In the words of my friend John Stokvis, “Put an AI on It.” A play on a phrase from a comedy skit in Portlandia called “Put a Bird On It” where a quirky shop puts a bird on literally anything and everything. That’s what we’re seeing today as companies rush to put AI on anything and everything, and often in a format reminiscent of Clippy.
Microsoft’s Clippy was part of a vision that provided proactive and reactive help while using Microsoft Word. Clippy had the best of intentions, but as Microsoft Word became more commonly used, and more people understood how to use it, Clippy became the butt of many jokes. Viewed as an annoying character who inserted itself in your work at inconvenient times.
In the race to integrate with AI, many companies have created chatbots that jump out at you unannounced like you’re going through a haunted house. Your website or app shouldn’t be scary or remind users of a haunted house. I have thoughts on how to ensure that you deliver value to your customers with AI in a less scary way.
I will give a disclaimer that chatbots are not inherently bad. They are frequently misused. I can’t count the number of times I landed on a website, and before I can even finish reading what the company does, a chatbot is already asking me if I have questions. Yes, why can’t you let me finish reading the page, thank you. If you have to make your chatbot pop out 3 seconds after the user arrives on the page and they are new to the site, you’re unnecessarily boosting your engagement metrics with it. I’d love to see the analytics of companies who do this and see the number of unique visitors that prompt the chatbot vs the number of unique users who click or engage with the chatbot in some way. My theory is most people are trying to figure out how to close the darn thing. Microsoft may have killed Clippy, but Clippy’s GenAI grandchildren have prolifically multiplied across the internet. If you need to have your chatbot auto-open like that, it’s also possible that your users do need help, but are expecting their core needs to be met in a way that is integrated with the UI, and not a chatbot. This could be a reason why if you don’t auto-open the chatbot, they just pass on by.
Let’s dig into that point: integrating the way to deliver value into the UI of your website or app itself better meets users’ core needs. This can take on many forms depending on who your customers and users are, the core value you provide them, and how they naturally engage with a specific activity or flow. Just because a chatbot CAN help them at a specific juncture, doesn’t mean that is the best way to deliver value to them depending on what they need.
A month ago, a friend of mine was talking to me about a chatbot they were building for an online course website. It was clear there wasn’t a defined user problem. It was more of a mandate to give students a study partner. I love the idea of a study partner, but study partners are most helpful when they have context about you, what you struggle with, and things that help you unlock understanding. The question needs to be asked, will someone new to a topic know the right question to ask? A teacher or tutor can ask questions in a non-quiz format to assess how well the student is understanding, but chatbots struggle to do that in a seamless way. It’s a good sign that more can be done in the interface with AI to help the student learn, leveraging data on what they struggled with in previous lessons and what they excelled at in previous lessons to shape the learning experience. It would be amazing if there was a digital notebook for students as they watch lessons where they can highlight things they don’t understand and the AI leaves a comment next to that specific part of the notes about what it means. Meanwhile, it can live side by side with the rest of the student’s notes. The notes now become something special to that platform that the user needs to go back to get the value of the notes. It also becomes data for AI to leverage to customize future lessons, quizzes, and more. Chatbots can help in a learning environment, but there are so much more creative and effective ways to leverage AI in a learning environment.
Stop and think:
Who is my customer?
How can I leverage data in a way that adds significant value that wouldn’t be available without AI?
The key here is that “significant value” cannot be delivered if the implementation of AI is done in a way (even for non-chatbots) that is not intuitive for users, or not high enough value that people will force themselves to learn a new way of doing things.
Here is where chatbots are helpful before building more AI into your product. The gold is in the prompts. Analyzing the prompts to understand repeated themes in what people are asking and where they are asking it helps you understand two key things:
Where your chatbot is acting as a band-aid for things that should be in your product
What people need that they’re not getting in terms of value in a specific part of the flow
If you’ve already implemented a chatbot, this will help inform you of what to build and why in a more integrated way.
One of my favorite things is thinking through the best ways for people to engage with products powered by AI. Context is king. That’s why I recently added a custom GPT named “Claudia” to my private Slack space because I treat Claudia, who is an AI, as a collaborator, and I collaborate the most in Slack. To me, that is better than constantly opening another tab to talk to her. This is just one example of the fun ways we can integrate AI into the right channels for our users.
I will wrap up here for now. If you’d like to learn more, I will be speaking at AI in Production on July 18 in Asheville, NC. My talk is called Avoiding the Chatbot Trap: Designing AI Interfaces That Unlock Real Business and Customer Value.
Be thoughtful. Be curious. Be creative.
If you’d like to learn more about me and my work, visit annetgriffin.com.
Share this post