Introduction to Chatbots
Chatbots are making interactions between customers and businesses more streamlined, more informative and more productive for all parties. Whether it’s an automated Facebook Messenger bot to answer your queries, a chatbot on a dedicated mobile app to deal with customer support issues, or simply a tool to help business owners reduce repetitive work for their human staff, chatbots are problem-solvers.
To help you understand what chatbot could streamline your processes and drive new customers to your business, this guide will walk you through everything you need to know about chatbots.
What is a chatbot?
In its most basic, stripped-back form, a chatbot is artificial intelligence (AI)-powered software that is designed to simulate conversations with users – typically potential customers – in natural language. The goal is for the chatbot to mimic an interaction with a human, and it is most often deployed on websites, mobile apps, messaging platforms and even over the phone as a voice chatbot (for more information, see ‘What’s the difference between a voice chatbot and a text chatbot?’).
Chatbots are a modern evolution of the traditional question answering systems, and their ability to use natural language processing (NLP) makes them stand out from clunky, robotic response systems that are often unable to discern meaning from human-generated questions.
How do chatbots work?
It’s important to clarify from the outset that chatbots are not a set-and-forget solution. Businesses looking to adopt chatbots in their systems must understand that there are different types of chatbots for different types of needs. That means it’s important to conduct your own due diligence to find the chatbot that matches your requirements.
In terms of how chatbots actually work, we can categorise them into three different pillars: rule-based, learning (AI) and live chat.
As the most widespread chatbot online, rules-based chatbots are essentially a repository of set responses and specific information, in which the customer is guided through the question-and-answer process according to how they respond. The interaction begins with an initial greeting from the chatbot, the customer then clicks a question or response from a set list. There may be further qualifying questions as they go along, and then the chatbot will give them their ‘answer’ – this may be the information they need, a link to further information, a request for the customer’s contact details so a human support person can reach out to them, or a variety of other ‘solutions’.
Learning (AI) chatbots
Learning chatbots leverage the power of AI to actually become smarter over time by learning as they go. The chatbot developers deliberately create these bots to ‘learn’ from humans with each new interaction. When a customer interacts with a chatbot, the bot analyses the request or question and then determines the intent and entities. The chatbot then replies to the customer and that information is held in its systems for future reference – and faster, more accurate and more detailed replies next time.
An example of a learning (AI) chatbot is Question Bot, aka QBot. QBot is a world-first, AI-infused chatbot that can be integrated into Microsoft Teams or another platform. Focused on the education sector, the bot is able to answer student questions. It learns from every interaction and builds a knowledge base that can be used to answer questions in the future. A learning chatbot that utilises AI is essential for remote work and remote learning, and promoting collaboration between employees and/or students. In fact, having an overall streamlined unified communications strategy to accompany your chatbot is also essential.
Live chat chatbots
Contrary to its name, a chatbot can – and in many cases should – be used in tandem with human support. For a live-chat chatbot, a customer will ask a question or make a request which will cause the chatbot to ping the human support team. Whether it’s customer support or a salesperson, the human can then see the customer query and make a direct reply. You may have interacted with this type of chatbot on popular social media platforms like a Facebook business page.
Why a chatbot?
No matter the type of organisation or the industry in which it operates, there is incredible value in using chatbots for improving your business. Think about all the ways you want to reach out to consumers, clients or stakeholders – chatbots can potentially help you streamline those conversations and improve your bottom line.
But perhaps even more valuable is how specific internal use cases can improve productivity in your team – from chatbots streamlining your HR support, to them training up your new employees, to bots simply taking care of repetitive day-to-day tasks. This lower barrier to entry means you can ‘try before you buy’ and see if there is actual value to chatbots for your organisation. One such innovative solution is askHRplus. askHRplus is the answer to all your HR needs: it combines three solutions to help you drive employee engagement.
