Chatbots have come a very long way in a short space of time. When they first started appearing on brand websites, they were more of a novelty and didn’t offer much in actual value. Jump forward to the present day and chatbots in all their forms are a vital tool for businesses to attract new customers, provide a better service to more people, and make significant cost savings. Crucially, with the development of no-code chatbot building tools – anyone can build and deploy one without needing a developer.
In this blog we will provide an overview of chatbot technology, describe all the available types of chatbots, and the basics of chatbot conversation design. Finally, we will showcase some real-world chatbot examples and how they are proving to be game changers for the businesses they serve.
What types of chatbot are available today?
If you were on the internet around 2005 your first interaction with a chatbot was most likely when one butted in on your browsing uninvited.
“Hey – it looks like you are browsing for car insurance. How can I help?”
These early chatbots were bolted onto business websites to generate a bit of a buzz. They didn’t add much value to be honest. Usually built in isolation by the IT department, they were rigidly scripted with the interaction breaking down if the user didn’t ask just the right question in the way that the bot had been designed to understand.
A bit like the old Windows paperclip guy – they were often more annoying than useful.
That has all changed with the maturation of chatbot design principles and the technology that underpins them. Designers came to realize that creating a chatbot that people want to use requires creativity, flair, and a real understanding of how humans think and communicate. As a result, chatbots these days have much better manners, are more sophisticated, and far more knowledgeable than their Web 1.0 ancestors. Crucially they can be completely integrated into an organization’s tech stack and can be deployed on multiple channels, including SMS, and messaging apps like WhatsApp, Messenger, and Viber.
As a result, they are now hugely popular with consumers who value their speed, accuracy, and 24/7 availability. Infobip’s 2022 Messaging Trends Report which analyzed over 153 billion interactions on our platform in 2022 found that a startling 99% of customer support and chatbot interactions took place on WhatsApp.
Irrespective of the channel, there are broadly speaking two main types of chatbot. Firstly, bots that guide the user to what they need using a set of options are called rule-based chatbots. You then have chatbots that are trained to provide more human-like conversation. These are called conversational or intent-based chatbots because they can be trained to recognize the intent of the person interacting with them and respond accordingly.
Both have their strengths and use cases that they are particularly suited to. Let’s take a closer look at both.
What are rule-based chatbots?
Rule-based chatbots can be considered as the digital equivalents of the interactive voice response (IVR) systems that are often deployed in call centers to direct customers to the correct department. We all have all been there.
“You have reached Acme Co. Your call is important to us. Please press 1 for accounts, 2 for customer service…. etc“
Rule-based bots operate on the same principle – although thankfully they are much quicker and offer a far more flexible and engaging experience. Under the hood they simply use a series of IF THEN statements to work out what action to take. So, if a particular condition is met when the person types a number or letter, then a specific action is triggered – for example relaying to the requested service branch or responding with a predefined message.
In this way, the chatbot can provide very specific and detailed information quickly and easily, with the added bonus that most live chat and messaging platforms offer features like lists and customizable quick-reply buttons so that users don’t have to type a response. With clever copywriting and attractive visuals. the whole experience can be made almost pleasurable.
The main benefit of rule-based chatbots is that they can be developed quickly and implemented easily. They are ideal for businesses that want to take a first step into the world of chatbots and start adding value immediately. They are perfect for all sorts of use cases including:
- Customer service: The vast majority of customer enquiries and calls to contact centers are to get answers to a handful of common questions – things like “What are your opening hours?” and “Where is my closest store?” or customer-specific queries like “What is my account balance?” or “Has my order been shipped yet?”. All of these can be easily handled by a rule-based chatbot.
- Sales support: Driving sales is becoming a highly successful use case for chatbots, especially for online businesses that operate 24/7. The chatbot can answer any product questions blocking the customer from completing checkout. It can also sweeten the interaction with rich media like videos and image carousels or send entire product brochures as attachments.
- Engagement: Chatbots are a great tool for engaging with customers and prospects at scale. Brands are using all sorts of clever strategies using chatbots to inform, educate, and inspire people. Games, quizzes, customer polls and interviews can all be facilitated via a rule-based chatbot.
- Two-way interactions: This is the latest area of chatbot progression that is facilitating some groundbreaking use cases. It is a way for people to share information about themselves via a private, secure, and always available service and receive targeted information back from the chatbot. From registering mobile phone SIM cards to medical applications, for example outpatients using a chatbot to submit blood pressure or heart rate readings to their medical practitioner. The key is that people now trust chatbots with their most private and sensitive details.
If you want to see just how easy it is to build a rule-based chatbot, check out our blog how to build a WhatsApp chatbot in 5 easy steps. If Viber is your preferred messaging app, then we have another blog that covers it.
