Chatbot automation is rolled out by brands to help reduce load on staff. Companies already working with chatbots are extending where they deploy the tech. Indeed, according to Gartner, by 2027, chatbots will be the primary customer service channel for around a quarter of businesses.

Below, we explore how to make chatbots work for you and your customers.

Are chatbots effective?

According to IBM, chatbots significantly improve customer satisfaction and decrease time spent by employees dealing with requests. Meanwhile, a survey by Gartner highlighted that in January and February of 2022, 54% of respondents used some form of chatbot for customer-facing applications.

That being said, IBM also drives home the point that “Continuous improvement is a critical component of Virtual Agent Technology performance.”

The tech solution you choose has a big impact on the user’s experience. For example, IBM’s Watson Assistant solution is one of the most robust and natural systems available in the market. Unfortunately, we often see chatbots with huge potential that are implemented without proper research or understanding of customer needs. Influencing factors include:

• Internal skills of the team (team upskilling required)
• Lack of data around customer behaviour
• Insufficient scope or goals around the problem you’re trying to solve

Building first-rate chatbots doesn’t need to be difficult. Just keep in mind there are no corners to cut and that doing it improperly is worse than not having them at all. There are a few simple points to focus on that we outline below.

What problem are you solving?

There are typically two areas where chatbots excel:

1. Quickly surfacing information for customers so they don’t have to search online help centres, order systems, or get assistance from a support agent.
2. Collecting data for top-of-funnel marketing, such as requesting demos.

Creating user stories/scenarios is a valuable way to work these out. Examples include:

• I want to book a software demo
• I want to ask about my order and see shipping information
• I want to find developer documentation for a specific API call

Outlining customer goals helps develop a script to test against as you build your bot and its dialogue. It also helps create intent phrases needed by natural language bots to understand how a customer may ask a specific question.

Take the story “I want to book a software demo”. You can create a few intent phrases around that to help train your bot, taking a lot of the guesswork out of the setup and helping your engineers produce better training for the bot. Here are a few examples:

• Booking a demo
• How to book a demo
• I want to book a demo
• Can I book a demo online

Who are you solving the problem for?

This is important for a number of reasons. Users of your chatbot will have typical journeys with specific goals they want to achieve, needs they want fulfilled, and pain points they want alleviated.

So, understanding who you’re assisting and what they want to do and need helps you design the bot in a more human-centred way.

Understanding how your users talk

Keep in mind that different demographics may respond in varying ways to chatbot responses. Meanwhile, terminology and numerical systems may diverge depending on languages and cultures.

Understanding how your users search for information is important as well – use site search logs from your own website as well as queries from Google Search Console to piece a picture together. Combine that with human agent requests to build a comprehensive view of the language your customers use to find answers to their problems.

Furthermore, continual monitoring of queries through site search and working with SEO teams to understand what users are looking for enables consistent improvements to your bots.

Creating content that leads to satisfaction and engagement

A leading use of chatbots? Reducing the number of support agent requests. However, if the information the bot serves up is unhelpful, you frustrate the user even more before they speak to a person. 

Your content needs to directly answer a problem or an FAQ. To that end, regularly review the data you retrieve from chatbot interactions. Queries that lead to an unsatisfactory response indicate a need to improve existing content or create new copy. 

Watson Discovery is a particularly powerful tool. Why? You can scan content hubs and use AI to index copy and images of interest that can be delivered via chatbot. That allows the bot to answer specific questions using the content you have with incredible accuracy and speed.

Choosing a platform that reduces frustration and enables success

Not all chatbot platforms are equal. There’s a huge gap in features and capabilities even among market leaders.

Here are important features to look out for:

1. Natural Language Processing (NLP) capabilities. Sometimes, the inability of a bot to understand human requests is down to poor NLP, rather than the team that’s implemented it. You should have full control over the language flow, including setting intent, entities, conversation responses and detection types.
2. Integration with data sources and API capabilities. If your chatbot can’t talk to basic internal systems – for example, order, booking and shipping – your bot ends up being a virtual and very expensive Magic 8 Ball. 
3. Context. The ability to store and recall crucial contextual information provided by users is what drives the best engagement and experience. To that end, if a bot repeatedly asks the same questions, it will frustrate the user. 
4. Agent handoff. Accept the fact that no chatbot can answer every single question. It’s essential to have a clear point and process for handover to a human, delivered through the same chat window.

ChatGPT has entered the chat

Of course, it’s not possible to talk about chatbots and AI-driven anything without ChatGPT factoring into the discussion. With its seemingly mind-boggling ability to answer any question you throw at it and ask follow-up questions to try and get to the bottom of what you’re asking it, ChatGPT seems like the future of chatbots, and everything else if you believe the hype. Though some extremely clever people aren’t getting swept away with the wave of excitement.

Wherever you stand on it, there’s no denying that ChatGPT’s incredible ability to pull together information, delve deeper and probe your needs, and formulate clear and accurate responses (a lot of the time) will make it extremely powerful as a chatbot on a site. 

Handling routine (and even more complex) questions, directing users to resources, and even escalating contacts to an actual person, are all things the platform could handle with relative ease. And the naturalness of the interactions should yield positive experiences for users, and offer the opportunity for personalising those experiences as well. 

Of course only time will tell how it all plays out, but the possibilities seem impressive. 

The future is now and it’s all about data

We regularly see businesses using multiple systems and logins, often separating support customer accounts from their main website account. Sometimes, there are different chatbot platforms for each department, and none of the information is shared or accessible.

So, customers have to repeat info again and again to what they assume is a single system, leading to a disjointed and jarring experience.

What if you could have a chatbot that takes care of a customer through their entire lifecycle, in some cases potentially being the only touchpoint? With the right data, that’s possible. 

A customer could make first contact with a brand through a chatbot that notices they’re browsing the pricing page, and it pops up, asking cost-related questions. The person provides their details which are passed into the CRM. If they make an order, they receive shipping updates through the same chatbot. If there’s an issue with the product, the bot helps them troubleshoot. And in the event they’re not satisfied with the item, the bot helps process a return and refund.

The key here is sharing data between systems. The bot needs to identify a sales lead, store it in the CRM, and access the order system and support documents. The easy part is then taking and testing what you’ve learnt from making super bots and applying your user stories to the process.

Chatbots are the final and potentially most powerful piece in the marketing automation stack. With the right implementation, they reduce workloads and increase positive customer experiences.