And it is here that the biggest drawcard for business owners lies: chatbots can reduce inefficiencies and increase productivity in customer interactions, which ultimately improves the customer experience and may lead to a more streamlined sales funnel. This is because they cut down on the need for humans to respond to identical, repetitive questions from different customers.
Taking the software a step further, chatbots can offer businesses clearer insight into the customer-engagement process so they can spot – and plug – gaps in their operational efficiency, and therefore reduce the average cost of their customer service.
To maximise the power of chatbots, however, they must be deployed as a support tool for your human workforce. No matter how intelligent the solution, a chatbot needs humans to configure, train and optimise its system – not to mention there will always be instances where customers want to speak to an actual person rather than a dedicated chatbot.
What is the best chatbot?
Like all forms of technology, it is difficult to quantify what is the ‘best’ solution because it depends on your individual needs and the problems you are trying to solve.
For many business owners, the best chatbot will be the one that is easy to integrate into their systems and creates the fewest hurdles in their day-to-day operations. However, this may be seen as a controversial stance to take, especially for technology lovers who keenly follow the latest developments in AI and machine learning.
That is why this statement requires some clarification: high-quality and constantly evolving solutions are integral to the technology sector. It’s what keeps the competition high and forces developers to bring newer, more powerful solutions to market. However, it’s just as important to recognise that many chatbot projects fail due to issues that are not attributable to the technology itself, but instead to the organisation deploying it.
When you have made the decision to use chatbots, we recommended asking the tried-and-tested ‘why’. Why do you want adopt chatbots? Is it to improve the customer experience, reduce your staff overheads, or create a more streamlined sales funnel? Once you have answered that question, you can start looking into the different types of chatbots on the market and determine which solution will best match your needs. (See the next section ‘What are some examples of useful chatbots?’ for a more detailed look at some of the best chatbots currently being used).
Whitebox vs blackbox chatbots
For non-technical business users in particular, whitebox provides a critical level of transparency in the AI decision-making process. This includes clarity around how the AI arrives at a certain result or decision, making it easily understandable for everyone from product managers and customer service staff to legal and compliance departments. This feature of whitebox chatbots is critical for businesses that are in highly regulated environments, such as government. The other major benefit is that, in most instances, the result or decision can be easily changed. This makes whitebox AI chatbots highly agile.
Blackbox AI is characterised by deep-learning algorithms, typically derived from well-understood annotated datasets that are carefully programmed. In most cases, these programs can only be fully understood by data scientists and programming engineers. Any changes or enquiries into the results provided can only be resolved using very technical resources, with a process that sometime takes many months. This may make for a more robust overall process, but one where the business will struggle to be agile.
What about a hybrid solution?
Typically, platforms such as Amazon, IBM Watson and Microsoft Cognitive Services combine the best of both worlds in a hybrid approach. This allows users to harness the incredible accuracy of blackbox deep-learning technology while also giving them much more transparency around the results – not to mention user interfaces that are accessible for even the most non-technical user.
What’s the difference between a voice chatbot and a text chatbot?
Voice chatbots were in huge demand back in 2018, when companies wanted everything all at once – even before they had successfully rolled out a chatbot pilot in text. This thirst for fully functioning voice chatbots spurred a lot of companies to ramp up their R&D in the voice space, where their voice capabilities are now unmatched and provide a positive experience for their customers.
For the most part, chatbots powered by voice technology don’t differ too much from their text-based counterparts. Perhaps most important to note is that the limitations of text are carried over to voice, and sometimes magnified because of how users interact in voice as opposed to text. This may make the choice between voice and text – at least if this is your first foray into using chatbots in your business – an easier decision.
Another consideration when using voice chatbots is how you structure conversations. This is critical as you need to ensure your solution doesn’t reference information or context that is too far back in the interaction – potentially confusing the user, and thus the chatbot for the remainder of the conversation. Intent can also become complicated when using voice rather than text, so businesses need to carefully consider how they can frame conversations with clear intent.
For the most part, voice chatbots in service today are transcribers – they transcribe the user’s voice query into text to find the most appropriate response, and then provide the response to the user in voice format.
