We recently decided to add an AI-enabled chatbot to Blanchard’s most popular leadership development program, SLII®.

We did so with a clear purpose in mind: to help our students transition from a positive training experience to the world of work. You see, many students finish an SLII® class eager to use their new knowledge and skills, but the workload and bad habits get the better of them and a month later they find that their SLII® skills are rapidly declining. We believed that a chatbot with artificial intelligence could help them overcome this obstacle through a combination of prompts, tools and assistance at the right time. So we set out to create SLII® Chatbot™.

Together with our platform partner Mobile Coach, we got down to work.

representation user experience interface design

Prepare for two types of interactions

Chatbots can be structured or unstructured. A structured interaction is planned in advance, with written instructions to guide users. For example, we have all had the experience of “dialing or saying 1” when calling a toll-free number.

Unstructured dialogues are not written in advance. ChatGPT is a good example of this type of interaction, where an AI system is trained using a Large Language Model (LLM) to interact with humans using our natural language. We wanted both a structured and unstructured system in our SLII® Chatbot.

The structured part of the chatbot would focus on helping trainees through reminders and assistance for one to three months after their training. We have drafted a dialog and questions for the bot. We use preformatted answers; for example, “type 1 for tools, type 2 for tips”. Our structured script ended up being about 50 pages long. The most important thing we learned was to be clear about the objective, allow enough time for development and testing, and combine information with entertainment.

We also noticed that text messages use a completely different writing style and syntax including acronyms (e.g., lol, btw), use of emojis, lack of punctuation and capitalization, etc. We have chosen to adopt them in the SLII® chatbot. It is helpful to be clear about writing style guidelines from the beginning.

Entering new territory

Creating the structured dialogue was a lot of work, but it was work we were familiar with, as we had done a lot of scripting over the years. The unstructured and conversational part of the chatbot was a new field.

Our goal was to help students with an SLII® chatbot that could answer any question. This required an intelligent AI engine based on an LLM.

Once again we relied on our partner, Mobile Coach, who offered a private LLM so that we could help our clients securely without having to transmit our proprietary content to the public ChatGPT system.

We wanted our SLII® Chatbot LLM to be an expert in SLII®, so we started by feeding it with existing documents from our SLII® program, facilitation notes and supporting materials. The LLM was very flexible in its ability to process documents, having learned it by absorbing a wide variety of content on the Internet.

Designing messages

The next step was to design the ads. To understand an LLM notice, think of a sandwich. A user could type a question to the bot along the lines of “I just had a disagreement with a team member, what should I do?”. Think of it as the center part of the sandwich. Even if the LLM is able to answer that question on its own, the answer will be much more useful and controllable if you include the question in a broader question. Let’s think about adding bread, lettuce, tomato and mayonnaise to the sandwich.

Suggestions can be used for many things, but we have focused on context and writing instructions. The context is to provide the background to the question so that the LLM can better understand what is happening. This can take the form of descriptive text that explains what the user is doing or trying to accomplish. An example is “I asked an employee how his one-on-one conversation with a team member went and he commented: [respuesta del usuario]”. It’s like when we provide context when a new person joins an ongoing conversation, “Oh, we were just talking about the merger announcement.”

The writing instructions you provide to the LLM in these prompts can cover many aspects of writing. Whether you know it or not, you probably have an idea of what appropriate and inappropriate responses look like. I remember one of the first responses that was too long, so we added a length instruction to reduce the LLM responses to that length. You can also tell the LLM the tone you are looking for and where to extract the content from. Here is an example: “Please write an optimistic response that validates your struggle, is less than 100 words and extracts at least one concept from the following content: [lista de fuentes]”.

Attention and care during the first days

Combining a wide range of content sources, providing context and background, and defining the tone and style of written responses will get you off to a good start. The last tip is to invest in care and nutrition. LLM chatbots can easily be improved and enriched over time by reviewing bot performance and working on interactions that are not serving users well.

It is still too early to talk about LLM chatbots. We are very excited to be able to experiment with this technology for such an important purpose. We strongly recommend that you try it if you can spare the time. We hope you are doing well.

Note: This article is a translation of the original article by Dr. Jay Campbell, Blanchard Product Manager.

Share this news! Choose your platform.

