CASE STUDY

Introducing AI

When it became clear in early 2023 that AI was changing the tech industry, we held a hackathon to ensure Mighty was using this awesome technology to our advantage, and ultimately released a powerful set of AI features.

The Problem

With AI setting new consumer expectations, we needed to raise the bar for the services we provided at Mighty to show we were on the leading edge. Our challenge was to:

  • Be radically more helpful to our Hosts and members by incorporating AI into the core of our platform.

  • Maintain our position as a leader in tech innovation. The timeline was tight—if we didn’t move quickly, someone else would.

  • Learn to work with the technology. With AI advancing rapidly, it was crucial to dive in and start experimenting with real-world applications. We needed our product team to fully embrace the use of AI, and learn how to work with it early and often.

All of that being said, we already had a full roadmap of critical projects for the year, leaving us wondering where we could possibly fit in this brand new set of work. This led us to our solution - a two week, company-wide hackathon.

PS: In the spirit of embracing AI, I used ChatGPT to help me write this case study. You’ll have to let me know — how do you think it went? 🤖

The Solution

We held a two week, company-wide hackathon to kick off our AI journey. But why a hackathon?

  • It allowed us to integrate AI into our product workflows quickly and organically.

  • The hackathon format demonstrated how fast and agile our teams could be when given a clear goal and a short timeframe.

  • Our squads were already adept at delivering meaningful results, so this was an opportunity to adapt under pressure and hone our skills in incorporating cutting-edge technology.

To keep the focus sharp, we set several criteria for the hackathon:

  • Features must help Hosts or members – Any solution needed to improve the efficiency and success of our Hosts or increase member engagement.

  • AI-powered tools, not chatbots – We sought tools that leveraged AI to enhance our platform’s unique needs, not just replicate what’s already available.

  • No popularity contests – Projects had to deliver real value to our users and align with our product goals.

  • Attach to an AI identity – We explored how to present AI within our platform, whether through an “AI Co-Host” identity or more subtly, with no explicit mention of AI.

In addition, the product design team played a very specific role during the hackathon. Not only were we participants in the ideation alongside engineers, PMs, researchers, and more, but we needed to think holistically about the our platform, and how these AI features would incorporated into our existing systems. This included the following key tasks:

  • Learning prompt engineering to optimize user interaction with AI features now and in the future.

  • Creating a clear visual language to distinguish AI-powered elements and ensuring consistency throughout the platform. A major goal was to make AI features feel interactive and conversational, rather than static.

  • Consulting across every hack team to help define use cases and create visuals to sell their vision.

How Did It Go?

The hackathon moved quickly over the course of two weeks. We began with a brainstorming phase that generated over 50 ideas. Squads sponsored the projects they were passionate about, working on 22 hacks and preparing them for our internal “Shark Tank” demo. After intense collaboration, we selected 11 projects to move forward into production, and were ultimately able to release them all over the course of the next couple of weeks.

Visual Language

One of the most important problems the product design team tackled, along with participating in building hackathon features, was creating a consistent visual language for our AI features. We aimed to create an AI presence in our platform that felt…

  • Alive - One of our product principles at Mighty is that a Network should always feel alive with activity, and AI should be no exception to that. We wanted to make sure that during any loading, processing, or down time when using an AI feature, that you felt like you were interacting with an entity that was deserving of a spot in the Network.

  • Approachable - While people in the tech industry are all super familiar with AI and its benefits, our user base is extremely broad, and often includes those who are not as technically savvy. We wanted to make sure our focus was on the power of the features we were providing, rather than the technology of that AI is at its core.

  • Contained - As much as our AI features needed to blend into the rest of the Network, we also wanted to be clear about what was making use of AI and what wasn’t. We did this by using the classic sparkle icon, but also by being extremely consistent with our visual language.

Below is an example of a spec we created for our buttons, which addresses all of these areas.

As with every project at a fast moving startup, especially during a hackathon, you’ll notice that even with these guidelines our visual language is not as consistent as we’d like. It’s a constant work in progress, but the most important thing from the design perspective is that the groundwork for the rules were laid early, and are referenced often.

Favorite AI Features

While we managed to release a whopping 11 features to production during our hackathon, there are a select few that continue to stand out to me as meaningful additions to our product and our mission at Mighty.

Show Similarities and Conversation Starter

Found on member profiles, this feature suggests commonalities between users based on engagement, location, bios, and more, and uses those commonalities to generate conversation starters to break the ice in chats.

These features speak to one of the core principals at Mighty - People Magic. In this instance, AI helped us find ways to alleviate some of the awkward moments one encounters when trying to connect with someone new, allowing the software to feel magical in it’s ability to facilitate human connections.

Suggested Hashtags

This feature appears in Quick Posts and Articles, providing hashtag suggestions where none exist, marked by a sparkle icon. Hashtags play two roles in our product - allowing people to express themselves, as well as categorizing content around topics or ideas. Removing the work of deciding what hashtag to add to a post makes it feel less onerous and therefore more likely for people to use them. Plus, who doesn’t love when they can look a little extra witty, with no extra effort?

Inactive Member Re-Engagement

In this instance, AI helped us meet a longstanding Host ask in our platform more quickly than would have otherwise been possible. With this feature, Hosts can identify who in their Network is considered to be “inactive,” and quickly reach out to try and re-engage them. We use AI to create personalized prompts that allow the Host to connect with this member, learn why they may not be interested in participating in the community, and facilitate conversation to try and get them coming back.

Conclusion

In addition to helping us stay on top of the latest trends in the industry, our hackathon helped our teams gain great skills that we were able to incorporate back into our daily workflows:

  • Rapid demos and prototypes - culturally this was not a skill we were particularly good at practicing, but the hackathon helped restate the value of seeing and using features early and often to help make good product decisions.

  • AI as a product tool - Because we made such a big deal about AI early on, it gave permission to our product team to think of other ways we could incorporate AI into the product even once the hackathon was over.

As with any product release, and especially one of this speed and magnitude, we had a number of learnings and quick adjustments to make:

  • Ability to toggle AI features - some Hosts were particularly skeptical about the use of AI in the product, prompting us to add an opt out toggle for each Network.

  • Feedback mechanisms - While we acknowledged the risks of this decision ahead of time, we released our suite of AI features without any in product feedback mechanisms for uses. This meant that AI was generating content for Hosts and members, but there was no thumbs up/thumbs down feedback mechanism to quickly allow them to share the quality of the results. This made it hard to measure the effectiveness of the features, or know where to improve our prompts.

  • Mighty Co-Host - As a B2B2C business, we serve two sets of users: Hosts and members. We decided to call our AI “Mighty Co-Host,” which ultimately ended up causing quite a bit of confusion. The “Co-Host” entity was never mentioned to members, and Hosts had a hard time understanding that members did have AI features, but they were nameless compared to the “Co-Host” features they saw themselves. We ended up backing away from this name and moving towards a nameless approach for all.

In the end, this was one of our most successful hackathons ever at Mighty. And while not every feature we built was the most innovative or unique, we were able to meet our goals of accelerating AI adoption, proving the strength of our teams to quickly deliver impactful innovations under pressure, and embrace the wild world that is modern technology.