Laptop in home
News

10 AI chatbot best practices for 2022

08.02.22 Written by Futr

Chatbots are all the rage today, particularly in the fast-paced world of retail. If you’re planning to digitally transform your live chat capabilities this year, we’ve put together a list of 10 AI chatbot best practices to guide your journey.

Chatbots have never been smarter or more capable of facilitating genuine interactions with people. Today’s chatbots combine conversational artificial intelligence (AI), natural language processing (NLP) and machine learning to simulate human conversations and understand the context of words and phrases. This allows you to automate responses to repetitive questions, freeing up your employees’ time so they can focus on more complex tasks with higher value-add — such as direct contact with customers. 

If you’re one of the many businesses in 2022 looking to supercharge your customer service capabilities, you may be considering deploying a chatbot in your organisation. You’re probably wondering which platform to get, which channels to use and what features to look for, among other things.

This guide provides a list of 10 chatbot best practices and tips to remember when implementing chatbots and developing a deployment strategy. If you’re a complete newcomer to chatbots, be sure to read our chatbot terminology guide

Related reading: Chatbot terminology guide for newcomers: 12 terms you should know


But first, let’s look at chatbots by the numbers to see why they’re so valuable for businesses in the years to come.  

AI chatbot statistics to remember in 2022 and beyond

  • The global AI chatbot market is forecasted to be worth $102.29 billion in value by 2026. Mordor Intelligence predicts a compound annual growth rate (CAGR) of 34.75% from 2021 to 2026. (Mordor Intelligence)
  • Chatbots are predicted to generate up to $8 billion in annual savings from reduced customer service costs alone. (Juniper Research)
  • 69% of consumers prefer quick communications through chatbots. (Salesforce)
  • 40% of shoppers express no issues talking to an AI chatbot or human, so long as they get immediate answers to their questions. (HubSpot)
  • Virtual customer assistants can help organisations reduce their volume of inbound calls, chat messages and email inquiries by 70%. (Gartner)
  • Global retail spending through retail chatbots is forecasted to reach $142 billion in 2024 — an increase of nearly 5000% from $2.8 billion in 2019. (Insider Intelligence).

Follow these AI chatbot best practices for a smooth deployment

When correctly implemented, chatbots can optimise your customer service capabilities, driving cost savings and efficiency. But all this depends on whether you’re taking the right approach and have a proper strategy to guide your chatbot project. 

Below are 10 chatbot best practices that will help you do just that. 

Related reading: 5 Common chatbot mistakes that annoy customers.

Chatbot best practice #1: Identify your chatbot project objectives

A strong chatbot initiative is driven by clear purpose and objectives. 
Before you even reach out to a chatbot provider or build a custom chatbot, you’ll want to list all the problems your chatbot seeks to solve. If you’re not sure where to start, it helps to start with three primary categories in which these problems typically fall under:

Customer SupportSalesMarketing
Examples of customer support problems chatbots can address include:

• Slow resolution times
• Long call queues and long waiting times
• High contact centre costs
• High average cost-per-call (CPC)
• Burnt out customer support teams
• Low customer satisfaction.
When it comes to sales, common issues include:

• Slow and inefficient lead generation process
• Long sales funnels, with customers needing to do research before making a purchase decision
• Abandoned carts on eCommerce sites.
Finally, marketing problems that align with a chatbot’s capabilities include:

• Low customer engagement
• High customer acquisition costs
• Poor customer retention rate
• Lack of signposts to offers and promotions.

Listing the problems you want to focus on will help you rank them by priority and inform your chatbot goals and objectives. While chatbots can take on multiple roles, every business’s needs vary. For example, an eCommerce site may need eCommerce chatbots to keep up with the deluge of customer queries about shipping, returns and payments. 

Chatbot best practice #2: Decide where to deploy your chatbots

One of the benefits of implementing chatbots is that it enables organisations to scale their customer service capabilities across a wide range of digital channels at minimal cost. This lets you create an omnichannel (“omni” for “all”) customer experience in which people can reach your brand at the right place and right time. 

According to HubSpot, today’s customers want to communicate with companies about their customer service, sales and marketing concerns through at least 13 different channels.

marketing chart
Image source: HubSpot

Before you choose a solution, you’ll first need to decide where to launch your chatbot. 

