You are a proud business owner. You take pride in providing excellent service and personal attention, and so do your employees. Making your customers smile is what makes all of you smile.
But, you’re worried.
Your staff drowns in simple questions that anyone could answer just by searching your site. They grumble. They feel their jobs have become boring.
How are you going to keep both your employees and customers happy, and develop your products and business all at the same time?
You don’t want to skimp on any one of them, but you can spend money only once.
Something’s gotta change.
Could chatbot software be the answer to your predicament?
You’ve heard stories, but you aren’t convinced just yet. You’re not even sure you understand what they are or do, let alone how they can help.
Well, then, read on. This guide tells you everything you need to know to make chatbot software work for you, and your employees.
Before we dive into the technicalities, let’s see what a chatbot is and why you would one.
What is chatbot software, anyway?
A bot, short for robot, is software that can autonomously perform tasks over a network, usually the internet. They perform simple tasks that need to be repeated more often than a human would care to do them. And at a much higher rate than a human could.
A chatbot, simply put, is a bot that can engage someone in a conversation, a chat.
They listen and respond using text or voice, though not all bots can do both.
Like a human, bots can respond with information or with a question that helps them tailor their next response. They can include audio, video and pictures.
When a bot is integrated with your back end systems, it can use someone’s preferences and history, and it can perform simple tasks on behalf of them. Like book a room, or place an order.
All very interesting, but why would you want a piece of software that can chat?
Keeping up with the Joneses, or is there more to it?
Chatbots are hip. And hyped. Everybody and their nanny is having one.
Keeping up with the Joneses, however, is the worst reason to invest time, money, and effort into anything. It would be like buying a car when you don’t even know how to drive and have no desire to travel.
The flip side is that you need to keep up with the Joneses, because chatbots are becoming the norm and your customers expect them.
Chatbots are part and parcel of the virtual assistant they use on their mobile phones, PCs and websites: Siri from Apple, Alexa from Amazon, Cortana from Windows, Google Assistant. And Facebook. Shortly after they opened Messenger to chatbots, the number of chatbots on Facebook grew to 30,000 and 100,000 just a year later.
What’s more, your customers expect you to respond within a couple of hours on social media. Not responding quickly enough is hurting your bottom line more than you think.
That’s where chatbots come in.
The perfect employee to grow your business
Chatbots thrive on messaging platforms where your customers spend much of their time.
They don’t tire, they don’t get grumpy, they don’t get bored. They’ll happily answer the same question for the 100th time. They take care of the drudgery and work 24/7 without complaint.
With a chatbot on the team, your human staff can enjoy a regular schedule and focus on interesting challenges.
A chatbot doesn’t feel any pressure to meet their targets. That makes them less daunting to chat with than to pick up the phone, talk with a human and face a hard sell or have to say ‘no’ to a friendly, helpful person. That means more interactions and more chances to present your marketing message.
When you integrate a chatbot with your backend systems, they can use everything you already know about a customer. They can help your customers be successful with your products and give them an excellent experience. It makes selling and upselling less obnoxious and more relevant.
What’s more, integration enables them to do small tasks for your customers. Like schedule a meeting with multiple participants, book a room, or re-order something.
Just imagine what all this can do for your bottom line. And then imagine what it would mean for your decision making when you can survey and get feedback from every single visitor to your site.
So, chatbots are not just fancy site decoration or playthings.
They can save time and money; improve employee satisfaction; improve brand/product loyalty and improve revenue as they can help retain customers by providing an excellent customer experience both in (up)selling and service.
The best results come from combining the strengths of humans and bots. Because, bots, as advanced as they may already be, still lack some crucial skills. To understand how and why that is, let’s take a quick look at the technical side.
The 2 types of chatbot software (and their differences)
Chatbot software comes in two main flavors. The plain vanilla ones that do their work based on simple fixed rules, and the sophisticated Italian espresso ones that incorporate natural language processing (NLP) and machine or deep learning as part of their artificial intelligence.
