The five different types of chatbot - and which to choose for your business
What are the true benefits of developing a chatbot for your business? Nastasya Savina of Just AI explains.
265 billion customer support requests are made every year, and it costs businesses $1.3 trillion to service them.
As a result, huge pressure is being placed on a business's finances and its customer service agents - yet up to 80% of support requests can be served without human involvement.
Content seriesView full content series
Through self-service and automation, it's said that you can reduce customer service costs by 30%, potentially freeing up support staff to deal with resolving complex queries.
Little wonder there is so much hype around chatbots.
The chatbot market size is estimated to hit $1.25 billion by 2025, according to Grand View Research. 57% of UK consumers know what a chatbot is and crucially - 35% say they want to see more companies using them.
The road to chatbot success is far from straight-forward, however. Starting your first chatbot project with an ambition to create a perfect AI-enabled tool might be an unfeasible task for any business.
Far better to begin with a simple solution and evolve from there onwards - and there are five different types of chatbot you may wish to consider.
1. Button-based bots
The purpose of a button-based bot is to lead a user through a pre-defined scenario tree. This type of bot resembles a text-based IVR system. Just like an IVR, a button-based bot offers choices and asks for data. Usually, it says: "I'm a bot that can help you with the following issues, please press the relevant button: X, Y, Z..."
Creation of this type of bot does not require any AI capabilities from the development platform and takes minimal time and resources (e.g. one to a few days for the simple scenario). Usually, such bots are comparatively easy to create, require minimal professional competence and can be designed in some intuitive graphical bot builder like Aimylogic or other. This bot can be published in any of the normal channels - as a web widget, in messenger apps, Alexa or Google assistant.
A button-based bot is perfect for onboarding, surveys, sales support and practically for any simple process automation task where the communication scenarios are clearly defined.
However, after a few weeks or months of operation, a company might want to cover more scenarios, so that the simple button-based bot would not be enough. Then the next step on the bot evolution continuum is a hybrid bot.
2. Hybrid bot
A hybrid bot is a button-based bot with an option to ask a question on natural language. Usually, it says: "I'm a bot that can help you with the following issues; please press the relevant button X, Y, Z or type your question in the field below".
Industry insightsView more
Frequently, the NLU ( natural language understanding) in these bots duplicates the buttons. A hybrid bot is a transitional stage between the button-based and an AI/ NLU bot.
A hybrid bot is suitable for the following tasks:
- Like a button-based bot, it helps users to find answers to the most frequent/common questions.
- It registers the subjects not covered in the initial bot scenario: what issues concern users, and what words are used to describe them. It is a form of customer research, helping to understand, e.g. what questions should be included in the FAQ or depicted on the website, and what words should be used for that.
- If the company does not have a conversation history logs but wishes to create an AI bot in the future, then a Hybrid bot might be one of the acceptable ways to collect context data over time for a bot training. However, it would be better to create logs from scratch by recording an operator, talking with a visitor.
A bot can exist in a hybrid stage for quite a long time, while developers are enriching its AI engine with a better understanding of relevant context.
The cost of initial creation is not much higher than for the button-based one, however, this type of bot usually requires a more advanced development platform with strong AI/NLU core. Otherwise, processing of queries on the natural language would be next to impossible. Moreover, the ongoing costs of the project are higher because a hybrid bot requires constant update and training to progress to a fully-functional AI solution.
3. AI bot
A company should consider implementing an AI bot when one or more conditions are met:
- If the company receives a few hundred or more queries per day.
- The company’s tech support or sales support staff exceeds 20 people.
- The company wishes to support more communication channels (messengers, web widgets etc.) without adding support staff.
- Most of the conversations are purposeful and structured.
- There are peak periods in communications when many customers reach out to the company simultaneously and are very unhappy about the response delays.
- When a large proportion of customers needs help and support outside working hours.