Many chatbot platforms support a multi-channel presence, while other frameworks are limited to a specific platform, like Facebook or WhatsApp. If you want your chatbot to be an HR solution for your internal communications, look for a solution with chat integrations for Slack or Microsoft Teams. 

Related reading: Chatbot marketing: 4 examples of practical use cases.

Chatbot best practice #3: Research solutions and platforms

Now that you have an idea of what problems you’d like your chatbots to solve and where you’d like to launch them, the next step is to look up tools and chatbot platforms that meet your needs.

For better or worse, the market is saturated with all kinds of vendors and SaaS solutions designed to automate customer interactions, generate leads and sales and collect customer feedback. Some platforms are better than others, but your choice will depend on your specific needs. 

For example, if you’re operating a nonprofit or charity, it’s a good idea to look for a provider specialising in chatbots for charities

Other tips we recommend include the following:

  • Signing up for free trials or demos to see how the product works
  • Looking for customer testimonials
  • Reading the provider’s case studies to see what kind of results they’re generating for their clients.

Chatbot best practice #4: Pick your KPIs

Like any other business project, you should measure a chatbot deployment against a set of metrics and key performance indicators (KPIs) to determine its effectiveness. This is precisely why you need to set appropriate objectives before launching your chatbot project. These objectives, in turn, help you identify your target KPIs. 

For example, consider the following objectives:

Relieve burnt out customer support teams

  • Ticket deflection rate: This KPI, also known as the self-service score, shows the average number of customer support tickets being sent to your live agents. A high ticket deflection score means fewer tickets are being sent to your agents because your chatbots are doing an excellent job of handling complaints and questions.
  • The total number of users: This KPI shows the total number of users that have used your chatbot. A high number means your chatbot answers questions and engages users as designed. 
  • The number of questions answered: This KPI is a good indicator of how successful your chatbots are at answering your customers’ most frequently asked questions.

Increase your leads

  • Completion rate: This KPI shows the success rate of a specific action performed by your chatbot. These actions can include getting users to:
    • Sign up for a newsletter 
    • Like your social media page
    • Complete a quick pulse survey within the chat.

Improve customer satisfaction

  • Human takeover rate: This KPI tracks how often users ask to speak with a human during a chatbot interaction. While not necessarily a sign of a frustrated customer, a high human takeover rate means your chatbot may not be giving users the information they need to resolve their problems. 
  • Engagement: This catch-all KPI measures how frequently and how long users engage your chatbot. If your chatbot has self-service features, such as the ability to pay for something or track something, you want this number to be high. 

Related reading: 3 practical applications of chatbots for customer service

Chatbot best practice #5: Craft your chatbot workflow

Most third-party chatbot providers will have workflow templates that illustrate the flow of an interaction between a human user and a chatbot. For example, a standard retail chatbot workflow can look like this: 

  1. A user lands on your website and clicks on the chatbot button
  2. The chatbot displays a welcome message
  3. The chatbot offers two options
  • FAQs > Show list of FAQs 
    • FAQ 1: Warranty > Show information about warranty claims
    • FAQ 2: Shipping > Show information about shipping fees and times
  • Track order > The chatbot requests the customer’s tracking number
    • The chatbot displays the customer’s order status. 

However, it’s still a good idea to develop a workflow of your own. This step, while simple, is key to creating a personalised chatbot experience. 

Depending on your services, niche and objectives, your chatbot workflow can have more or completely different options. Take the time to think of a thoughtful chatbot story that your customers will find genuinely helpful. 

Chatbot best practice #6: Give your chatbot a personality

As a virtual customer assistant, your chatbot is an ambassador for your organisation. Just as your website and marketing collateral are a reflection of your brand, your chatbot should also represent your values, image and personality. 

You want a chatbot to engage users not just in a human way but also using language that’s consistent with your audience, their needs, goals, challenges and pain points. 

For example, British footwear brand Hotter uses a retail chatbot with an upbeat personality to engage visitors. This personality is consistent with the excitement retail brands wish to convey when engaging shoppers.

Image by Hotter

On the other hand, Norfolk Police take a more neutral approach to their chatbot’s language and messaging — highlighting the more serious nature of the chatbot’s functions.

The key to designing a personality for your chatbot is to find the overlap between your organisation’s image, your target audience’s persona and your chatbot’s purpose and objectives. 