Chatbots using simple fixed rules
These bots offer a prompted – guided – conversation defined by a decision tree, aka conversation tree.
The ones that allow a user to type in free text use keyword and phrase recognition and are limited by what’s in their database. It means users can hit a dead end quickly when the bot doesn’t recognize what they type.
Some of these decision tree based chatbots don’t even allow free text entry, making them more of a guided menu or help system than a chatbot.
You might think that these limitations mean that simple rule based chatbots are undesirable.
Not so.
Their strengths lie in their limitations. Their highly structured interface and limited options provide a very good user-experience. It avoids frustrating misinterpretations. It is also quick. And it keeps users engaged by keeping them clicking around.
AI Chatbot: a bot incorporating artificial intelligence
Artificial intelligence in chatbot software can take a number of forms.
- Natural language processing (NLP) allows a bot to analyze a message and figure out what the user is talking about – for example named locations, named people, dates, objects and types of objects; what they want to achieve, aka user intent – such as wanting to know what the weather in London will be like in a couple of days; and what the context is.
- Machine learning allows a bot to learn from every interaction. They can then apply the patterns they discovered in their training to similar situations and slightly different questions. They are not limited to a predefined series of paths, and can come up with relevant responses to a large variety of input.
- The third is deep learning. Where machine learning is very much task specific, deep learning is fully task-independent learning designed to make sense of whatever data it receives and discover the patterns it contains. Bots with this type of learning are used where predefined responses won’t work and are still very, very limited in number.
Bots with NLP and machine learning still answer with “fill in the blanks” predefined responses. NLP has made huge strides in the last few years, but generating text is still in its infancy.
Inseparable buddies: chatbot builders, chatbot platforms, and chatbot software
The terms chatbot platform, chatbot builder and chatbot software are a bit confusing and don’t really cover what the bot does.
A chatbot solution requires four parts:
- A builder to specify your bot’s behavior. The software you use to create the conversation tree and – for AI bots – to train it.
- The bot algorithm: the software that uses the conversation tree, training databases and AI rules to process input and come up with a response from the builder’s database.
- The chat interface: the software that allows a user to type or speak to the bot and presents the bot’s response to the user.
- The bot platform: the machine, or machines on which the software parts run.
While every part can come from a different vendor and run on different – and even multiple – machines, you need them all to have something useful.
The term chatbot platform is a bit confusing. It can be:
- a chat / messaging platform allowing software from other parties to interact with it. Facebook, Twitter, Google Hangouts, Slack, MS Teams, etc.
- an application where the chat interface and the chatbot is just part of the deal. For example a CRM – Customer Relationship Management, email marketing, and customer service software.
Drawbacks of a perfect employee
While chatbots sound like the perfect employee, they do have limitations:
- Chatbots utterly lack the ability to create deep, emotional, human experiences. They simply do not have the empathy to understand someone’s individual emotions. Any empathy they display is pre-programmed.
- Keeping track of context, let alone switches in context, is mostly a pipe dream. Most simple rule based bots have limited context memory – just a few chat bubbles back. The ones with NLP are better, but can still easily get confused.
- NLP can’t handle complex questions or messages with multiple questions.
- NLP is still mystified by irregularities in our use of language. Typos, colloquialisms, accents, culture-specific meanings of words and expressions, are all still insurmountable obstacles.
- Chatbots have no clue about ethics, morals, and socially acceptability. It’s up to their human masters to protect their machine learning chatbots against picking up bad behavior.
4 critical components to chatbot success
1: Go “Big Brother” on yourself
Customers need – and have a right (data privacy) – to know what is done with whatever they type in to be comfortable using your bot.
So, be upfront about who has access to its logs.
And when people know a real human may read what they say, they are far less likely to be unprofessional in their conduct.
That said, humans are humans, so…
2: Turn the other cheek
Expect and prepare for abuse towards your bot. Some people are simply too eager to use abusive language with a bot, just for the heck of it.
If you have a self-learning AI chatbot, this is an absolute must. You don’t want to become the next Tay. Microsoft’s chatbot Tay was turned into a racist hours after its debut.