An AI bot can automate complex simultaneous communication with multiple users, covering many loosely connected topics. For example, a telecom operator chatbot can consult on the tariffs and upgrade options, at the same time report on the current client’s usage stats. A retail sales support bot consults on various product categories, terms of purchase and delivery. A tech support bot conducts initial diagnostic and offers help.
Simple and rigid button-based bot scenarios can not automate all these complex tasks – one really needs the flexibility of the natural language to capture a wide variety of users’ intents, collect necessary data and provide support.
The AI bot project usually requires a noticeable upfront investment for the creation of a bot (usually, more than £7,000-£20,000). In addition to this, a company should invest in the development platform subscription and ongoing support of the bot improvement.
An AI bot project should be managed and implemented by a professional team with strong conversational AI competence.
Despite the more noticeable investment, an implementation of an AI bot brings material benefits, including:
- Cutting total cost of customer support
- Improvement of the quality, timeliness, speed of customer communications
- Improved motivation of support staff, that doesn’t have to deal with repetitive and mundane queries.
One of the most famous bots - Julie from Amtrack – helps users to find a train and book tickets, conducting a human-like conversation. Such bots are appreciated by users too: Amtrak declares that Julie’s launch has led to increase of booking by 25% and $1 million saved on customer service.
The AI bot project usually requires a noticeable upfront investment for the creation of a bot - usually, more than £7,000-£20,000
4. An omnichannel ecosystem, including AI bot
Even when the ambitious goal of creating an AI bot is achieved, it is too early to get too comfortable. Those that cannot live without a challenge embark on the next project of creating an omnichannel communication experience for their customers. In omnichannel paradigm a conversation, started, for example, in Twitter, can continue with the chatbot and be followed by a phone conversation with a tech support agent and finalized over email.
Achieving a smooth work of all systems, clear and complete transfer of data from and across the channels is another level of complexity, which requires a very strong competence and significant resources. Instead of being a standalone solution, an AI bot becomes an integral part of a diversified communicational ecosystem.
A lot of customers get used to communicating with businesses via multiple channels – social media, phone, email, web chatbot. And when the switch between the channels happens within one conversation, they expect the whole history to be available in the next channel. Otherwise they get annoyed introducing themselves and explaining an issue again and again.
5. Voice – the future of bots
Global smart speaker ownership will exceed 200 million by the end of 2019 from 114 million with the UK growing 46%. This explosive growth creates a new communication channel, connecting businesses and customers. To use this channel the most advanced companies (usually B2C) create so-called “skills” (voice-based bots for Alexa or Google Assistant).
In recent years the speech and linguistic technologies progressed to the point when the creation of well-functioning complex voice solutions have become feasible, for example, voice assistants, smart IVRs or robocalls.
By utilising the most natural interface - voice - these solutions can appeal to users and bring a lot of benefits to companies. However, the development of voice-based solutions requires specific domain expertise from an architect, a linguist, a UX designer, a developer, and a range of other professions. To work effectively, each member of the team needs dedicated tools. There are not too many voice solution development platforms, that support the whole development cycle, for example, DialogFlow from Google, SAP Conversational AI, JAICP from Just AI.
Many market experts believe that in the near future the voice solutions would become a common and necessary part of the whole IT ecosystem - just like mobile has become in recent years. Therefore, forward-looking companies should start the first experiments with voice now, to build experience and test hypotheses.
A team ethos
Chatbots, as with any properly designed automation project, can bring tremendous benefits to a business – improving speed and quality of service, saving costs, increasing revenue and strengthening the brand.
Conducting the first experiments with chatbots does not require a huge budget, time or resources – a company can “test the waters” with a tiny button-based chatbot cheaply and easily. Once the initial hypotheses are tested, the company can evolve this tool further, incrementally adding resources, supporting new use cases, developing capabilities.
A corporate chatbot is not a single rigid solution, developed once and staying unchanged. Like a corporate website or a mobile app – it is rather a journey than a destination. And like in a real journey - the success of the project largely depends on your partners, their competence, power of their technologies and industry knowledge.