Training a chatbot
If you’ve never used a chatbot before or want to ensure you get the most value from your solution, it’s important to contact a chatbot expert, Antares who can deploy the right tool for your needs. After all, poorly trained AI chatbots can actually cause more harm than good, doing damage to your brand reputation and potentially causing you to lose loyal customers.
There are a few key stages when training your chatbot, and if you are unsure how to handle these steps effectively we recommend talking to us on how we can help you get the right chatbot for your business:
- Outline the problems you want to be solved: Your chatbot will rely on specific use cases to solve problems for your customers. Step one in chatbot training requires you to create a ‘wish list’ of what you want your chatbot to do, such as answering FAQs, tracking the status of placed orders, or any number of other business tasks that can be automated.
- Clarify your intents: Make sure your intents are clear and distinct. After all, if a customer wants to buy something from you but you haven’t made the purchase intent clear for your chatbot, then you’ll likely lose a sale.
- Ensure there’s a wide variety of utterances: These utterances will help drive your chatbot to the customer’s intended target. For example, if they want to buy something, saying “Buy now” or “I want to purchase this” or “Let me buy that” should all result in the same purchasing response from your chatbot.
- Build a chatbot support team: You’ll need to get humans to help configure, train and test your chatbot.
- Add a splash of personality: Just as humans prefer speaking to other humans, they also want to know their chatbot has a bit of personality – rather than a boring, static, robotic response. It should also match your brand: professional tone for white-collar customers; a casual tone that uses jargon for a younger customer base.
- Keep training: As with any business process, you need to keep updating it in order to glean the most value from it. It’s the same for a chatbot. Keep training and improving how it interacts with customers and the breadth of responses it is able to provide.
Tone of voice of your chatbot and the opening message
Thanks to the rise of hyper-connectivity and increased globalisation affecting almost every industry, businesses are finding themselves in more competitive markets than ever before. Because of this, any advantage you can get over your competition is valuable – which is why branding is so critical.
Chatbots should be part of your business’s branding conversation. After all, as the first point of contact for potential customers, a chatbot is representative of your brand values and the tone you want to display.
Tone of voice for your chatbot doesn’t necessarily mean the timbre or cadence of your voice chatbot – although this should certainly be taken into account. For example, if you market yourself as an Australian-owned and operated business, using a voice chatbot with an American accent may not play to your brand’s strengths.
More holistically, what it means is that you should be exuding a tone of voice through your chatbot that speaks to your established brand values. Chatbots should be able to easily engage with users and interact with them in a way that is reflective of your human staff. From the words your chatbot uses to its ability to detect user emotions, finding the right tone of voice could make or break the value of your chatbot.
In much the same way, the opening message is key to customer engagement. A dull greeting message will come across as a lazy attempt to get someone’s business, whereas a well-thought-out opening message that catches a user’s attention is much more likely to engage them, provide them with the information they are looking for, and potentially lead them through your sales funnel.
Value of chatbots
How you define the value of a chatbot will depend on what you are hoping to get out of it in the first place. A business that wants to use chatbots to reduce the number of humans required for generic customer-support queries will find value if the chatbot is able to save them time and money. On the other hand, for a business that wants a chatbot to drive more sales like the Kian app described above (see ‘What are some examples of useful chatbots?’), they will see value in how it increases their bottom line.
You can read more about the ways chatbot QBot was able to revolutionise teaching at UNSW in our case study about our work with the university. QBot helped to create a teaching and learning community for the engineering school, a great example of the ways AI can bring people together.
The real business value of a chatbot can be derived by how it adds value to three specific areas of the business:
- Customer focus: Chat logs left in the wake of chatbot interactions with users give businesses reams of valuable data to exploit. Insights pulled from this data can be used to not just improve the customer experience, but also detect new patterns in user behavior which can be leveraged with new business strategies.
