Chatbots have received a lot of attention and a lot of media coverage in recent years. From a corporate perspective, they hold great promise – which companies would refuse to consider automating actions currently performed by humans? However, the benefit to users is more rarely discussed: do they really enjoy interacting with a robot when they need help?
Because of companies’ enthusiasm for chatbots, it is sometimes difficult to determine what they are actually capable of doing compared to what would require advancements in the underlying technologies.
We will therefore start by presenting the benefits that a company can hope to obtain with a chatbot as well as what they require. We will then consider a very specific use case: customer support, and semi-automated alternatives to chatbots. Finally, we will conclude on a very different use case for bots: the services offered by SMS. Let’s go !
When to use a chatbot?
Before you start to see chatbots use cases, you should take a step back to consider how to determine whether their use is relevant or not in a given situation.
Indeed, it is good to take some precautions before implementing a chatbot to avoid situations where a badly designed bot meets a need … which does not exist.
Improved user experience
Chatbots that work are pleasant to use and it shows: they have real users who appreciate them. To arrive at this result, it is good to start by rigorously analyzing the current scenario of your situation to compare it with that which would involve a chatbot for the same effect. This will help you accurately determine the actual contribution of the bot. Here are some questions to feed your thinking on the subject:
- How is the user currently solving their problem?
- Is it possible to solve the same problem faster (fewer steps) or easily with a chatbot?
- On which channel do you plan to deploy the chatbot? Are your users already there?
The real cost of the chatbot
From a corporate perspective, chatbots are cost-saving machines. Once you are convinced of the relevance of a chatbot for a given situation, you still have to determine its cost. It is also worth thinking about the division of labor between the chatbot and the (human) agents: fully automatable situations are rare. Human intervention is often necessary to guarantee an acceptable result. Here are some helpful questions to help you understand the issue:
- What resources are you currently spending on the problem you want to automate?
- Are they enough to be worth the time and resources needed to develop a chatbot?
- How much of the workload will be taken up by the bot? Can you estimate the workload that your agents will have left?
Chatbots for customer support
Customer support is one of the textbooks for chatbots. Indeed, the interests are obvious as much for the company as for the user.
From a user perspective, this means that customer support is available at all times and will respond instantly. In addition, since the support is automatic, this removes the fear of asking the most basic questions: a robot cannot waste its time.
From a company perspective, chatbots are reducing the number of agents needed to run customer support. Depending on the implementations, the bot can also be used to categorize and sort incoming requests. This then optimizes the agents’ schedules and improves the accuracy of the statistics. Finally, as the chatbot is deployed on an existing channel (help center or messaging) this greatly facilitates its adoption.
Support process with a chatbot
Before we go any further, you may be wondering what customer support looks like using a bot? Here is a simplified version:
- The customer initiates a request for assistance (“ticket”): A customer has a problem with the product or service and is looking for help. He goes to the usual channels (the help center for example) and starts a conversation with the bot.
- Ticket categorization : Thanks to natural language recognition (NLP), the bot is able to recognize the main subject of its request. It is categorized, for example: delivery, payment, problem with the site…
- Immediate assistance: If the problem is one of the most common, the chatbot may have an answer ready. Otherwise, he can offer the user different supports to help him, such as guides.
- Followed : With targeted questions, the chatbot is able to know whether the problem has been resolved or not.
- (Optional) Reporting to an agent: Before handing over to an agent, the chatbot may collect the information necessary to process the ticket – such as details of the problem or the customer’s settings.
- (Optional) Agent response : The agent can then regain control. He has access to the conversation history and the categorization carried out by the bot, which facilitates and accelerates the processing of the ticket.
Prerequisites and limits of a customer support bot
The procedure presented above is rather fluid and effective if it is well implemented. There are however several clarifications to be made. Indeed, to achieve such a result, you need a substantial already existing ticket base that can be analyzed. In addition, the more tickets are similar to each other, the more effective the bot will be. Finally, the development and maintenance of such a chatbot requires technical skills internally.
With these limits specified, we realize that this use lends itself rather to B2C companies that have many users. Companies with more B2B products will have an interest in moving towards other solutions.
Automation with Interactive Voice Servers (IVS)
B2B businesses tend to have more modest user bases as well as significant differences in the configuration or even the installation performed for each customer. A chatbot would therefore not be able to recognize problems effectively. However, options remain to allow these companies to partially automate their customer support.
There are still many companies that manage support directly by phone: it has the merit of being simple and direct for customers. For manage the incoming call flow efficiently, they can use SVI (Interactive Voice Server).
IVR allows you to orient your customers through voice interactions or using the keys on their phone. The system plays pre-recorded messages to guide them. An SVI allows you to collect information about your callers. With this data, you can then prioritize important customers for example (by reducing / eliminating the waiting time) or even optimize the routing of incoming calls.
Gefficiently manage your incoming calls with CALLR
Let’s finish our overview of chatbots and other automation techniques with a very specific type of bot: SMS chatbots.
Chatbots by SMS
Chatbots are often considered in messaging applications (Messenger, Whatsapp, Telegram…) but they can also work by SMS and this has serious advantages. The first and most obvious is of course accessibility. Not everyone has Whatsapp, but almost everyone has a cell phone that handles SMS – over 5 billion people worldwide can receive text messages. Indeed, no smartphone is needed for SMS, even the most basic phones support them. Chatbots by SMS thus make it possible to offer intelligent services on very simple telephones like a Nokia 3310.
TextEngine is the perfect example. As its name suggests, this service offers an SMS search engine. It’s sort of like a portal to the internet that doesn’t require internet access yourself.
Go further with chatbots by SMS
SMS are particularly suitable for notifications: for tenir a customer informed about the status of his order for example. Text messages are read very quickly: 90% of them are opened within 5 minutes of sending them. Nevertheless simple SMS notification can sometimes be frustrating. You can go further by letting your users respond, this is the purpose of chatbots by SMS.
For example, an SMS chatbot can improve your delivery notifications and optimize the whole process:
Your ABC package will be delivered to your home 25 rue de la Paix in Paris tomorrow between 2 and 4 p.m. Respond OK to confirm delivery or reschedule to change the date.
If the customer reschedules the delivery, the chatbot may offer, for example, three other ranges that the customer will confirm with a simple 1, 2 or 3.
As the interactions in the scenario described above are very well framed, developing such a chatbot is easy. The process is nevertheless quite flexible and applicable to other scenarios such as appointment reminders.
Thanks to our partnership with Recast.AI, a natural language recognition service that helps in the creation of chatbot, you can quickly set up a similar chatbot.
Democratize access to your services with an SMS chatbot
Learn more about chatbots
If you want to learn more about chatbots, we have already spoken several times on the subject, in English. Our first article in 2017 dealt with the emergence of SMS chatbots. To better understand the technology that made chatbots (NLP) possible, we invite you to read Chatbot Masquerade our analysis which presents the challenges of programmatic understanding of language.
Article written in collaboration with CALLR