6 Tips for Designing Employee Training via a Chatbot

Ever since we created a chatbot-based product management tutor Eddy, I thought we could help aspiring creators by sharing 6 Key Learnings that we've taken away from the process.

The takeaways are the result of our experimenting and also the feedback from the user interviews – more than 2000 people tried our chatbot-based Product Management training. We then interviewed 30 of the most active users (read: happy users) and 10 users who abandoned the product pretty quickly (read: indifferent or unhappy users).

Without further ado, let’s see what you should think about in order to setup training or onboarding via a chatbot.

1. Less is More

Split long paragraphs of text into short lines, preferably 1-2 lines on the mobile device screen. It improves readability and looks like texting in messenger.

2. Set the Tone

Think about your audience – what conversation tone fits your audience the most? Once you have that figured out, use it consistently in all chatbot messages. Serious, sarcastic, funny, engaging, optimistic – select one and stick to it. Remember to use emojis and keep in mind that emojis should match the tone of the conversation.

3. Prompt the User to Act

To make chatbot-based training effective and engaging, it should feel like a two-way conversation. The best way to make it feel like a conversation is to give the user a chance to "speak". After the learner sees a couple of messages popping up on the screen, (keep your message not longer than 1-2 lines), it's the right time to react.

Create a few simple prompts in the admin system:

Check out how it looks on the user's side:

"Sure" and "hmm...not really" are two different reaction types, which aim to catch the mood of the user. You can think of more reactions and link different reactions to corresponding "branches" of the decision tree in the chatbot.

4. Make it Visual

Leverage external resources – videos, pictures, articles, podcasts, etc, and embed them into your chatbot. Keep the balance right – try to keep the "Text – Not Text" ratio around 70% – 30%. Remember, videos, pictures, and podcasts serve to supplement the text, but not to replace it.

It's good practice to ask the learner if she has headphones before playing a video or a podcast. For example, the learner might be on public transport with no headphones, and switching from text to video may be inconvenient and annoying.

Create a simple Yes or No prompt in the admin system:

Check out how it looks on the user's end:

If the user is okay to continue with the video - go for it. Otherwise, continue with the text. You may want to create a decision tree and prompt the user to continue with the video or text only.

5. Summarize, summarize, summarize

Prepare the content summary of the articles, videos, and podcasts that you share with the learner.

If you plan on including external articles in your chatbot, it's great to show that you've done your homework. Provide a summary of the article, video, or podcast before sharing the link to the external resource.

A summary is your courtesy to the user – what if she doesn't want to read the article you are about to offer – by summarizing you show that you're mindful of the user's time. Briefly mention what the article is about, add a short line on the author and or just give a short teaser of the story, which will motivate the user to read further.

Free resources such as Resoomer or Smmry can help you to summarize and get the gist of the text quickly.

Finally, an example of the article – Net Promoter Score Definition – and a short summary:

"Developed by Bain & Co, a Net Promoter Score provides companies with a simple and straightforward loyalty metric. A Higher Net Promoter Score indicates a healthier business, while a lower Net Promoter Score is an early warning to dig deeper into potential customer dissatisfaction. Check this visualization of NPS calculations."

...and another example - The Alternative to Roadmaps by Marty Cagan

"Marty Cagan from SVPG believes that roadmaps fit the command-and-control model, not a product organization. He posted this article in 2015 but it became even more relevant now. Maybe one day, the product management community will say: roadmap is dead, long live OKR.

6. Track the Progress

Prepare content-based questions to track the progress. For example, if you share an article on, say, RICE Prioritization Framework, you need to test whether the learner took the effort to learn from it.

Multiple-choice questions are an effective method to test knowledge. Based on the user feedback to our chatbot Product Management tutor Eddy, users prefer more than 1 question for knowledge testing. Make effort to explain the 'why' behind the correct and each incorrect option in the multiple-choice questions.

When we created Eddy, we didn't at first provide details on each incorrect option until our users pointed at it. We learned to take advantage of the incorrect answer – it is a chance to contribute to the learners' experience. Explain what is wrong with the given option and offer to try again. For correct options - take a chance to reiterate why it is correct and provide a solution – it helps to retain the material.

Below are the 3 Key Takeaways from the creation of multiple-choice questions:

1. Make smart incorrect options – a good indicator of quality is the time spent on the preparation of incorrect options. Smart incorrect options require 2x more time than correct options on average.

2. Don't try to test the empirical knowledge, instead create a problem that could be solved using the material learned.

3. Mind the platform – since the chatbot is accessed via a smartphone, don't include heavy calculations or complex spreadsheets.

At Corpedios, we put together the best features of text, images, and videos to create an exceptional onboarding experience for employees at your company. Try onboarding new talent via a conversation that we tailor to your needs in our text message-based onboarding solution. Your coworkers and new hires text – would it be terrible to put it to good use?

Get a Demo