- Freeing up human staff: From answering FAQs to solving common user problems, a chatbot means your human support team is freed up to focus on more complex business tasks.
- Ubiquity of service: The right chatbot can turn the average nine-to-five business into a 24/7 service. This eliminates geographic restrictions and staff-availability constraints, and can ultimately improve profit margins.
Read the blog:
There’s usually one very clear reason why you want to use a chatbot: to respond to queries immediately. Doing so can net you warm leads because you show potential customers that their needs are so important that you are willing to drop everything to respond to their queries.
While chatbots are great for responding to customers in real-time – even when you’re not online yourself – they can’t respond to queries beyond their capacity (i.e. the intents you’ve given them). And when the chatbot can’t answer a particular query or drive the customer to where they need to go, there’s a high chance they will bounce – and probably visit one of your competitors instead.
That’s why it’s crucial to build in a way to follow up on all interactions customers have with chatbots. In many instances, a follow-up won’t be necessary. When the chatbot does its job and answers a standard query, the customer can be driven through the funnel to the next stage without intervention from a human.
If your chatbot interacts with a customer and can’t adequately solve their problem or respond to their query in a way that is satisfactory to the user, however, you want to have built-in ‘fail-safes’ that alert your human support team and allow them to step into the conversation.
Remember that 86% of active buyers are willing to pay more for a top-quality customer experience, yet the majority believe vendors don’t consistently meet their expectations. A chatbot is a way to reduce the risk of missing out on a potential customer. Not only that, they can learn from every interaction, improve themselves and ultimately drive new business to you – a minor investment that could potentially deliver a huge return on investment.
Should I use chatbots as virtual assistants or human replacements?
That depends on a variety of factors, such as your business size, whether you currently have a customer support team, your online presence, the number of customer interactions you deal with on a day-to-day basis, as well as how you want to portray your business to customers.
Some business types may lend themselves to using chatbots as human replacements, especially if there is a high volume of identical queries that are easily solved with a static response. However, this is likely to be the exception rather than the rule.
In the real world, most customers prefer engaging with a human agent than a bot. And while advances in the technology are helping chatbots become smarter and more human-like in their responses, there is an argument that they should never – and will never – truly replace human support teams entirely.
The best outcome, then, is to harness the usefulness of chatbots as virtual assistants in tandem with a human workforce. You’ll be part of a growing contingency of businesses enabled by chatbots, as Gartner predicts 25% of digital workers will use virtual employee assistants daily by 2021, and chatbots will power 85% of all customer interactions.
Chatbots can be used as a ‘first line of defense’ to reduce wasted hours spent answering the same queries by your human team members. Customer requests can be clarified by the chatbots so when the time comes for the human to take over, they know exactly what the customer wants without having to ask an abundance of qualifying questions.
What are some examples of useful chatbots?
WHO Health Alert
In the age of COVID-19, the World Health Organization’s WHO Health Alert chatbot acts as a source of trusted information. Through the popular messaging app WhatsApp, users can ‘speak’ to the WHO chatbot to get travel advice, learn how to protect themselves from getting infected, and dispel fake news around coronavirus.
Margot the WineBot
European supermarket chain Lidl wanted to making finding and choosing wines easier, with the ultimate goal of driving users to purchase wine directly from Lidl. In addition to providing advice on the best wines to match with certain foods and explaining how different wine varieties are made, Margot the WineBot is an AI-powered solution that learns as she goes, even altering her own language style depending on how users speak to her.
Buying a car is one of the most expensive purchases for most people, but Kia has managed to make a serious decision more accessible through Facebook Messenger. Its Kian chatbot is, at its core, a typical sales tool, but the results speak for themselves: driving triple the interactions of the Kia corporate site and enjoying a 21% conversion rate.
Domino’s offers Bot
Ordering a pizza is made even easier with the Domino’s Offers Bot. Australian customers can jump on Facebook Messenger and ask the Domino’s chatbot things like “What deal can I get on my pizza today?”. Vouchers and the latest deals are then sent to the user, with a direct link to online ordering for fast, effective sales.
