It’s no secret that chatbots can be a great asset to a business that is looking to streamline their Sales, Service or Marketing processes. A highly advanced and well-thought-out chatbot can be impressive, but what might be seen as well-meant sophistication can quickly turn into tediousness and inefficiency. Making the right call when deciding the type of bot you want to develop will be crucial to guarantee success. Today, I want to take you through some of the things that should be taken into consideration when determining the right type of chatbot for your use case. 

This blog is a part of a series on Einstein Bots that Cloud Innovation will be releasing over time. In the first instalment, I went over how chatbots can greatly improve company processes and improve customer experience and satisfaction. 

Interested? You can check it out here

Scripted Bots

Traditionally, chatbots have been quite a bit less intelligent than some are today. This type is commonly referred to as a ‘Scripted’ or ‘Flow-based’ bot. Namely, because all interactions are entirely pre-defined by humans. Think of it like going through a guided setup flow or installer wizard on your computer. You get a set amount of actions and options to adjust, usually through a menu like structure, instead of naturally talking to the bot to accomplish your goal. 

They use clickable menus and buttons to navigate the flow. Scripted bots are hardcoded, they are unchanging. In turn this means that you accomplish the same result every time you use it, granted the options and parameters are the same. While this might result in a slightly less dynamic experience, they’ve proven to be almost foolproof given enough testing before deployment.

Natural Language Processing

Natural language processing (NLP) is a subset of Artificial Intelligence technology that enables computer software to analyze human speech input, also called natural-language. This technology is most often used to allow for convenient and foolproof human-machine interactions.


In NLP lingo, Intent is specified as the end goal the user wants to reach by interacting with the software. They can be statements with the goal  to invoke direct action such as “Send an email to John Doe” or questions and vague statements that the chatbot will need to deduct the intent from. A good example of this would be “My roof is leaking”. Even though it might not seem obvious  what action should be undertaken at first glance, NLP software is usually able to deduct what it should do just like a human would. In this case, referring the user to an article in the knowledge database or trying to set an appointment with a service worker could be the right course of action. To detect which action should be executed, the system uses a process called intent classification.


Now that we know that Intent specifies the end goal that a user wants to reach, which is usually defined as an action that should be executed by the chatbot i.e. schedule an appointment. It is also important to know about Entities. They can be defined as data that accompanies intent. We can make use of our example earlier to clarify what entities are. The intent there was to send an email to John Doe. What if we added some data to that expression: “Send an Email to John Doe with the subject ‘Dinner tonight at 8pm?’”. NLP is capable of understanding that there are two different entities in that statement. First of all we have our contact ‘John Doe’, which defines who the email should be sent to. Secondly, we added a subject that should be used for the email itself.

The example above is a pretty advanced combination of intent and entities that is usually reserved for products like personal assistants, nevertheless, it illustrates the difference between intent and entities quite well. In essence, intent is an action you want the software to undertake or an end goal you want to achieve. While entities are contextual data that is related to the intent.

NLP Based Chatbots

Quite recently, NLP-based chatbots have started to gain a lot of popularity as the technology continues to be improved. These intelligent bots use a trained Natural Language Processing model to correctly guide users to the expected end result instead of flowing through a menu like structure. Furthermore, they also have the ability to pick up on contextual information from the conversation to reduce the strain on the users and make conversations more natural.

Let’s say you want the chatbot to register the user’s phone number. What you normally ought to do, is set up a question in your conversation flow that will ask the user for the required data. But what if they've already mentioned their number during their conversation with the bot? Intelligent bots can naturally recognize if certain information already exists in the context and pick it up for later usage. This also hooks into Intent and Entities. Chatbots using a platform like Einstein Bots make this easy by allowing the developer to suggest possible patterns and structures in data that the bot should be looking for.

Which one should you use?

So far, it might sound like scripted bots are not even worth considering, but this is simply not the truth. There are pros and cons to both types, which are important to highlight and take into consideration.

The rigid nature of scripted bots might seem like a downside, but it’s actually a blessing in disguise. First of all, because the conversation and possible inputs are limited, there is little chance of the bot becoming ‘confused’ and thus not being able to complete a task. Secondly, this approach might be more comfortable or less confusing to users who are not as technologically inclined. Improving accessibility leads to higher adoption rates as well.

NLP-chatbots are more dynamic and feel more natural by using artificial intelligence to determine the intent of the user and conversational context. This style of bot is similar to a natural conversation with a human. If the human was great at keeping things simple and concise that is! Though, owing to the fact that they are more dynamic, problems and confusion tend to creep up more often. This can usually be partially alleviated by providing the bot with more testing data to improve its accuracy.

Want to know more?

Interested in building a chatbot using Sales Einstein Bots? Having trouble deciding on what kind of setup you want to go for? Cloud Innovation would be delighted to provide you or your company with guidance on setting up your own chatbot.

You can contact us here.