Still, ensure that your chatbot displays unconditional positive regard for your customers. Most people are decent, so assume positive intent. And deflect – with a bit of humour – what obviously isn’t.
3: Never rest easy
Continuous training keeps your bot up to speed on your products and services.
Use continuous monitoring to learn where your bot is successful in helping your customers, where it is failing, what abuse is thrown at it.
It will give you hard data to decide how to train and develop it further.
4: Love a one-trick pony
Have your bot do one (small, even tiny) thing well. It’s far more helpful than doing many tasks and not being good at any.
When you are ready to add features, consider adding another bot.
Keeping each bot focused on a single topic or task, makes developing and training them much easier.
7 ways to make your customers love your chatbot as much as you
1: Don’t take hostages
Don’t keep your customers hostage to a bot that is lost or stuck. Give them a way out: an easy way to get hold of a flesh and blood support person. Present it early and preferably keep it in view.
Help your staff by giving them a copy of the conversation at handover. It makes the transition seamless for the customer and saves time for your staff.
2: Woo them
A bot integrated with your back end systems is like an assistant with a photographic memory. It can make your customers feel known and personally cared for.
For example, it can retrieve information like order history and status and your customers’ preferences, use that to provide information on relevant new products and services, promotions, and perhaps even to suggest new or different ways in which to use the products they’ve already bought.
3: Don’t help
Ever been to a store where a sales person jumped on you the minute you set foot in the door?
Very few people like that.
So have your bot sitting unobtrusively in a bottom corner of the screen. Only pop it up unsolicitedly when someone has spent some time on your site, visited a couple of pages or searched for information. Use what they have done so far in your offer to help.
4: Like hearing “No”
There are few people who like to say “no” – even to a bot.
When you make saying “no” easy and guilt-free, it reduces pressure and makes it more likely that your visitor continues the conversation.
And that’s what you need, because when you then do make a suggestion or an offer, it’s more likely to be positively regarded.
5: If you don’t have it, flaunt it
Give your chatbot a bit of personality and a “voice” of its own. Make it sound like a human, without pretending to be one.
Being upfront about your chatbot being a bot, inspires trust and makes your customers more forgiving of its limitations. After all, it’s “only” a bot.
Even flaunt its limitations: let your bot comment on them itself. A bit of humor adds to its voice, and helps having it accepted and liked.
6: Provide a free lunch
With a bot, being helpful comes cheap.
A bot can handle many customers simultaneously and it’s a lot cheaper than an employee.
So you can afford to provide free lunches: answers and solutions without pushing for a sale. It evokes positive regard and gets people to come back and more likely to buy from you.
7: Return the favor
Remember Tamagotchi? How we as humans respond emotionally to gadgets behaving in a way that resembles a living being with emotions?
So, when you get your bot to act human-like, expect your visitors to express empathy with it.
Let your bot respond in kind. Use emoji. You may think it unprofessional, but their use is so ubiquitous that not using them would seem strange for a human-like bot.
Aim low, end high: achieve success by iterating baby steps
That was a lot to take in. But you don’t need to do it all at once. The trick is to start small, even tiny, and keep growing.
The two most important things are 1) to provide value for your customers and 2) to keep learning and building on what you learn.
Providing value always beats offering a lot of irrelevant stuff.
Keep things simple. Take time to watch and learn. Let the success or failure of a step inform your next step. Only deploy more advanced features when you have a good use case and the data to back it up.
Drift chatbot, Facebook chatbot, Google chatbot: one or all?
Do you need a Facebook chatbot? Or a Google chatbot for Google Hangouts? Does a Drift chatbot sound like the way to go? Maybe you think you need them all?
What you need is a chatbot that works on the social media platforms where your customers – and their friends – spend their time. That’s it.
It wouldn’t do you much good to have a chatbot that can sing and dance and do the dishes when it can’t talk to your customers in their “natural” environment.
So figure out where that is and pick a chatbot that talks on the most important of them.
