Chatbots are becoming an increasingly common asset to businesses. They’re being used in every industry and the use cases for where they can help are continuing to increase. Due to them becoming cheaper and more powerful over time.
So what are chatbots exactly? A chatbot is a software designed to talk with humans, via various communication channels like messaging apps, to automate typically repetitive questions or tasks. They are primarily powered by rules but can be driven by machine learning. A chatbot can be designed to work independently or in synchronous with a human operator.
A lot of enterprises such as HSBC, Airbnb, Mastercard and Sephora, have started their own bot programs. By starting early, they’ve been able to take advantage of an untapped market channel it into improving their marketing efforts and customer retention.
Table of content:
- Why are chatbots the future
- Seeing beyond the hype
- When shouldn’t you get a chatbot
- The different types of chatbot
- Facebook Messenger
- Social media
- Machine learning
If you haven’t seen one in action yet, here’s a quick demonstration:
Say you wanted to buy some clothes from an online store. The popular thing to do is go online and visit the site via your desktop or mobile browser:
This is a perfectly valid process, sometimes it’s nice to browse in your own time, explore options and so forth.
Where chatbots start to shine are in instances such as, what if you have a question such as when does it come back in stock? How long does the sale last?
How about if you’re tired of browsing and you’d like some recommendations based on your personal profile (rather than similar items).
Asos’ Enki isn’t 100% there on its ability to answer all those questions, but it’s definitely getting there.
While it’s features are mostly the same as the website with its “visual search” capability, the potential is enormous to develop Enki to be your personal shopper and learn about you and your personal preferences over time. Good for the shopper and Asos!
An added benefit we cover further in our Facebook Messenger chatbots section is instead of needing a separate app or website, you can do it all within Messenger. At EBM, we predict that the number of apps that exist on your phone will reduce. Much like what happened in China with Wechat.
We cover more about this here:
Why Are Chatbots Important?
Chatbots are powered by some of the latest technologies that allow computers to have a meaningful conversation with humans in everyday business at a scale not possible before. They combine (NLP) Natural Language Processing, Voice Recognition, Semantic search and in some cases (OCR) Optical Character Recognition to perform tasks a human being would do.
With these combinations, they are able to process customer’s information and turn them into actions like completing orders, booking hotels, making reservations and other tailored actions.
What makes them important is that they allow humans to focus on tasks that need more human interactions. This allows companies to scale and focus on more challenging tasks. It is a truly revolutionary technology.
According to Gartner, By 2022, 70% of white-collar workers will interact with chatbots on a daily basis. These numbers will only continue to increase.
Whoever you believe wins the messaging app race, whether it’s Tiktok, snapchat, or messenger, you as a business get to win regardless.
Don’t be put off by the fact this is a 2017 metric (we’re still waiting on the BI 2019 public release of this data) businesses are continuing to Invest:
Seeing beyond the hype:
It’s true, chatbots have a bit of a bad rep in certain circles and for good reason. Around 2017, lots of hype were flying around. Chatbots were the next big thing. People splurting:
“Chatbots will make apps obsolete!”
“Messaging is huge, 80% of companies will have chatbots by 2020! If you don’t adapt you’ll fall behind!”
And so on.
2 years on, we can confidently say apps are going nowhere. You will, however, be losing ground on competitors if you haven’t adopted a chatbot.
Our intention saying this isn’t to restart the hype cycle or drive sales by fear. We’ve personally build bots, both simple and complex and both have delivered tangible returns of investment and improved customer experiences.
Fortunately, we’re now also exiting the trough of disillusionment and entering the slope of enlightenment. The next few years of chatbot development will start delivering some exciting capabilities and increase the range of chatbot use cases – drive home that ROI even further.
So if you haven’t already, it’s well worth at least creating a minimum viable product and getting the basic infrastructure in place so that you can scale faster and easier as technology progresses.
If you’re wondering where to start, Drift has an excellent report on the current state of chatbots and where you can start solving pains your customers are likely to have:
Will other channels become completely obsolete? Unlikely.
What we believe will start happening in the lines will start to blur which is why everyone keeps suggesting omnichannel is the way to go. We cover more about Omni-channels later on in this article;
The different types of chatbots
In 2019, if you still have a basic live chat on your website with little or no automation, then you quickly need to review the platform and process you’re using.