Chatbot best practice #7: Make a good impression with your welcome messages

Your chatbot’s welcome messages set the tone for the conversations users will have with your chatbot. Your first messages also set expectations, primarily if you use them to signpost to information about what your chatbot can do.
Here are some great welcome messages from Hotter and the anti-suicide charity, CALM.

Image by Hotter and CALM

Both chatbots immediately greet users with a list of FAQs, helping customers instantly figure out how the chatbot can help them. With Hotter, the brand knows their customers constantly need information about tracking and returns, so it makes sense to push the information right away. 

In CALM’s case, the chatbot uses welcome messages to explain what the organisation does. Users can also click the chat widget to see the same welcome messages.

Image by CALM

Chatbot best practice #8: Keep messages short

Speed and convenience are two of the biggest draws of chatbots. According to a 2018 report by Salesforce, 69% of consumers prefer engaging chatbots for quick communication with brands. 

This is hardly surprising, considering how people typically communicate on live chat channels, whether SMS or WhatsApp, using short and snappy messages. 

So, it makes sense for your chatbot to provide concise messages. Here are some chatbot tips to remember:

  • Avoid overwhelming users with a wall of text that occupies the entire chat window. 
  • A good rule of thumb is to keep individual messages from occupying no more than a third of the chat window.
  • You can break them apart into smaller chunks for unavoidable lengthy messages  instead of sending them as one long block of text. These chunks also encourage the user to participate, making the interaction a real two-way conversation.
Image by Hotter

Short and snappy messaging also go both ways. By telling users to provide short questions (see the example above), your chatbot has an easier time processing the question and providing an accurate response. 

Chatbot best practice #9: Use thoughtfully-designed fallbacks

As smart as AI chatbots are, they will inevitably fail to understand a user’s inputs, either because the query is new to the chatbot or does not align with its capabilities. When this happens, your chatbots should send a fallback message. 

A fallback message is used if the chatbot fails to understand a query. It allows your AI bot to recover the conversation by asking the user to do any of the following:

  • Rephrase their question
  • Clarify their question with a follow-up query
  • Providing information that closely matches the user’s query.

Here’s an example of a fallback message from believe housing, a housing association in North East England.

When creating fallbacks, it’s essential to keep the conversation going so users don’t abandon the chat. Here are a few tricks to remember when designing your fallback messages. 

  • Use creative responses: “Sorry, I didn’t get that” isn’t a very helpful response. Telling users what they should do through a request such as “Can you rephrase your question?” can elicit a more helpful response. 
  • Design multiple fallbacks: Fallback variations make your chatbot sound more human. This also prevents frustrating users with the same error message repeated over and over again. For example, you can have a fallback for a query with typos and another for a query that may be related to your products and services.
  • Signpost to options: Turn the chatbot’s failure to understand the query into an opportunity to generate a quick response. For example, your chatbot can signpost to a question or a topic with a button (see believe housing’s example). The point here is to help users figure out how to steer the conversation towards a productive direction. 

Chatbot best practice # 10: Provide a clear path to escalate chatbot conversations

Some organisations that use chatbots make the mistake of not having a mechanism that allows users to quickly escalate the conversation to a live agent. In 2018, a Forrester Research report noted that chatbots were frustrating users because of “poor escalation paths to agents.”

Some of your customers will have questions and concerns that need a human touch, whether it’s due to their complexity or urgency. Make sure your chatbot can detect when it can no longer help the user, following it up by providing instructions on how to escalate the conversation to a human. If the interaction happens outside business hours and a live agent is unavailable, your chatbot can generate a ticket and schedule a chat the next business day.

CALM offers an excellent example of how to signpost to a live helpline, ensuring that anyone who needs help knows how and where to get it.

Image by CALM

Position your chatbot project for success

Rolling out a chatbot to engage your customers is a major project that requires planning and strategy. The good news is that today’s advanced chatbots can help organisations in a wide range of sectors, whether it’s social housing, charities and nonprofits, universities or retail, embrace AI and automation. 

Follow the Futr blog to get more insights on how our AI chatbots can transform your organisation’s customer service experience. Contact the Futr team to sign up for a free four-week trial or schedule a FREE demo of our live chat and chatbot platform.

  • Published 08.02.22
  • By Futr
  • In News
Share