messages, Erica can understand voice commands.
eBay first offered a chatbot as a means of the virtual bidding assistant. The solution used to help customers to win auctions with as low bids as possible. It could send a reminder to a user via Facebook Messenger 15 minutes before the end of the auction thus enabling the auction participant to make the last-minute bid.
Then the company expanded the functionality of this bit and turned it into a full-fledged virtual personal shopping assistant that could help customers find products they needed within the required price range. The AI-based ShopBot can ask relevant questions to offer the most convenient items. Since its launch, eBay customers started communicating with the company more often by 300% thus increasing the retention and engagement rate.
Enterprise chatbots help companies organize their work with databases and separate services in a more efficient way. An AI-based chatbot called Meekan offers an alternative for meeting scheduling web apps. It allows managers to assign meetings in the conversational form. Instead of comparing schedules on your own, you can ask the bot to recommend the best time. Meekan will synchronize with Google Calendar, iCloud, and Office 365 and check the availability of each specialist. Then the chatbot will recommend the best suitable time for a meeting for all participants and inform them via Slack. NLP enables the chatbot to understand human-readable requests.
To connect bots based on messaging platforms, you should simply add its name to your contact list as you do it with your friends or colleagues. If need a specific type of bots, you can find a suitable one in bot directories.
- Botlist. It’s the largest registry of virtual assistants that includes both voice- and chat-based bots. The directory provides filtering options such as category, platform, and rating that will help you find a suitable solution faster.
- Chatbottle. This marketplace provides tens of thousands chatbots for different platforms like Skype, Slack, Telegram, and Kik. However, the directory has poor filtering that’s why it’s quite hard to find what you need among over 10,000 chatbots for Facebook Messenger.
- Storebot. Storebot has the catalog of both bots based on messaging platforms and installable mobile apps. You can find a suitable virtual assistant in corresponding categories or top chart, best new, and recently added collections. Furthermore, the directory has its own Slack-based chatbot that can help you navigate through lists of other bots.
Pros and cons
Let’s cover the advantages and disadvantages of the chatbot development.
AI chatbots have a wide range of benefits that allow companies in different industries like insurance and retail save their time and focus on complex tasks while entrusting routine issues to virtual assistants.
Anyone has ever faced long minutes or even hours staying on the line to wait for an answer from the customer support department. When the system finally connects you to a company’s representative, he or she often answers with huge delays since employees need to manage multiple chats at the same time or they just need some minutes to find the information you need. AI chatbots can understand your requests and respond to them in a few seconds. This eliminates the need for wasting your time in the queue.
Bots are much faster than humans. While an employee can serve only one customer at the same, one chatbot can communicate with multiple users and generate instant answers to routine questions. Therefore, one virtual assistant can replace several employees thus allowing businesses to save their budget on paychecks.
In the U.S., working days have the limitation of 8 hours, but customer inquiries can come even at night and during weekends in the case of emergency. Bots need no rest. They can process customer requests in the 24/7 mode. This can help companies improve the quality of their customer support and increase client loyalty.
Bots are trendy
The popularity of virtual assistants is continuously growing. Statista reports that the global chatbot market share will reach the value of nearly $1.3 billion by 2025 compared to only $190.8 million in 2016. Another reason why companies create their bots is that this technology is a current marketing trend. A chatbot launch sounds like a good idea for starting a new PR campaign.
Bots are as smart as their creators are
Regardless of all advantages of using chatbots, they still have a few important drawbacks emerging from brands’ attempts to replace a live person-to-person conversation with automatic replies from bots, even though from smart bots enabled with natural language processing technology.
Bots are not humans
Natural language processing allows developers to create bots that imitate a human conversation. However, AI-enabled chatbots can’t imitate human feelings and express sorrow or empathy. A customer who approaches the company with a problem realizes that a robot behind a display doesn’t care even if it sends automatic emphatic replies. Bots are good at providing information but awful at pretending to be a human.
Neural networks can allow virtual assistants to learn on each interaction with real users. The more often a bot talks to humans the faster it becomes smarter. However, bots are as smart as their creators are. These are developers who define how a certain chatbot should learn. The thing is that this learning process should be under control. Otherwise, “too smart” bots can violate the idea they have been created for. In this case, a virtual assistant can respond inadequately to user requests and exceed the supposed conversational logic. This can lead to significant reputational losses
Since bots like The Washington Post’s one can be used for content delivery purposes, this forms a risk of becoming annoying with too frequent notifications. Chatbot developers should wisely manage how often their solution will deliver news, links to posts, or promotional content to users. Otherwise, tired of spam, users may simply remove this bot from their device or contact list in the messaging app.
If you don’t know how to build a chatbot, you should remember that you always have to ways: to create it from scratch by hiring a software development team or use one of the special platforms that greatly simplify the overall development process. With a dedicated team, you can develop an advanced solution enabled with artificial intelligence. However, the AI chatbot development will cost you on average from $5,000 to $10,000. Using chatbot platforms, you can build a simple bot for $10-$20 per month without any programming.
- Chatfuel. It’s one the easiest chatbot development platforms that allow you to create a bot for Facebook Messenger in only 7 minutes by linking it to a Facebook page and gradually filling in blocks with questions and answers.
- Manybot. You can use the Manybot platform to create a simple chatbot for Telegram for free in a drag-and-drop way.
- Dialogflow. This platform allows you to develop a chatbot enabled with the NLP technology. With a Premium plan, you can even get a bot with the speech recognition feature.
- Meya. Using the Meya chatbot platform, you can build a bot capable of synchronizing with CRM systems for a wide range of messengers including Twitter, Slack, Kik, and Twilio.
The chatbots vs. mobile apps comparison
AI chatbots were supposed to replace mobile apps due their advantages:
- Capability to function in one chart interface. No need to switch between different screens in the menu.
- Chatbots need less traffic than mobile apps and can work even at a low Internet speed.
- Bots can work within a messenger platform. Developers may not need to create native mobile apps for Android or iOS. As a result, users don’t need to download and install them thus reducing a free storage space of their mobile devices.
- Bots are much cheaper to develop than mobile apps.
Will chatbots replace mobile apps?
Instead of the aim to replace mobile apps, chatbots should have been considered as separate tools with other tasks. The initially incorrect idea may have formed an obstacle for further improvements of bots. Are they better than apps? Indeed, they have a set of advantages mentioned above, but this doesn’t make chatbots better than mobile apps.
They aren’t faster than apps as for now. Virtual assistants need a couple of seconds to process your request. Furthermore, to book a hotel with a chatbot, you need to answer several questions until it retrieves all necessary data to accomplish your task. You can do the same in a few clicks using a mobile app.
Bots aren’t easier to use than apps since the latter ones aren’t difficult either. What really can complicate the booking process is trying to reformulate your request to make a bot understand you. Even with natural language processing, bots may misunderstand your inquiry and generate a useless answer.
Keyword-based chatbots make it easy to create a list of scenarios they’re aimed at. However, everything that exceeds this list makes bots useless and make users disappointed. When it comes to mobile apps, they limit a set of available scenarios with their functionality and interface, unlike bots that allow a nearly unlimited number of scenarios in a form of replicas these assistants aren’t ready for.
Until NLP can understand 99% of user requests, chatbots won’t be better than mobile apps. Until companies develop bots that really make sense, they won’t be able to benefit from this technology. As for now. Brands shouldn’t try to fully replace their customer support department and mobile app with a virtual assistant. Instead, they may build a chatbot that will improve user experience and automatically connect employees when a user needs human assistance.