A guide to conversational marketing chatbots

Conversational marketing uses a combination of human conversations, which could take place on voice channels, via live chat, or messaging app, and interactions orchestrated by AI that could take the form of chatbot interactions or triggered messages that provide customers with the information that they are looking for, immediately and in the right context.

Indeed, contextually appropriate communication is one of the key attributes of conversational marketing. With the goal of removing friction from the customer journey it enables brands to always provide the right information, at the right time, taking into account all previous interactions and data gathered to remove repetition and make the customer feel understood and valued.

Conversational marketing has become established in recent years as one of the most effective ways of building customer relationships and enabling people to discover your products and services at their own pace, whatever the stage of the customer journey.

Using the right conversational technology, it shouldn’t matter what channel the conversation takes place on – just like chatting with a friend on voice, email, messaging app or social media they will always remember your last interaction and you can pick up where you left off without skipping a beat.

The ability for a brand to do this with all its customers, all the time, across all channels is why conversational marketing is so successful.

The growth of easily accessible chatbot building technology has been of the key drivers of the growth of conversational marketing. Chatbots are perfect for many use cases that depend on the immediacy and 24-hour availability that chatbots provide.

These days they can also be deployed on any digital channel including web, WhatsApp, Messenger, and more. Instead of looking for a helpline number many customers first look for a chatbot link so that they can get their query resolved immediately without having to navigate an IVR menu and wait in a queue.

Whether they realize it or not, people intuitively understand the 80/20 rule and know that for most of their queries they can get a quick reply from a self-service chatbot that has been trained to provide answers to the company’s most frequently asked questions. For the other 20% they will generally be happy to hold for a human agent. To be honest, with the rapid advancement of chatbots and tools like ChatGPT, we might be calling it the 95/5 rule soon!

In the meantime, let’s look at some of the ways that you can start conversations with potential customers, and some of the use cases that chatbots are ideally suited to.

For marketing use cases, you should be aware that in most countries there are restrictions on how conversations can be initiated on digital messaging channels.

While it is perfectly acceptable for a web chatbot to politely ask if it can be of any help when someone is browsing your website, a chatbot cannot initiate a new chat with a person on messaging channels. Companies like Meta and Viber have strict policies in place, so it is a good idea to familiarize yourself with these before you start building out your conversational marketing use cases. For example, the rules for WhatsApp are explained here.

Brands are finding all sorts of interesting and effective ways to use chatbots in their marketing. This is driven by two trends – firstly chatbot technology is advancing quickly, so they are smarter and more flexible than ever before. Secondly, people are getting more and more used to dealing with chatbots and now trust that the experience will be positive.

When someone has a need but doesn’t know exactly how to satisfy it, then a chatbot can be a great way of helping them to research the subject and find the best products to help.

Say I’m a budding YouTuber and I want to upgrade my sound recording equipment. I walk into a technology store to be faced with a whole wall of microphones and audio devices, but I have no idea which one is right for my setup and budget. In store I can ask a shop assistant for help but trying to make the same purchase online would be tough – unless there was a virtual assistant that could help me hone in on the ideal mic after a asking a few basic questions.

Chatbots are perfect for this use case as they can very quickly process huge amounts of information in milliseconds to come up with a shortlist – that they can then display in an engaging and dynamic way, for example using an interactive image carousel.

What’s more, store opening hours aren’t an issue for chatbots and they can keep on helping to drive sales all through the night and weekends – when budding YouTubers are most active.

When customers sign up for a service online there may be multiple paths through the process, depending on their personal circumstances and information. Rather than presenting an overwhelmingly detailed web form covering all scenarios, use a conversational chatbot to guide people through the process, asking for small amounts of information at a time, and answering questions when anything is unclear.

When integrated into a customer engagement platform, chatbots can also be deployed to reduce drop off rates and re-engage with people that have dropped out of the process. The ride-share company Bolt did this for their driver sign up process, increasing conversion rates by an amazing 40%.

It’s one thing attracting customers to your service business, but if they don’t get up and running smoothly and start getting value from your service then they are in danger of abandoning you.

Chatbots are excellent at providing a hand holding service as customers explore your platform and start getting the hang of it. No-one reads manual these days, do they? A chatbot can drip feed the right information at the right point in the process so as to not overwhelm new subscribers.