Even on a tight budget, there are powerful tools such as Drift that can:
- Book meetings with sales reps
- Answer popular FAQs
- Provide customised introduction messages and flows depending on the webpage
- Push prospects further down the sales funnel with engaging and personal content
- 24/7 responses
In the customer service and marketing sectors, there’s a huge change going on. The term “conversational marketing’ flies around a lot. It’s basically a synonym for creating a seamless and personalised experience for your customers – and chatbots are the things that can enable your business to do that.
A particular example where chatbots and tools see increased growth is something Drift call “sequences”. https://youtu.be/B16rOkMgqUU
Sales reps can target and send emails to prospects via Drift email sequence. If the potential client doesn’t reply before they visit the website again, the Drift sequence is triggered and can either send the email on the live chat widget, or send a personalised message unique to the sales reps and prospects conversation.
What are Facebook Messenger chatbots?
Messenger is alone is a powerhouse.
Facebook Messenger really should be on your priority list of chatbot assets.
The key reason:
Its adoption rate.
1.3 billion people use Messenger.
Regardless of your opinions of Facebook currently (which has some very-deserved criticism), Facebook Messenger and the likes of Chatfuel are huge influencers the chatbots resurge of momentum.
(Chatbots are nothing new and have been around since 1964)
Messenger bots are slowly changing the way customers shop.
Today, there are over 300,000 bots on Messenger. They’re collecting information, making product recommendations and taking your fast food orders.
Should You Create a Facebook Messenger Chatbot?
I know, I know. I’ve built an incredibly strong case and you’re now wanting to create one yourself.
But before you do: make sure you have these boxes ticked:
- Are your customers using Messenger?
- Is your brand personable enough to talk to your customers via Messenger?
- Can any other of your marketing campaigns be optimised first?
- Are there any other channels that may be better suited?
- Do you actually have the money and time to support and grow it?
- What are your social media goals?
If you’re looking to launch you’re on Facebook Messenger chatbot and you’re on a tight budget, we highly recommend checking out the tools and these get started guides:
If you’re a medium-sized company or above:
While these tools are great, they have many limitations and aren’t the best if you’re wanting to:
- Integrate into multiple systems,
- Have high-security requirements,
- Need deployment management and version control
- Require on-premise hosting
- Scale across multiple use cases and grow into larger user numbers of 5000+
Then you’ll want to either consult an agency or use tools built specifically for large companies like EBM.
Social media chatbots
You also have the challenge of figuring out which channel your bot will be on the most effective.
Social media bots such as on Twitter and Instagram work differently to those on Facebook Messenger.
While on messenger you can build conversations using logic or NLP, social media chatbots typically to three things, which we cover in a moment in our business applications section.
Streaming management chatbots
Along with the meteoric rise of eSports. Chatbots are also popping up on the popular streaming service Twitch.
Twitch is basically Youtube dedicated for live streaming games like the insanely popular Fortnite or Minecraft.
These twitch chatbots can:
- Moderate the streamers chatroom
- Greet new viewers
- Answer popular questions
- Greet new users
- Post pre-made messages
- Add Extra function to the Livestream
All in the effort to improve engagement and optimise the streamer brand.
Setting up these bots is relatively easy.
There are a variety of free and paid chatbots that are used by Twitch streamers, many of which can also work with broadcasts on other services such as YouTube and Mixer.
This field is becoming so popular, that there are dedicate chatbot providers for it!
At EBM, we see voice bots as a layer on top of a chatbot. Usually, because the infrastructure stays the same, you but add a speech to text convert on top.
Of course, with the likes of Alexa and Siri, you can have voice-only bots. We find, however, that going for a hybrid approach produces better results for most business cases.
What are Omnichannel Chatbots and why is everyone saying you should have them?
Thanks to social media and messaging apps, customers want speed, convenience and the ability to contact your business on whatever page or channel they’re on at the time.
Dictating one single channel where your customers can contact you won’t cut it anymore.
Omnichannel chatbots focus more on the user experience by giving the customer a seamless interaction regardless of which channel a customer uses.
Imagine a scenario where you have to contact a business for support on SMS but decided to continue your conversation with the business on Skype.
When you switch between these channels, in most cases you are required to start the conversation all over again. This can be frustrating to a customer.
A customer does not care about technology, they care about only one thing, and that is getting their problems solved.
With an Omnichannel approach, this hurdle is eliminated and a customer can pick off from where they left off on any channel their last conversation with your company’s chatbot was made.
Machine learning chatbots
Lastly, we have end-to-end machine learning (also known as natural language generation) chatbots.
This is where we use machine learning to create models based on historical conversations that create response messages from scratch. Rather than looking up a response in a database that we directly train the chatbot to respond with.
