Review: Lowe’s Chatbot MyLo
As a conversation designer, I want to grow and improve, and I think it helps to see what else is out there. In that spirit, I want to try out other conversational AI products, observe what they do, and learn. What does the interface look like? How does it feel? Anything particularly impressive or frustrating?
For background, I was thinking about buying metal shelves on which I could put some plants. I had been searching on different sites, and eventually ended up at lowes.com. When I get there, I see a drawer in the bottom right:
There is a picture of someone (unnamed), an online indicator, a short welcome message, and the Lowe’s logo. The design for the unopened drawer feels fairly inviting and large enough to see but not be too distracting, though the colors and background obscure it. And it worked! I clicked on it! Assuming that was the goal.
If you start by clicking the (x) in the top right, it minimizes to this:
The red circle feels jarring, in contrast to the blues and greens everywhere else, as if I’d done something wrong. If I click on either the “chat now” bubble or the red circle, it makes an audio beep and loads this page:
This is the start state. At the top it says “Message us”, while the chat bubble says “Hi! I’m Mylo…”. I’m not sure what the “Message us” adds, and it confuses me to be messaging “us” when there is only one person (MyLo) that I’m talking to. The “…” at the top does offer a nice “Print transcript” feature, though the look and feel of the transcript could be better.
The content positioned at the center of the interface includes MyLo’s speech bubble and eight buttons, presumably to suggest likely questions or searches. At the bottom is a text input field with a paperclip for attachments (at this point I don’t know what I would want to attach), a smiley face to add emojis, and an airplane icon that is greyed out. Once you enter text in the input field, the airplane changes to a solid color, which you could click to submit (or just hit enter). The “Type your message” text feels simple and intuitive, and once you start typing it goes away.
The welcome message
As shown above, the welcome message says:
We could break this message into the following parts:
1) Greeting: Hi!
2) Introduction: I’m MyLo, your Lowe’s automated assistant.
3) Orientation: Here’s what I can help you with.
4) Pre-repair: Type “menu” if you need to start over.
I think I’ve written welcome messages with all of these components myself at some point! I find writing the welcome message challenging. But it is also important, since we wouldn’t want people to drop off if they could get their needs met. I wonder if this message needs all of these components. My sense is the Orientation is the core, with the Greeting and Introduction serving to humanize the bot, and the Pre-Repair setting the stage for recovery if the user runs into trouble. The optionality of 1) and 2) becomes clear when you return to the start menu later:
This version of the message includes just an Orientation (same as above but with an added👇) and the same Pre-Repair. It makes sense to drop the Greeting and Introduction after returning to the start menu.
At this point, I’m wondering about what the bot’s capabilities are. I’ve gotten two messages on that front. First, in the unopened drawer icon, the message is “Ask me anything!”
In the screen after clicking to start, the message is “Here’s what I can help you with.”
My expectations may have started as “anything” being on the table, but then they get updated to what is “here”, presumably referring to the eight buttons on screen. There is additional ambiguity because of the text entry field, since I could try to ask about things beyond the eight buttons.
Eventually, I click on the button “Help me shop”. Then I get this:
Once I say I’m looking for help shopping for shelving, I get connected to someone (a human?). The first time I did this, I was taken aback. I wasn’t ready to talk to a person! One of the benefits of bots and other automated conversational interfaces is that we can interact without the fears or stresses of talking to an actual human. Being transferred automatically was jarring. I would have preferred being offered to talk to a human, at which point I could choose to do so or not.
While I didn’t like the way the handoff was done, I do like how seamless it is implemented. The handoff was explicit, the new agent was introduced, and the interaction continues in the same chat window. Presumably the full chat history is available to this new agent as well, so I wouldn’t have to repeat myself.
The technology
From inspecting the web page’s code, it looks like this bot is being powered by https://www.liveperson.com/. I don’t know much about the company, but from their website it looks like part of their value prop is being multichannel (mobile, web, chat, social, voice). And given their name, and the fact that I was quickly transferred to a human (I think), they may also focus on AI-to-human use cases.
Conclusion
This bot seems to be optimized for a simple, scalable entry point that can quickly transition into a chat with a real person.
Copyright 2021 Joseph Tyler All Rights Reserved
