Roger Chen, Cofounder & CEO of Whatsgood.
As a founder in consumer social for eight years and having built products that have reached No. 1 on the App Store, I’ve come to find that building social products is a deeply rewarding experience. You get to help people connect and build relationships. However, as every product person in the social space knows, one of the toughest problems in building social products is “breaking the ice.”
Icebreaking is a problem that you see in many scenarios across many products, from dating to friends and even the workplace. When people get matched on Tinder, for example, most matches don’t lead to conversations, and the reason is simple: people don’t know what to say. They agonize over what to send in that first message. You can’t be boring, but if you’re too out there, you end up as a meme on @tindernightmares.
The same problem exists in the workplace. In the workplace, the water cooler conversations have always been critical to building team culture. However, as a lot of teams move toward remote or hybrid, it has become harder to continue this casual banter online. It’s harder to strike up a seemingly aimless conversation in Slack, and, as a result, a lot of conversations that should have happened didn’t, and this can cause teams to drift apart.
Again, the key problem is the cold start—the icebreaking. Once a Slack thread has two or three people talking in it, it becomes easier to keep it going.
Using AI To Break The Ice
Icebreaking requires context. It takes work to create conversations between people and keep those conversations alive, and this work is perfect for large language models (LLMs). If there is one thing LLMs are especially reliable for, it’s chatting.
Sound vague? Let’s explore how AI can be used to help break the ice in the workplace.
A key element typically missing online is casual conversations between different teams. The water cooler conversation happens for a reason—you are there getting something to drink, and someone is also there. Now, to make things a bit less awkward, you break the silence and chat for a bit.
Perhaps things were a bit weird going into the conversation, but after the chat, you might feel great about the discussion you just shared. It’s the slightly forced nature of it that makes water cooler conversations possible. However, it’s hard to recreate this online because every single behavior in the digital world is optional. Therefore, there is no serendipity.
I believe we can solve this by using AI as an effective—perhaps slightly annoying—”orientation leader” to create group chats between teammates. Let’s explore how this might look.
Meeting People
One of the most basic functions of the water cooler chat is for you to meet new people, but this is much harder in a professional setting online. It can be seen as odd to just DM a colleague on Slack and say, “Hey there, what’s up?”
To solve this, I suggest creating a Slack bot that pulls people into group chats. Here are a few elements I believe would be critical to ensuring this approach is effective:
1. Every week, the bot finds a specific time to create a Slack group chat with three users: you, the AI and another colleague.
2. It selects a colleague who is online at the moment and not busy. This can be done by providing access to everyone’s Slack status and Google Calendar.
3. The bot positions itself as an orientation leader and initiates the conversation by sharing some notable information, like what you both might be up to. It gets the data from your public Slack threads and your commits in Github and Notion.
I imagine the message might read along the lines of “Hey, Cameron! I think you should meet Taylor. Taylor is our new backend engineer and has already shipped an update to our recommendation system. Cameron is our long-time designer and is currently working on a new sign-in flow. And you’re both from Atlanta!”
This gets a conversation started that may have otherwise never happened without the use of AI.
Impromptu Jamming
One of the most magical things about working with an in-person team is the impromptu jamming. Needless to say, this is another aspect that can be difficult when working remotely.
Beyond initiating water cooler conversations, a Slack bot could go a step further by helping people jam online:
1. Every week, the bot reads the latest Notion, Coda or Google Docs from each team.
2. The AI then summarizes the latest developments and, in its own language, announces what each team is working on in an open Slack channel and separate Slack threads.
3. While everyone can ask the bot follow-up questions, they can also jump into a thread. The corresponding team members can jump in to offer more clarification, and people from other teams can jump in and jam ideas.
Few people actually enjoy having to summarize what they did and then “present to the class.” However, having a third party take care of this work can help save time and may even help colleagues feel less defensive about the current state of their project.
Due to hallucinations, it’s important to note that the third party may not always have the details completely accurate. However, I see this as another potential benefit, as this offers colleagues the chance to clarify specific points to everyone and spur more discussions. Again, the AI is simply here to be the middleman that links people together.
Conclusion
The biggest takeaway is that since people can help break the ice between others, LLMs can accomplish this as well. The AI, in this case, is just a means to an end. We find meaning not in talking to the AI but in talking with other, real people. The goal of AI should not be to displace human interaction but to spur more and strengthen it in the process.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