These types of bots aren’t often used in companies and large scale applications yet as, frankly, they don’t perform as well vs NLU and flow-based chatbots like the ones shown above.
What are the common business applications?
According to recent studies, an increasing number of companies across industries are planning to deploy AI chatbots in the near future to help them offload manual tasks and make space for more strategic work and customer relationships.
The use of AI chatbots in the travel and hospitality sector, for example, is projected to grow by a whopping 241% over the coming 18 months.
In the past 3 years, financial services, travel and fashion industries have led the way with the early adoptions of chatbots, typically using them for customer service and virtual assistants.
The most common applications for bots in businesses are:
Chatbots have an array of applications, especially in social media
1) Most commonly, bots on Instagram and Twitter are just software that automates certain against like re-tweeting, following accounts and liking comments. Which strictly aren’t chatbots.
2) Provide single answers to questions and tweets.
3) trawl all of the relevant hashtags to collect data to analyse with techniques such as sentiment analysis or text classification
Lead generation & sales
As we mentioned briefly using the Drift example, chatbots are making the sales process more personal and seamless.
Facebook Messenger is becoming the next “email marketing”. Thanks to Messenger enabling you to create messages that are interactive and personal, conversion rates are going through the roof:
Secondly, email is no doubt becoming saturated. It’s becoming increasingly difficult to get website visitors to input their email in popups and signup forms.
A more practical way is to make use of chatbots to ask for customer leads and then follow-up with your email marketing campaigns.
A customer will be more likely to submit their email to a chatbot than an email newsletter.
This is because a lot of website visitors now translate signing up for a newsletter form as spam.
Other ways chatbots are improving lead gen & Sales:
Sales assistant: Integrated to your CRM system, your chatbot can act as an assistant to sales personnel, by notifying them when they are assigned opportunities, simplifying lead creation and updates.
Improving response times: Harvard Business Review shared an interesting study from Sungkyunkwan University. According to the report:
“Companies that try to contact potential customers within an hour of receiving queries are nearly 7 times as likely to have meaningful conversations with key decision-makers as firms that try to contact prospects even an hour later. Yet only 37% of companies respond to queries within an hour.”
By using a chatbot to respond immediately, you’re increasing your chances of a meaningful conversation.
Provide attended digital support to your staff or team by doing basic tasks like data validation, customer complaints etc.
JPMorgan launched COIN, a chatbot that primarily analyzes legal contracts exponentially faster than human lawyers can.
The chatbots also autonomously grant access to software systems and replies to common FAQs, whether that’s resetting passwords, finding out how much leave they have left and so forth.
Customer service is easily the most popular area for chatbot adoption currently and for good reason.
It’s much easier to upsell to current customers than it is trying to find new ones.
Customer experience has always been a staple to successful companies. As we’ve touched on, improving response times is one step to a happier customer.
The average customer expects you to respond to their query within 24hrs.
A chatbot can work 24/7 as customer support by responding to user queries. Not only that, 80% of customers believe that the experience a company provides is as important as its products and services.
The current digital transformation we’re facing is raising customer expectations. Customers are becoming more informed and it’s now easier to be less loyal.
More than two-thirds (67%) say their standard for good experiences is higher than ever.
The areas chatbots are assisting (but also increasing this level of expectations!) are:
- Reducing customer waiting time. We touched on this enough already.
- Efficiently directing customer enquiries to the right agent. By using a combination of machine learning and bots, we can dissect a customers request and make sure it gets diverted to the right company, department or person. Especially useful for holding companies such as Unilever.
- Prioritising the waiting line: We’re currently working with multiple mental health charities, using machine learning and chatbots to identify if anyone needs urgent attention by using sentiment analysis. Rather than that vulnerable person having to wait behind a queue of 100+.
To give a specific example with tangible results in the financial sector: HSBC released AiDA, which has had huge successes so far: cutting the average handling time in half when handling a conversation from end to end and delivering customer satisfaction scores that are comparable with human agents.
AiDA has been shown to be up to 6.5X more productive than the average human agent and has led to increased efficiencies at HSBC.
When shouldn’t you get a chatbot for customer service?
Before we run through more detail about why you should invest in a chatbot and the various areas they can benefit.
For this, we revert to our good friend Alex Mead, who is currently driving global Customer Experience Innovation & Transformation within the Travel & Airline sectors.”
“Many companies are actively pushing chatbots and virtual assistants as solutions to improve their customer service experiences.
These solutions do have a place, but many are failing.