Platform integration: Microsoft chatbots vs Google, Amazon and IBM
All the major operators – Microsoft, Google, IBM and Amazon – have spent years heavily investing in the field of chatbots. Each of them has also developed their own iterations of chatbots and the flurry of products that have spawned out of the technology – such as the emerging market of conversational AI services that includes Microsoft LUIS, IBM Watson Assistant and Google Dialogflow.
While the layperson hears chatbots and immediately thinks about Apple’s Siri or Amazon’s Alexa or Microsoft’s Cortana, the potential of chatbot technology goes far deeper – especially for businesses.
The Microsoft Bot Framework is a comprehensive hub for building and deploying enterprise-grade bots powered by AI. Microsoft also offers the dedicated Azure Bot Service. This managed service has been purpose-built for the development and deployment of chatbots, helping you create intelligent, enterprise-grade bots that can easily integrate with your current business processes and boost the customer experience – all while giving you complete control over the data.
Despite the technical complexity of some of the most advanced chatbots on the market, creating your own business-ready chatbot is achievable, especially when using the right bot platform like Azure Bot Service.
When you’re ready to get started with a chatbot, there are a few steps to take before actually building one:
- Explore a variety of different bots: You won’t know what type of chatbot you want to deploy for your business until you’ve used a variety of different bots personally. Whether it’s chatbots on websites, via SMS, on Facebook Messenger or even over the phone with a voice chatbot, do your own research and make notes about each different solution.
- Pick the right platform: Do most of your customer queries come through your website? Are your clients asking for customer support on Facebook? Pick the right outlet for your chatbot to get the most value out of it.
- Think about the chatbot’s personality: How do you want your chatbot to interact with users and potential customers? If you’re unsure where to begin, a chatbot personality guide can help you highlight what you should include as part of its persona.
- Write down all the functionalities of your chatbot: This is essentially a brainstorming exercise, but the ultimate goal is to create a comprehensive – yet succinct – list of all the functionalities you want your chatbot to perform.
From there you can build your prototype, test it, reconfigure it (as needed), train it and ultimately deploy it – keeping in mind this is a living, breathing part of your business that must be managed and improved consistently.
Business expectations and training for chatbots
The idea or ‘end state’ of chatbots is, in most cases, to provide 24/7 serviceability. Generally, a chatbot may take around three to six months where the business should be willing to allow it to respond to enquiries without much granular oversight.
Typically, we see that chatbots become highly competent after around 1,500 interactions, so depending on the traffic flow and the amount of training you are willing to dedicate to it, your expectations should be set around a corresponding timeframe.
Ultimately, chatbots must be viewed in through the lens of an investment – that will improve exponentially over time. By leveraging the power of AI and machine learning, your chatbot will soon be giving better answers and giving customers a more user-friendly experience – thanks to its ever-expanding dataset.
In order to achieve these improved results, however, you must invest in consistent chatbot training and continual monitoring of how it interacts with users.
How to get feedback on your chatbot
It’s one of the most overlooked aspects of using a chatbot – especially for business purposes – but integrating a chatbot feedback loop is an easy way to help your solution grow, learn and become even better over time.
Without a feedback loop, your chatbot’s potential is limited to what it can do right now. But with a feedback loop, you’ll be able to leverage all the data your customers (unconsciously) provide when interacting with your chatbot, and this will help you deliver more value to your users.
In-conversation feedback sources are the most obvious form of feedback (i.e. having the chatbot ask for specific feedback during conversations with customers), but don’t discount the value of external feedback sources. This might include looking at your chatbot logs and measuring whether there’s a lower churn rate for customers who had a positive interaction with the chatbot.
The best part about getting feedback on your chatbot is that you can do so without having to ask anyone for it – whether it’s internal staff or your customers who are interacting with the chatbot. Every time a chatbot converses with someone, a log of that conversation will be stored in your chatbot’s database. Over time, this repository will turn into a valuable hub of customer insights exploring the success – or lack of success – of your chatbot’s interactions.
