Highlights from 2023 User Conference
Our chatbot experts shared tips and tricks for chatbot authoring during three sessions in the 2023 Mobile Coach user conference. In these sessions, we covered chatbot writing best practices, measuring success, and building a chatbot with a knowledge base. Below is a recap of each chatbot authoring session and the recordings.
Chatbot Writing Best Practices
The secret to your chatbot project’s ultimate success? It’s actually not in the technology but rather, in the quality of the writing! After all, your end-user is reading each message to determine whether the chatbot feels valuable.
Chatbot Authoring Framework
- Modality: The chatbot channel best suited to reach your audience considering ease of access and security requirements.
- Content: Your chatbot script or library of messages.
- Context: The rules that power chatbot logic to ensure the right messages are sent at the right time to the right users.
- Goal: What problem is your chatbot solving and what metrics do you need to prove it?
Set the big picture
- Who is your audience? How will they access your chatbot?
- What is your chatbot value proposition?
- Describe the user experience:
- Is it push or pull or both?
- What is the nature of the content?
- How long is the user experience?
Understand the interactivity plan
There are three types of messages your chatbot can send. We recommend creating a user experience with a mix of these pushed messages. The right mix is designed with the user’s needs and the business case of the chatbot in mind.
- Non-interactive: This is a message sent by the chatbot that does not ask a question. An example of this could be a reminder, a resource, or a motivational message.
- Multiple choice: There are many different ways to ask a multiple choice question (ABC, Likert, True False, Fill in the Blank, etc.). The benefit of asking a multiple-choice question is you know the user’s response. That allows you to create different chatbot answers based on the predicted user responses. These types of questions are also typically easier for users to answer and so see higher response rates.
- Open-ended: With an open-ended question, there is no way to predict a user response. These types of questions require more thought from a user and so often see a lower average response rate. However, having a user reflect and provide a thoughtful answer can be an important part of a chatbot conversation. Since there is no way to predict a user response, your following chatbot answer must be generic and acceptable for any answer. We suggest acknowledging that they responded and then pivoting or providing universally applicable details.
Prepare to onboard users
Joy from Dell put it perfectly when she said: “You can have a great script, great chatbot, but no one knows about it.” Some ways you can incorporate this in your larger chatbot authoring design plans are:
- Pitch the value (Roula from IC Axon shows an example of that here)
- Invite in-person/in-session
- Message in a high-priority channel
- Set expectations for user experience
Measuring a Chatbot's Success
One of the most frequent questions asked by organizations looking to implement a chatbot is, “How do we measure success?”
Chatbots are incredibly fast, scalable, smart, and efficient. But organizations still need to measure just how much a chatbot benefits them.
Chatbot Measurement Framework
Here is a basic framework to use to think about putting together your ROI strategy.
- Define success metric: You want to define success upfront, so you’ll know it when you see it.
- Identify data collection plan: Any ROI plan will require data or inputs to measure the results. You need to identify your data collection plan in advance.
- Compute program cost: Think of this step as, “How much am I spending to buy my success metric?”
- Create business plan: This can be formal or informal but is vital to success. We recommend you create an “elevator pitch version” and a fuller, more robust version.
- Multiple choice
- Links to content
- Subscription rates
- Opt-out rates
- Response rates
- Response times
- Link click-throughs
- Video views
- Knowledge base inquiries
- Support tickets
- Qualitative feedback
Create a Knowledge Base
One of the most common features for chatbots, in general, is to have it answer a user’s question. But how do you do this for the workplace? How do you design a chatbot to answer specific questions an employee might have while on the job? The answer is a knowledge base!
Why a knowledge-base chatbot?
- Your audience is not getting the answers they need
- It takes too many clicks
- A person is involved
- Ready answers don’t exist
- Scaling is needed
- Existing solutions aren’t working
- Getting info at users’ fingertips (frictionless)
Try it out:
The three chatbot authoring categories for a knowledge-base chatbot are:
- Conversational flow
- Branching logic with narrowing questions
- Additional actions
- A large library of FAQs
- Follow-up conversation isn’t needed
- Transactional experience
- Pull information from another system
- Existing system has API end-points
Common messaging categories:
- Job Aids
- Talk to HR