This is the sweet spot for conversational chatbots – people love to engage with them when they are well designed and offer something fresh and entertaining. The possibilities are almost limitless, and some unexpected organizations are using them from government agencies to sports teams. Infact Arsenal football club was one of the first to employ a chatbot to interact with fans. Robot Pires (named after Arsenal’s legendary mid-fielder Robert Pires) can be reached on Messenger, Skype, Telegram and Slack and offers the latest news about the club, fixtures, and is a font of knowledge about player stats and historical results. Need to settle a pub debate about a key result in the 2015 season? Robot will know!

Some other engagement use cases for chatbots that you may want to consider:

This is an important use case as elements of it can be built into all the chatbots that a business deploys – including those whose primary purpose is transactional or informational.

For example, AI product recommendations can be built into chatbot conversation logic to help customers find products that they may be interested in that could help to solve issues that they are reporting.

However, care needs to be taken that the offers are relevant and targeted to each individual and are not a distraction from the primary purpose of the chatbot. If the person is not interested, then they should be able to dismiss the offer and opt out of receiving more like it.

In a recent blog on the latest developments in chatbot technology we explained the difference between rule-based chatbots and intent based chatbots. In summary, a rule-based chatbot enables people to find the information they want by providing a cascading set of options to select from. Although this is not strictly a ‘conversation’ it is still a great tool for marketing use cases where a person wants to learn about the products and services that you provide, or simply find out where their nearest store is.

On the other hand, intent-based or conversational chatbots are designed to mimic a two-way conversation with another person. Indeed, with some highly trained conversational chatbots you may find yourself questioning if you are actually interacting with a human rather than AI.

Conversational chatbots use something called natural language processing (NLP) to analyze the textual inputs from the people interacting with them to work out what their intent is, so that they can provide appropriate and factually correct answers.

Conversational chatbots are more complicated and time consuming to create because every possible branch of the conversation flow needs to be designed to provide appropriate answers. This process is called conversation design.

The ability to accurately identify human intent is the key to a successful conversational chatbot. This in turn drives the ‘next steps’ that the chatbot makes, for example to trigger a particular conversation flow or a requested action.

The notion of an intent can be easily explained with an example.

Say that a person wants to find out how many miles they have accrued with their frequent flyer account.

This is a fairly clear-cut scenario, but the process is exactly the same for any use case. The only difference is the total number of phrases that the chatbot has to be trained with for each intent. It is not uncommon for a single intent to have over 400 phrases where it is more ambiguous.

In other words, what use case is the chatbot being created to satisfy. Be specific about what its capabilities should be, and just as importantly, specify what the chatbot won’t do. This avoids what is called scope creep – which slows down projects and causes them to drift away from their original purpose.

Conversational marketing chatbots are all about generating engagement and interest but this does not mean that you can radically depart from your brand identity and tone of voice. You want your chatbot to be trusted by customers and it should therefore be consistent with your brand in terms of its values and your customers perception of you. This will dictate:

Google has an excellent guide on creating a persona that you can refer to. We also think it is a great idea to give your chatbot an easily remembered name and an avatar that reflects its personality.

The script for a conversational marketing chatbot should look less like a linear timeline and more like a family tree – going back several generations and with lots of branches. This allows the conversation to be more free-form and less rigidly guided. At each branch in the flow the person interacting with the chatbot should have multiple options, so can therefore take whatever route they like to achieve the required outcome.

Always refer back to the original goal that you specified to make sure that you don’t waste time over-engineering low value side quests that people might try and explore.

Now is the time to get into the detail of creating all the phrases and responses that the chatbot will use to respond to the intent of the people interacting with it. This should be significantly easier if you have done a good job of specifying the chatbot’s persona and mapping out the conversation flow.

A few helpful pointers:

The testing phase is not a box ticking exercise that you can skimp on. You need to test every possible branch of your conversation flow to ensure it make sense and it contributes to achieving the goal of the chatbot.

Where you have a very large number of intents and complex logic, there are specialized tools that can help you to automate some of your testing. However, nothing is better than using real people, ideally from different backgrounds and of different ages. These could be trained testers, a sample of your target audience, or even members of your staff that haven’t been involved in the design and build of the chatbot.

Thorough testing will identify flows that might need modifications or additional phrases and words that you need to cover.

Here is just a small selection of chatbots built by Infobip customers using our chatbot building platform.