What are conversational chatbots?
Rule-based chatbots become less useful when the person interacting with them doesn’t know exactly what they want. Rather that enquiring about something specific like store opening hours or the outstanding balance on their account, they perhaps have a problem that needs solving, but need to be able to explain what the issue is.
This is the domain of conversational or intent-based chatbots which use natural language processing (NLP) to work out what the person is trying to achieve, in other words what their intent is. This aims to replicate the normal conversational experience between two humans.
But how do they do this?
The answer is by meticulously designing every stage of the chat flow taking into account the goals and values of the chatbot, its personality, and its scope.
This process is called conversation design.
Conversational design – how to teach chatbots to speak human
Intents are the basis of every conversational chatbot. They inform which service it needs to activate, and which conversation flow to trigger at a particular point.
Say that a person wanted to check their bank account balance. The flow would look something like this:
- Person initiates a chat with their bank’s chatbot and requests “Give me my account balance”.
- The chatbot’s NLP engine ‘reads’ the text and after discarding non-relevant words such as ‘give’, ‘me’ and ‘my’ identifies that it needs to trigger an intent for “account balance”.
- This in turn triggers the pre-programmed flow for ‘account balance’, fetches it from the banking system, and formulates a reply – “Your account balance is: $XXX.00”.
All this depends on the chatbot successfully recognizing that the person wanted their account balance, and not say, their account details.
To achieve this the AI engine has to be trained to recognize every possible version of the same intent.
Best practice dictates that at least 50 different versions of each expected utterance (phrase) is required to be able to train recognition, but this number can be as high as 400 utterances for intents that are more ambiguous.
Consider the number of different intents that a chatbot needs to support in order to cover the required scope, and you start to realize that most businesses will need some help. Luckily, it is possible to crowd-source intents and there are providers that specialize in just that, providing you with all the training phrases required to build a conversational chatbot.
Testing a conversational chatbot
Once you have defined your chatbot flow and sourced all the phrases required for your intents, then you are ready to test. This is a crucial step that cannot be skipped.
No matter how diligent you have been during the design and build phase, there will always be scenarios that you haven’t accounted for and phrases that you haven’t covered. No chatbot is perfect, but testing will quickly surface any obvious issues that you need to fix.
A quick chatbot testing checklist might include:
- All supported entry points available
- Welcome and goodbye messages
- ‘Are you a bot’ questions covered
- Appropriate responses for compliments
- Appropriate responses for insults and abuse
- All intents and phrases required for chatbot scope
- Default intent and fallback options defined
- Consistent tone of voice
- Appropriate content and language for target persona
- End to end chatbot flow for chatbot goals are covered
- All data required for personalization is captured
- Consent obtained for retaining first-party data
- Concurrent interactions and scalability
To test a large number of intents, there are tools available that can help to identify those that are not distinctive enough, or which are incorrectly categorized. These can then be easily amended.
Testing of complex chatbots can also be simplified using specialized tools that simulate user behavior. Another option is to recruit a sample of your users to help you test. Testing with real people of different ages and from different backgrounds will help you identify flows that might need modifications or additional phrases and words that you need to cover.
Once you have tested every possible branch in your conversational chatbot flow, and you are satisfied with the experience it provides, you are ready to go live and start enjoying the benefits just like hundreds of Infobip customers.
Real chatbot examples
Here is just a small selection of chatbots built by Infobip customers using our chatbot building platform Answers.
- Nissan: The car manufacturer’s Saudi Arabian business saw a 138% increase in leads generated when they introduced a chatbot that provided 24/7 customer care and lead nurturing when dealerships were closed.
- Unilever: The brand experienced 14x higher sales with a clever chatbot campaign that promoted product awareness and generated interest.
- Flamingo: The financial services provider grew its conversion rate by 11% and NPS score by 21% after implementing a self-service chatbot to support its busy customer service operation.
- Carsome: The e-Commerce platform increased lead conversion by 10% with the help of a keyword chatbot that helped connect customers with their required service.
- Bolt: The ride-hailing company boosted the conversion of new drivers by 40% with a chatbot that guided drivers through the sign-up process, including an automated ID check.
- Unicef: The global children’s charity decreased donor churn by 33.3% with an omnichannel communication solution that incorporated a chatbot to serve as a failover channel for key communications.
You may also like:
- How to Engage Customers with Conversational SMS
- A Guide to Effective Conversational Marketing
- How to Design and Build a WhatsApp Chatbot
Messaging Trends 2022
Deep dive into the interactions that took place on our platform in 2022 and what they mean for the future of business-to-consumer communication.