Chatbots, virtual assistants, digital humans – there are many conversational AI technologies designed to interact with us in a way that mimics chatting with a real person. The catalogue of conversational AI platforms is set to become a $17-billion industry within the next four years.
Chatbots are currently the most widespread among businesses, with 80% of brands expected to be using them.
So clearly we’re all well versed in chatbot technologies, right? Well, not if you look at some of the most googled questions people have about chatbots. In fact, it seems many are asking the simplest and most basic questions.
So, let’s answer them.
As a bonus, we’ll also give you some insight into the difference between chatbots and digital humans – and how you can use both to create a more well-rounded, experience-focused digital workforce.
What is a chatbot?
Let’s start with an easy one. Chatbots are computer programs that can simulate human conversation either through a rules-based approach or by using artificial intelligence (AI) and natural language processing (NLP).
These conversations are usually through written text. Meanwhile, platforms that recognise and respond primarily to speech commands – such as Siri and Alexa – are more commonly called virtual assistants or voice assistants. When AI uses voice, written text and has a face, then it’s best described as a digital human.
How do chatbots work?
You’ll usually see chatbots either on an organisation’s website or via messaging app or social media sites (Facebook, Slack, Telegram, etc). They pop up on the screen and users type in a question or statement, which the chatbot can then interpret and respond to.
A rules-based chatbot will come back with a list of pre-determined actions or answers based on a ‘playbook’ set up by developers on the back end.
More sophisticated chatbots use AI and NLP to better understand sentence structure and intent. It means they can return more accurate, nuanced answers. Rather than rely solely on pre-determined outcomes, these AI-powered chatbots learn more with each interaction. They become better at distinguishing between ‘right’ and ‘wrong’ responses and adjusting accordingly.
What is natural language processing (NLP)?
Natural language processing is the technology that lets you process language data – like the words a customer types to a chatbot or says to a digital human – and come up with a response. Think of it like the brain, designed to process information and choose the most appropriate reply.
Conversational AI platforms like Google’s Dialogflow, IBM Watson, Amazon Lex, Microsoft Bot Framework and many others handle the NLP. So, if you have a chatbot, you’re likely to be using one of these platforms for writing the conversations your chatbot or digital human will have, based on what users ask.
What are chatbots used for?
We recently asked this question to a range of chatbot employers, leaders and enthusiasts, and found there to be a rather even split.
In total, 14% say they use their chatbot solely for converting new customers. Less than a quarter (23%) use their chatbot for customer support only. And 63% have a hybrid strategy of both business development and ongoing support.
This multi-functional strategy makes sense. In both scenarios, chatbots help lower the burden on human teams by redirecting people to useful areas of a website, answering frequently asked questions and other simple tasks.
Chatbots can be programmed to assist with administrative duties like password resets and updating contact details. They react quickly and are available 24 hours a day and from any location. This frees up regular staff to focus on the things only a human can do.
Chatbots do have limitations, though. Only 14% of customers say chatbots are friendly and approachable. They may be great when simple answers are needed urgently, but they struggle to build deeper, more empathetic connections with customers.
And there are situations where this is important, such as responding to people’s urgent financial concerns, helping them buy a dream home or handling sensitive queries. For these emotionally impactful journeys, digital humans and real people can offer a more human touch. It’s why brands today are focusing on building digital workforces and experiences that offer more of a human touch than chatbots can alone.
How do I make a chatbot?
We’re going to be a bit annoying and answer a question with a question: how much coding do you want to do? No coding is required if you use a service like Chatfuel or WotNot, as you can create a chatbot in just a few hours using a drag-and-drop style function. These chatbots are ideal for much smaller businesses that just need a little support with their customer services.
Coding will be required if you want something that’s tailor-made for your business, so that’ll usually involve more heavy lifting. Whether that’s a little or a lot depends on how bespoke you want your chatbot to be.
That said, chatbot frameworks such as Amazon Lex, Google Dialogflow and IBM Watson include helpful, off-the-shelf natural language tools and open-source libraries. It also makes sense to use these frameworks if your chatbot is going to exist within the wider development ecosystems of these companies.
How much does a chatbot cost?
There’s no easy answer to this question because not all chatbots are created equal. Generally, the more sophisticated the bot, the more time, money and resources have gone into developing it.
