Height was founded in 2018 with the idea that people deserved a better project management tool, in a landscape dominated by JIRA. Height 1.0 offered a flexible task system, with a chat and a simple inbox to manage notifications. I joined in February 2023 as head of design, where I took a hands-on approach and designed alongside two other designers.
With the release of ChatGPT, we saw the opportunity to reinvent our app, by integrating AI deeply where it worked best, and automate the most mundane tasks. We set ourselves to create the first autonomous project management tool.
To evaluate ideas, we made a list of features based on feedback, and brainstormed the most tedious aspects of project management: triaging and tagging tasks, creating subtasks, etc.
In the Spring of 2023, we set an aggressive sprint schedule to design, build, release one AI feature every two weeks. To mitigate the risks for existing users, we lived with the feature internally before releasing it to a wider audience. We called the result Copilot.
Narrow questions work better. LLMs are good at picking a value in a list, like tagging a task as a bug or setting its priority. Asking them to decide if an item is important enough to be added to the specs of a project is much harder, even with lots of context.
People also want to review what the AI generates, because they're worried it will shuffle things around, and because keeping things "human-approved" reinforces their value.
Introducing a companion (Copilot) was confusing. It was not clear who was speaking (the app or the AI agent), and it gave the impression that the rest of the app wasn't smart. Visually, rainbows and sparkles contributed to the feeling that AI was bolted on top of the existing app. To paraphrase Kathy Sierra: users don't want to use AI, they want the result.
Based on these learnings, we gave ourselves one year to launch a radically different app that would truly showcase what autonomous project management could be. We ranked our tentpole features, missing basic expectations, and ways it could feel like a next-gen app.
1.0 was purposely unopinionated to stay flexible when used in big teams with established workflows. As a result, new users were sometimes lost. For 2.0, we updated our data model to introduce more hierarchy: teams, projects, documents.
We had to rethink some of the app layout and navigation. In 1.0, task descriptions were hidden behind a tab (and easily forgotten). In 2.0, we kept the chat partially above the document, and introduced a side-by-side mode for quick glances and edits.
To prioritize what to build, we kept the features that worked best from the Copilot phase, and improved them with simple settings, defaults, and a coherent visual identity.
From a design point of view, we strongly believed that chatbots were often not the best way to interact with AI inside a specialized tool. We focused on creating features that solved narrow problems really well, and that could be discovered contextually.
One example of this is the flow to create subtasks with AI.
While we were building these features, we felt the need for the app to show its work, to avoid surprises, and to build trust in the automations. We started with the idea of a drawer, and explored the concept of a log showing the live activity. But it felt artificial, and nobody was going to keep the drawer open at all times. So we scrapped it.
What we kept was the idea that Height was scanning data before taking action. It also made a lot of sense to make Height, the app, the main character. To give it life, we gave it the distinct look of a monochrome multiplayer cursor. We then derived this logic to all activities where AI was involved.
In the example below, Height appears to auto-fill the teams values based on the task name.
Another key part of the 2.0 app was the UI refresh. While 1.0 had a good foundation, we took the opportunity to modernize our design system. On top of a visual refresh, we used the opportunity to fix usability issues: we introduced a second font size, lighter text color, and redesigned all our icons to make them clearer and more playful.
To prepare for the 2.0 launch, we also rebranded to a more meaningful logo, as the zap didn't specifically remind people of Height. The project was carried out by our web designer, with input from the entire team.
In the process of giving Height a unique feel, we took a look at customization. While the app needed to be usable and approachable out of the box, we studied ways for people to tailor the app to their personal preferences. While many of these ideas haven't shipped, they've definitely informed our work.
After a beta phase during the summer, followed by a lot of iterations, we launched the 2.0 app at the end of 2024, with billboards in San Francisco.
After a few months, it was clear that the concept of autonomous project management was resonating well. On the product side, we worked on simplifying some of the concepts we introduced with the new app, to get users to their aha moment faster.