How to pick and use a bot that works well for you (5 easy steps)
1: Resist temptation
It’s all too easy to be lured in by shiny features that sound like you need them.
Save yourself some headache. Don’t even go near sites of bot providers, builders or platforms before you figure out:
- where your customers hang out;
- what the simplest, most immediate task is that brings them value
- what the next two or three goals are;
- what questions you want answered to decide if your bot is bringing you the value you seek and what data you need to answer that.
Then keep these answers at the top of your mind when you go shopping for a bot.
2: Learn to walk before you run
Resist the urge to go for “the best”. Instead, pick the simplest chatbot that:
- can meet your first, simplest, most immediate goal;
- is available on the platform where you customers spend their time;
- can collect the data you need to decide whether it’s worth moving forward.
Then, watch and learn:
- Keep tabs on how your customers engage with it.
- Ask your customers for feedback on their interactions with your bot.
- Measure how effective it is in meeting the goal you set it.
- Measure its technical performance (speed, reliability, etc.)
- If your first, simplest, most immediate goal is achieved, work on your next one.
3: Be like a kid
While you are learning, be prepared to ditch it and try again.
Either with the same or with a different chatbot. After all, it may not be the chatbot’s programming, but its training or the rules it follows that keep it from achieving your goal(s).
Be like a kid in a candy store. Try different bots and configurations until you find what works for you.
4: Divide and conquer
Do one job, and do it well.
When you need your bot to do more than one thing, consider having more than one bot. Have each bot do one job well. Use simple bots for simple tasks, use more sophisticated bots for more sophisticated tasks.
Your customers won’t even notice if you give them the same kind of voice and a shared user interface.
5: Be demanding
When you have learned all you can from your experiments and it is time to pick the one to use for some time to come, pick the chatbot that:
- Covers most or all of the goals you want it to achieve.
- Offers “white label” use so you can make it look consistent with your own branding.
- Is available on the social media, voice and messaging platforms where your customers hang out.
- Uses industry standards like JSON and REST for integration with your back end systems, so you don’t get hemmed in and switching to another chatbot at some point remains feasible.
- Offers all the monitoring, logging and auditing you need. Auditing is essential when performing sensitive tasks like placing orders, or dealing with sensitive information.
- Scales well. Meaning you can spin up more copies of the same chatbot implementation when usage intensifies and stop a few when interactions drop off.
- Offers asynchronous execution for long running tasks. Meaning a customer doesn’t have to wait until it finishes but can continue to do other stuff and get notified when it’s done.
- Can be embedded in other applications, such as your other web applications, mobile apps, internal and external back end systems.
- Can be programmed using a language that your staff knows well – or is easy to pick up considering what programming languages they already use.
The last three of course only apply if they are relevant to your business.
The 10 most popular chatbot solutions (in no particular order)
1: Botsify
Botsify creators aim to have it “chat like a human” and help you increase sales and decrease costs.
Like many NLP bots, it has a builder to create guided conversations.
Botsify has been around a while and is well rounded. It can sync with your email list marketing so you can send out interest based articles and “make people think you know what they like to read.”
To get started you can use one of the many pre-built templates or examples in the Botsify Bot Store.
| Conversation | guided, natural language |
| Artificial intelligence | natural language processing |
| Targets | Website, FB Messenger, Slack, Alexa |
| Integrations | Many |
| Offers an API? | Yes, JSON REST API |
| Code | No code needed |
| Free plan? | No, 14 day free trial |
2: Chatfuel
Chatfuel aims to help you automate support on Facebook, drive sales up and costs down. It has many settings to control its behavior, which can feel a little overwhelming.
Its builder offers a flowchart to visualize the guided conversations you define and you can create simple AI rules to specify how your bot should respond to free text.
| Conversation | guided, natural language |
| Artificial intelligence | natural language processing |
| Targets | FB Messenger |
| Integrations | Many |
| Offers an API? | Yes |
| Code | No code needed |
| Free plan? | Yes |
3: FlowXO
FlowXO focuses on communication with customers. It offers guided conversations and can answer simple free text questions.