Technically, you can build a chatbot for nothing if you use a free trial on platforms like DialogFlow or Watson. But most brands will usually hire a chatbot company to do much of the strategic planning and building necessary, so prices will undoubtedly vary.
Want to build a digital human? You may want to talk with a chatbot company to ensure you create conversational elements that are specific to your brand and customer. We’re happy to recommend some great chatbot providers.
How long does it take to create a chatbot?
At the risk of sounding like a broken record, it depends! The length of time it takes to make a chatbot will vary significantly based on the type of solution needed, the team working on the project and the resources you’ve got in place.
Let’s look at an example. Arcus Lending sought our help in building an innovative digital human, Rachel, who could help answer any questions people have about the mortgage process. Arcus chose an NLP company from our Partner Program to develop the dialogue (the ‘chatbot’ bit). Meanwhile, we elevated the conversation to the next level in digital human form.
It took two months to take Rachel from ideation all the way through to a full customer-facing launch. From here, Rachel will be iterated upon, as she’s trained to be better and better at her job.
How do you break a chatbot?
No chatbot is failproof, so there will be times when it’s unable to interpret or respond to requests correctly. In other words, you’ve ‘broken’ your chatbot. But doing this intentionally does have its benefits. Knowing what questions will stump your chatbot is a great way to start planning ways to prevent it from happening.
For example, chatbots often struggle with non-traditional answers and ‘filler’ words. You may have programmed your chatbot to recognise ‘yes’ and ‘no’, but what about ‘yeah’ and ‘nah’? or ‘yup’ and ‘nope’? Even saying ‘goodbye’ can break a bad chatbot if it’s not been set up to end a conversation properly.
Filler language is also a problem. Humans often type how they speak, adding interjections (think ‘hmm’, ‘oh’ or ‘bah’) and colloquialisms (‘wanna’ and ‘y’all). Fortunately, conversational platforms like Dialogflow cover a lot of these bases for you, saving you from having to manually program in every colloquial greeting.
People could also mumble, horrendously misspell words if people really wanted to break your bot. But even then, having fallback dialogue such as “could you repeat that” or “I don’t understand, could you please rephrase the question” can cater for a lot of errors.
All in all, it pays to see how your chatbot can be broken and work backwards to fix it.
How to tell if you’re talking to a chatbot
The fact this is still one of the web’s most-asked questions shows how important it is for people to know when they’re talking with a chatbot and when they’re in a live chat situation. It also goes to show how convincing some conversational AI technology is becoming.
So, how do you tell? The simple answer: ask. Chatbots should be programmed to respond to this question, while a live chat agent will identify themselves.
Interestingly, digital humans are designed to clearly be non-human and non-chatbot. We believe transparency and honesty should be underlying principles in the design of AI-driven technologies. In fact, it’s one of the five laws of ethical digital human design – a digital human should never pretend to be what it’s not.
How do you deploy a chatbot?
Your deployment strategy is likely to be influenced by what NLP platform you’re using. As we’ve mentioned, many chatbot frameworks have their own development ecosystems and integration requirements, so the process of setting up and launching your chatbot may be specific to that environment.
For instance, Dialogflow has its own documentation with specific step-by-step instructions for deploying chatbots that are built using its conversational AI. Meanwhile, Amazon Lex has ‘one click deployment‘ capabilities.
Deploying a digital human also depends on how you intend to use it. You can have custom integrations into a website, app or even a physical kiosk. You can also use our pre-built UX tools to integrate your digital human solution in just a few clicks.
How do I make my chatbot better?
OK, OK… cards on the table, this isn’t one of the web’s most googled questions about chatbots. But as chatbots continue their exponential growth, we’re sure it will be a question most brands ask to make their chatbot stand out from the crowd.
And if you’re serious about the potential of chatbots and conversational AI technology, we know you’ll regularly pose yourself this query.
We asked people what their next priority is for the development of their chatbot, and the top response was “to create a more human experience”. If you’re not looking at how to make your chatbot more human, you’ll soon find yourself trying to catch up to those who currently are.
We outline exactly how to do that with a customer-experience-first approach, not a technology-first approach, in our eBook, ‘Building a digital workforce’.
You can download for free below. And if you have any questions google can’t answer, we’d love to help.