Here’s why I think that is:
1 – There is quite simply very little ‘Intelligence’ in most chatbots, many people confuse them with ‘Intelligent Assistants’.
2 – Most are actually just Avatar layers overlaid over simple digital knowledge bases, and can’t operate very freely at all.
To truly add a layer of intelligence and value to customer experience, the next 3 areas first need to be resolved…
3 – Most chatbots are simply not able to handle complexity or consider a customers ‘bigger picture’ requirements. I.e. They can recognize keywords, but the vast majority cannot take into consideration wider customer service aspects such as delayed deliveries, lengthy product issues, ongoing complaints etc.
So integration to CRM, ERP, etc really is required.
4 – There is often a lack of personalisation in the customer contact chatbot journey. Only the best designs will pull through the individual customers existing orders, deliveries, quotations etc and allow the customer to interact with those within their chatbot journeys.
So, personalise the chatbot journeys and personalise the customers service experience with data relevant specifically to them.
5 – Finally, they mostly fail in terms of continuity and convergence. Continuity means allowing a customer to seamlessly move to agent interaction, with full context pushed through, and convergence means letting the customer even jump across channels with no loss of context.
So, integrate the chatbot interaction into your multi-channel routing strategy so customers never have to repeat information across chatbot, agent or channel.
To truly add value to a chatbot, please think about addressing points 3 & 5 above. If not, you may as well just invest in smarter digital self-service designs. To me no chatbot is better than a dumb chatbot.”
Also known as digital personal assistants, the most popular are the likes of Siri, Cortana, or Alexa who I’m sure you’ve asked to “turn your volume to 11”.
There are also virtual assistants helping customers navigate their day to day with greater ease.
Here’s a bunch of examples:
- Personal shopper for beauty products.
- Mental health bot A non-judgmental and open ear ready to listen at any time.
- Tech News bot Personalised tech news.
- Personal finance advisor bot. Helps you manage your money with some wit & sass to boot.
- Meeting Scheduling bot. Your personal assistant to save you from the “when are you free next?” email chain!
- A bot that’s your friend. In China there is a bot called Xiaoice, built by Microsoft, that over 20 million people talk to.
- Chatbots that help GPs diagnose patients before seeing them
These are but a few scenarios on how businesses and people in day to day life are leveraging chatbots to improve their lives.
How do chatbots work?
In a simple summary, there are three types of chatbots:
- Functions based on a set of rules and logic: These bots are very limited and only answer very common FAQs or commands. Usually part of a premade package like Drift or Zendesk. Human handover is considered a must for customer service with these bots. Rule base chatbots work incredibly well for Messenger marketing – as long as you specify it capability early!
- Machine learning and logic: A more advanced version uses a combination of machine learning and logic (like most of what we build). Large companies such as ICAEW or HSBC where they want to expand into more complex use cases like handling accounts and automating HR enquiries need powerful natural language processing tools to capture and understand more enquiries.
- Natural language generation chatbots: that are entirely built using machine learning. Such as the Xiaoice example we gave earlier. Good for gimmicky things, but still have a long way to go before they can be used in businesses (The ethical debate also still rages on with the likes of GPT-2)
We cover these categories in much more details in our what are the 7 levels of chatbots and which is right for you? article
How to get started with your own chatbot:
Let’s look closer at 3 of the approaches:
- Chatbot builder platforms
- NLP tools
- Building from scratch
Chatbot builder platforms
If you’re on a tight budget or don’t have the in house expertise to use the more advanced NLP tools like Dialogflow, then these bot builder platforms may be able to solve your needs such as:
These platforms while quick to start and cheap to use, come with some downsides, to mention a few such as limited configurations, lack of scalability, dependence on one service and lack of custom integrations and development.
Natural language processing (NLP) tools:
If you’re an enterprise looking to create your first chatbot and MVP, then we recommend building your first bot on a system that’s built to scale specifically for large businesses such as:
Building from scratch, either in-house or through an agency.
Of the two options when building from scratch: building in house or an agency, we definitely recommend you outsource to an agency like Filament.
Unless that is, you have a dedicated team of chatbot developers each with extensive experience behind them, much like Ankit and his team at RBS.
The advantages of building a custom bot with the help of the advanced NLP tools are:
- full compliance with your needs
- control over data the bot receives
- the possibility to make corrections easily
- best for customer-facing chatbots for enterprise
- Easier to scale and grow into other verticals or use cases.
This method, however, will require significant time and investment. We cover exactly how much here.