What you do with that information is up to you, but the savviest businesses will use the most insightful feedback to make improvements – whether through direct human calibration and training, or more automated improvements where the chatbot teaches itself over time.
Read our white paper on AI-driven employee engagement: How to attract and retain talent in the modern workplace
Chatbots in general terms are not an owned entity, so there’s no specific price point – costs run the gamut, from the most sophisticated AI-powered chatbots that can be deployed on your systems immediately, to building your own chatbot at a cost relative to the man-hours your team puts into it.
Depending on your requirements, such as its problem-solving capabilities and where you want to use the chatbot (e.g. on Facebook Messenger, SMS, a dedicated mobile app, etc.), pricing will vary greatly.
Cloud Collective is made up of chatbot experts who can work directly with your organisation to understand its needs and how a chatbot can help improve your customer experience. We can get meaningful use cases up and running so you can determine what type of chatbot will integrate best with your current processes, at a price that suits your budget.
If you’re looking to adopt chatbots for your business, we can help you find a solution that perfectly complements your needs. Contact Cloud Collective on +61 2 8966 1496 or email us at firstname.lastname@example.org.
Here are a few blogs on chatbots
7 essential tips to build a better FAQ Bot
Chatbots are becoming a critical digital channel for businesses, and it’s not difficult to see why there has been an incredible level of interest in them. The tips in this blog will ensure your FAQ bot ticks all the right boxes.
How to create an Intranet Chatbot for employees In SharePoint
Using a chatbot in SharePoint, an employee can type in their question and get an answer. Most employees are familiar with this interaction channel, so there is little to no learning required. Sounds great, but how’s it done? Read on.
Why do you need an Intelligent Bot?
“Conversation as a Platform”, or the newly coined term “Conversational User Interface” (CUI) or simply “Bots” is becoming an integral part of our online digital experience. Find out why you need an intelligent bot?
A chatbot is a program designed to imitate humans in order to interact with customers and provide them with valuable information. Chatbots are powered by artificial intelligence software and their main purpose is to simulate conversations in natural language – whether that’s via text or voice.
Chatbots are typically broken into three different pillars: rule-based chatbots, learning chatbots (artificial intelligence), and live chat. All three operate in different ways, so how they work and how you apply them in your business will depend on what you want to provide your customers and how you want your chatbot to streamline your operations.
How long is a piece of string? There is such a wide variety of chatbots on the market that there is so ranking system of the ‘best’ chatbot. Yes, some chatbots are more generic than others, but the trade-off for that may be a lower price point – which could be exactly what your business needs. To find the best chatbot for you, it’s important to ask yourself, “What do I want to use this chatbot for?” From there, you can start looking at solutions that will match your needs.
Modern voice chatbots are, for the most part, transcribers – meaning they receive a customer’s spoken query, translate that to text, parse that information and send the customer a voice response. There is therefore very little difference between basic voice chatbots and text chatbots – the major differentiator is the delivery.
If you’re a business owner just diving into the world of chatbots, you will want to engage an expert who can help you get the most value from it. Cloud Collective can help you deploy the right chatbot solution for your specific needs.
First you will want to explore the market and select the type of bot that’s right for you and your chosen platform (e.g. Facebook chatbot or website chatbot). Then it’s time to think about your chatbot’s personality – how you want it to interact with customers. Then it’s a matter of creating functionalities for your chatbot. If you are unsure how to get started, Cloud Collective can provide expert support.
While they sometimes serve the same purpose, you will get different benefits out of chatbots vs humans. Chatbots can reduce your overheads and eliminate time-consuming, repetitive tasks. Humans, on the other hand, are capable of providing more sophisticated support for your customers.
That will depend on your specific needs, such as whether you want a basic chatbot to send a set of predefined automated replies on Facebook Messenger, or a custom-built AI-driven chatbot which will be more expensive than generic options.