We recently decided to add an AI-enabled chatbot to Blanchard’s most popular leadership development program, SLII®.

We did so with a clear purpose in mind: to help our students transition from a positive training experience to the world of work. You see, many students finish an SLII® class eager to use their new knowledge and skills, but the workload and bad habits get the better of them and a month later they find that their SLII® skills are rapidly declining. We believed that a chatbot with artificial intelligence could help them overcome this obstacle through a combination of prompts, tools and assistance at the right time. So we set out to create SLII® Chatbot™.

Together with our platform partner Mobile Coach, we got down to work.

representation user experience interface design

Prepare for two types of interactions

Chatbots can be structured or unstructured. A structured interaction is planned in advance, with written instructions to guide users. For example, we have all had the experience of “dialing or saying 1” when calling a toll-free number.

Unstructured dialogues are not written in advance. ChatGPT is a good example of this type of interaction, where an AI system is trained using a Large Language Model (LLM) to interact with humans using our natural language. We wanted both a structured and unstructured system in our SLII® Chatbot.

The structured part of the chatbot would focus on helping trainees through reminders and assistance for one to three months after their training. We have drafted a dialog and questions for the bot. We use preformatted answers; for example, “type 1 for tools, type 2 for tips”. Our structured script ended up being about 50 pages long. The most important thing we learned was to be clear about the objective, allow enough time for development and testing, and combine information with entertainment.

We also noticed that text messages use a completely different writing style and syntax including acronyms (e.g., lol, btw), use of emojis, lack of punctuation and capitalization, etc. We have chosen to adopt them in the SLII® chatbot. It is helpful to be clear about writing style guidelines from the beginning.

Entering new territory

Creating the structured dialogue was a lot of work, but it was work we were familiar with, as we had done a lot of scripting over the years. The unstructured and conversational part of the chatbot was a new field.

Our goal was to help students with an SLII® chatbot that could answer any question. This required an intelligent AI engine based on an LLM.

Once again we relied on our partner, Mobile Coach, who offered a private LLM so that we could help our clients securely without having to transmit our proprietary content to the public ChatGPT system.

We wanted our SLII® Chatbot LLM to be an expert in SLII®, so we started by feeding it with existing documents from our SLII® program, facilitation notes and supporting materials. The LLM was very flexible in its ability to process documents, having learned it by absorbing a wide variety of content on the Internet.

Designing messages

The next step was to design the ads. To understand an LLM notice, think of a sandwich. A user could type a question to the bot along the lines of “I just had a disagreement with a team member, what should I do?”. Think of it as the center part of the sandwich. Even if the LLM is able to answer that question on its own, the answer will be much more useful and controllable if you include the question in a broader question. Let’s think about adding bread, lettuce, tomato and mayonnaise to the sandwich.

Suggestions can be used for many things, but we have focused on context and writing instructions. The context is to provide the background to the question so that the LLM can better understand what is happening. This can take the form of descriptive text that explains what the user is doing or trying to accomplish. An example is “I asked an employee how his one-on-one conversation with a team member went and he commented: [respuesta del usuario]”. It’s like when we provide context when a new person joins an ongoing conversation, “Oh, we were just talking about the merger announcement.”

The writing instructions you provide to the LLM in these prompts can cover many aspects of writing. Whether you know it or not, you probably have an idea of what appropriate and inappropriate responses look like. I remember one of the first responses that was too long, so we added a length instruction to reduce the LLM responses to that length. You can also tell the LLM the tone you are looking for and where to extract the content from. Here is an example: “Please write an optimistic response that validates your struggle, is less than 100 words and extracts at least one concept from the following content: [lista de fuentes]”.

Attention and care during the first days

Combining a wide range of content sources, providing context and background, and defining the tone and style of written responses will get you off to a good start. The last tip is to invest in care and nutrition. LLM chatbots can easily be improved and enriched over time by reviewing bot performance and working on interactions that are not serving users well.

It is still too early to talk about LLM chatbots. We are very excited to be able to experiment with this technology for such an important purpose. We strongly recommend that you try it if you can spare the time. We hope you are doing well.

Note: This article is a translation of the original article by Dr. Jay Campbell, Blanchard Product Manager.

Share this news! Choose your platform.