FlowXO has a “Welcome Mat” feature you can use to inform users about what your bot can and cannot do and how to ask for human assistance.
Another very powerful feature of FlowXO is that it allows your customers to start a conversation on one platform and seamlessly continue it on another.
| Conversation | guided |
| Artificial intelligence | none, free text processing uses exact phrase and keyword recognition |
| Targets | Website, FB Messenger, Slack, Twilio SMS, Telegram |
| Integrations | Many, can accept payments using Stripe |
| Code | No code needed |
| Free plan? | Yes |
4: Manychat
ManyChat focuses on helping you grow your business by generating more leads, engaging prospects and driving sales.
It has a drag and drop interface to define conversations and allows you to send preview messages to FB Messenger, so you can ‘see’ (and demonstrate) what your bot does.
Its FB growth tool helps you deliver digital freebies to grow your audience. On the Pro plan an interested visitor can submit their details through the bot, winning you qualified warm leads.
| Conversation | guided |
| Artificial intelligence | none |
| Targets | FB Messenger, SMS |
| Integrations | Many |
| Code | No code needed |
| Free plan? | Yes |
5: BotKit
BotKit is a bot algorithm and development kit. You can use it to create your own AI chatbot from scratch and run it on your own machines. But you don’t need to. Glitch offers a free place to host and edit your BotKit bot and you can start by “remixing” an existing project.
BotKit has a visual conversation builder and can attach to a BotKit CMS (content management system) to import content into conversations automatically.
In addition to its standard response types – such as dynamic text and quick replies – you can use “interrupts”: messages that will be handled with precedence. For example: help and quit interrupts to create an escape hatch.
| Conversation | guided, natural language |
| Artificial intelligence | natural language processing |
| Targets | Website, Webex Teams, Slack, Google Hangouts, Twilio SMS, Facebook Messenger, Microsoft Teams; you can add adapters to target any platform that has an API you can talk to |
| Integrations | Comes with ready made CMS plugin; you can add your own |
| Offers an API? | Yes, the development kit is an API |
| Code | No code needed, all code available and open sourced, plus dozens of plugins |
| Free plan? | Yes, software is open source. |
6: Aivo
In their own words, Aivo aims to help you “Learn from your customers, train your virtual assistant without any linguistic or technical skills, and solve their requests in real time.”
It consists of 4 products: AgentBot, Live, Help and Voice. They all work together with Live being the chat solution, Voice providing speech processing, Help acting as the knowledge base and AgentBot takes care of the conversation flows and decision trees.
Aivo is one of the few chatbots that includes voice recognition. Its machine learning doesn’t just learn from interactions with your customers, but also assists Help to create SEO optimized content helping you to rank higher in the search engines and offers suggestions to your staff for better and faster answers in Live.
| Conversation | guided, natural language |
| Artificial intelligence | natural language processing, machine learning |
| Targets | Website, Whatsapp, FB Messenger, SMS, RCS (Rich Communication Services), Live (Aivo’s own chat solution), Telephone |
| Integrations | Zendesk, SalesForce, Zapier, Genesys |
| Offers an API? | Yes |
| Code | No code needed |
| Free plan? | No |
7: TARS
TARS focuses on lead generation and qualification using the more human feel of answering a bot as compared to filling out a form.
The 400+ templates available in its builder make it a ready-made store for digital agencies, but really allow anyone to get up and running quickly.
TARS offers a meeting chatbot that is integrated with Google Calendar making it super easy to automate appointment booking.
TARS’ speaking chatbot can talk in many languages, and you can even customize the accent it uses, but it is deaf – it doesn’t have voice recognition and wants it input using fixed options.
TARS’ plans include bots designed by experts. Two important features: running the bot on your own domain (as opposed to a TARS’ subdomain) and sending data to a CRM are only available from Business plan upwards.
| Conversation | guided |
| Artificial intelligence | none |
| Targets | Website (using a subdomain on TARS for Pro plan) |
| Integrations | Google Analytics, Facebook Pixel and Adwords to track results; Google Calendar for meeting chatbots, Zapier and anything with a REST API. |
| Offers an API? | Yes |
| Code | No code needed |
| Free plan? | No, 14 day free trial |
8: Chatbot
Chatbot, developed by LiveChat, aims to help you grow your business. It offers a visual builder to develop guided conversations with many templates to get you started. They also offer a marketplace for templates.
