Digital Transformation Today

Understanding Bot Design Language and Processes

With the rise of artificial intelligence and machine learning, the use of intelligent bots in the workplace will soon become the norm. So how can businesses ensure their customers (and other users) have the best — and most human — experience? By understanding the intricacies of bot language design.

What Is Bot Design?

Bot design, more specifically bot language design, is the process of designing a network of scripted conversations for an intelligent bot to follow. The primary goal of the best AI bot designers is to have any and all conversations mimic human interactions as best they can. 

Natural Language vs. Menu Driven Design

Simply speaking, there are two popular kinds of AI bot design being used today — Menu Driven Design and Natural Language Design. Menu Driven Design is ordered and structured, providing a list of responses (options) for users to choose from during their interactions. The more human approach is Natural Language Design. Natural Language design involves understanding the psychology of how users ask questions and look for information, in order to deliver an open-ended conversational experience. Users should be able to successfully interact with the bot interface as if they were speaking directly with another human being. 

Using Bots for Business

It’s important to map out how a bot is going to be used within a business, because that directly impacts how the intelligent bot will be designed. Most often, bots are created to save time and effort by relieving employees from having to answer repeated questions and perform mundane search-and-retrieve tasks. By utilizing bots, companies can have their employees focus on revenue-driving tasks, or on tasks that require human intervention.

Diving Into a Human-Centered Approach to Bot Design

To be able to design human-centered bot interactions, there are a few steps that must be taken:

Step 1: Understand the client’s current state & business processes

Being familiar with a client’s existing infrastructure and uncovering pain points is arguably the most important component to bot design. For any bot deployment to be successful there needs to be visibility into request channels (email, phone, chat, ticketing system, website forms) in order to:

  • Understand how the underlying technical infrastructure works
  • Help determine bot use across channels
  • Pinpoint specific problems and challenges

Bot deployments are most successful when there are trackable goals to hold the bot accountable to — as in 35% of future help desk requests will be handled by the bot freeing up time for employees to focus on other work. Goals can take various forms, but often include a specific number of requests to be handled per day/month/week.

Step 2: Define use cases & logical branches based on Step 1

Once the challenge(s) the bot will be helping to address have been determined, the next step is mapping out all the logical branches, or paths each interaction can take. This is best done through a series of brainstorming sessions, in which an analysis is conducted on how to keep the user focused, streamline processes, and deliver information as quickly as possible. This is where Natural Language Processing and Design come into play. It’s important to know and understand how people speak and ask for information so the bot’s conversation abilities can be designed accordingly. Some ways to make this step easier include:

  • Identify FAQs
  • Use third-party tools like Visio/Excel to define question & answer paths
  • Engage in resource gathering & mapping activities like cataloging all resources, videos, tutorials, etc. 
  • Determine whether people should take action inside or outside the bot interface

Step 3: Categorize branches based on lexical commonalities

Once conversation and interaction branches are catalogued and defined, it’s best practice to categorize them based on commonalities. This ensures that we are providing accurate and consistent bot responses for similar questions. All conversation paths must be analyzed; grouping together similar interactions that lead to the same end-points. It also requires a deep understanding of how common keywords, slang, and the vernacular impact bot language design. To provide an example; all paths containing the keyword “leave” could be bucketed into categories of Medical/SICK, PTO, Short-term disability etc. 

Step 4: Determine the type of language to be used

As mentioned above, there are typically two types of bot language design, Natural Language Processing and Menu-Driven Design. To deliver the most valuable and engaging experience, it’s best to rely on Natural Language Processing (NLP), or a combination of NLP and Menu-Driven design. This gives designers greater flexibility to provide a more human experience. Once the type of language is determined, consider:

  • If the bot language will be technical, or more informal
  • The personality of the bot (professional, friendly, witty, enthusiastic etc.)
  • How the bot will set expectations from initial interaction
    • Direct, Vague, Menu-Driven
  • How the bot will conclude interactions
    • Contact page, feedback survey, open-ended

Choosing a Bot Developer

Vetting and selecting a bot designer can seem like an overwhelming process, but it’s easy to narrow down the choices. To provide customers and employees with the most human-like bot interactions possible, choose a partner like Withum who has both the technical expertise and UX background needed to create truly powerful bot experiences. Our Human-Centered bot design approach is a collaborative effort between our consultants, designers, UI/UX teams, and our clients — leading to the creation of transformative solutions with a direct impact on business productivity. 

In our next piece on bot design, we’ll dive into some of the fascinating tools that are used to setup and deploy intelligent bots. Stay tuned!

The second piece of this series discusses tips and tricks in deploying the chatbot.
Click here to read part two.

Previous Post

Next Post