If we managed to inspire you to create your own chatbot, here are some tips to help you get started. In a very oversimplified summary, there are 4 main stages:
- 1. Define the goals.
- What’s your business goals?
- What success metrics do you need?
- What should your chatbot do?
- 2. Choose a channel.
- Where do you customers/users hang out?
- What channels do you need?
- Mobile app,
- Facebook Messenger,
- 3. Create, test and launch.
Again this is a bit of gross simplification which is why we’ve created a bunch of resources you can take your time reading.
We walk you through each part clearly to get you up and running in no time:
- What does it take to build a chatbot?
- How to build your first chatbot on a budget
Artificial Intelligence, machine learning, NLP and NLG
Some complex topics and analogies here.
Let’s break them down and understand the difference between them all:
Artificial intelligence is a difficult term to define since it’s so brazenly used in so many ways.
According to the Oxford Dictionary, the trusty ol’ book defines Artificial Intelligence (AI) as:
“The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages”
From our perspective at EBM, artificial intelligence is a subset of computer science. It exists to allow computers to behave like humans. While we’re making tremendous progress, it’s still decades away from competing with humans on many tasks. So no need to worry about Skynet and the terminators just yet.
The majority of the time, we use the term machine learning rather than AI:
Machine learning is a subset of artificial intelligence.
The term artificial intelligence is usually used when the marketing department wants to create hype around something.
Machine learning is a field of computer science which enables computers to learn without specific programming. Sometimes it can be assisted and guided by humans, sometimes unassisted.
Machine learning’s main focus is to provide algorithms which can be trained to perform a task.
It is closely related to the field of computational statistics as well as mathematical optimization. It contains multiple methods like Supervised Learning, Unsupervised Learning, Semi-supervised Learning and Reinforcement Learning which each have their own use cases and algorithms.
Machine learning can be broken down into:
- Expert systems, where computers can be programmed to make decisions in real-life situations. The integration of machines, software, and specific information allows the system to impart reasoning, explanation, and advice to the end user.
- Natural Language processing, which we cover more in a moment
- Robotics, are programmed computers which see, hear and react to sensory stimuli, such as light, heat, temperature, sound and pressure.
- Gaming systems: such as AlphaGo or Alphastar
Natural language Processing
We cover the topic of natural language processing here.
To learn how it applies to chatbots, read here.
How do AI and ML apply to chatbots and your business?
As we’ve covered, chatbots don’t need AI to work well. But in the instances when you need NLP to get your chatbot to work well… isn’t that really hard to do?
Short answer? No, you don’t have to be an expert at artificial intelligence to create a chatbot that has NLP capability.
Again, you’ll likely be limited to what capability your chatbot can have. But if you have enough time and grit, you can get really far with the likes of chatfuel and Dialogflow which need zero-to-minumum code to make work.
The other exciting news is the industry is progressing rapidly and tools are becoming ever-more accessible. For example, with our EBM platform, it takes care of the complicate natural language understanding integration stuff, leaving you to just worry about the bot management and user experience.
Decided not to build your own?
That’s ok! Contact us and we can help.
The Future of Chatbots
It’s a very exciting time to be involved in chatbots. It’s hard to say exactly how the technology will evolve.
What are the possible applications when we merge bots with 5G technology?
How will Augmented Reality (AR) come into play?
What happens when algorithms like GPT-2 get even better and apply across more use cases?
Asking these questions and quickly turns conversations philosophical ones.
What we are confident in predicting is chatbots will start communicating with each other, everyone will have their own personal assistant and you soon won’t be able to tell when you’re talking to a chatbot. (Except, of course, when it explicitly tells you so!).
Whatever happens, we’re just happy to be part of pushing the frontier of this technology.
Where do other chatbot enthusiasts hang out?:
Do you have a list of questions even after reading all our guides? Do you just want to ask people some questions? Maybe get a bit of help with your project from a super friendly community?
You’re in luck because there is a slack channel dedicated to chatbot enthusiasts like you’re now becoming.
Simply click here to get an invite link to join and join the movement to make the greatest bots.
- Botlist, an app store for bots.
- How To Build Bots For Facebook Messenger by Facebook
- Building Your Messenger Bot [Video] by Facebook
- Creating a Bot by Rob Ellis
- A Beginner’s Guide To Your First (Slack) Bot by Slack
- Slackbot Tutorial by Michi Kono
- Create A Slackbot Using Botkit by Altitude Labs
- How to create your own Telegram bot who answer its users, without coding by Chatfuel