Its artificial intelligence comes from both natural language processing to make sense of free text entered by your customers and machine learning to get better at answering their questions (even new ones) over time.
Chatbot falls back to a default answer when it can’t work what the user wants. As long as it hasn’t learned (or been trained to recognize) the connections between phrases and keywords, you will want to display escape hatch information prominently on this page.
| Conversation | guided, natural language |
| Artificial intelligence | natural language processing, machine learning |
| Targets | Website (using LiveChat or ChatWidget), Slack, FB Messenger |
| Integrations | Many, including Zapier |
| Offers an API? | Yes |
| Code | No code needed, but you can leverage their API to customize Chatbot to specific needs using code |
| Free plan? | No, 14 day free trial |
9: MobileMonkey
MobileMonkey focuses on leads, sales and relationships. It has been named “Best Marketing bot platform of 2019” by MarTech Breakthrough and is a favorite of SEO guru Neil Patel. Little wonder perhaps, as content marketers love its ability to set up Drip Campaigns delivered via FB Messenger.
The artificial intelligence you can leverage through its OmniChat user interface comes from multilingual natural language processing and machine learning. It lets you talk to any customer and not worry about which chat platforms they use: your staff only needs OmniChat.
NLP AI bots can have difficulty associating phrases and words. Instead of offering a default page, MobileMonkey can be set to match keywords broadly, so “hi”, “hello”, “hey” are understood to mean the same thing. It’s still experimental, but should reduce learning and training needs.
| Conversation | guided, natural language |
| Artificial intelligence | natural language processing, machine learning |
| Targets | Website, SMS, FB Messenger |
| Integrations | Many |
| Offers an API? | Yes |
| Code | No code needed |
| Free plan? | Yes |
10: Wit.ai
“Wit.ai” touts it is “NLP for developers”. It’s more than that. It’s multilingual, and includes machine learning and voice processing.
Using “Wit.ai” you can create apps with which you can simply interact to get answers – like a chatbot, but you do much more than that. You can also use it to control hardware devices – by voice as, it includes voice recognition.
“Wit.ai” performs even with limited data. This hinges on the license which allows wit.ai to use the data from the community (the other open bots) to become smarter. Check out their FAQ for the differences between open and private bots.
While you can keep your app private, it is recommended and promoted to make your app open source as well. (Data always stays yours). It’s also why and how you can find many examples to copy and start your own bot.
| Conversation | guided, natural language |
| Artificial intelligence | natural language processing, machine learning |
| Targets | any, docs come with information, recommendations and examples for using “Wit.ai” in your apps, with FB Messenger and on your site. For use in your apps you can start with the official clients (Node.js, Python and Ruby) or the HTTP API. |
| Integrations | It’s code, so you can add your own. |
| Offers an API? | Yes |
| Code | No code needed, all code available and open sourced. |
| Free plan? | Yes, software is open source |
Your turn to make a bot work for you and your team
So, there you have it.
Kudos for making it all the way down here. You have come a long way since you started reading this. And not for nothing. You can leave your worries behind, now.
You have learned how chatbot software can help you save time and money by automating a good chunk of your customer service. Even improving it and making it available 24/7 without hiring an army of employees.
What’s more, you have learned how a chatbot can make your employees’ jobs more interesting again by taking care of the grunt work and handing over smoothly when it reaches the limits of what it can do.
And, you have a strategy to embark on your journey to create your own bot. How to start small and learn with every step you take, so your bot has the best shot at delivering the results you want.
The future looks bright again. With all the time, money and resources getting a bot could free up, you can think about growing again.
So, get your team together and take the first step on that journey: decide what you want your bot to achieve and how you are going to measure its performance.