Loom Workflows

A note about this case study

This case study is a deviation from my others. Instead of digging deep into a single project, my goal here is to tell a story of my ownership and impact in a single problem space over the course of a year. From Feb 2025 to Feb 2026, I shipped three major improvements to make Loom's AI Workflows suite meaningfully more contextual, visible, and compelling, with the metrics to back it up.

Although my major design decisions are covered below, I've cut a lot of the explorative and iterative steps to focus on the bigger picture: my ability to advocate for and solve a problem space. Please get in touch if you'd like to see more of anything specific. Thank you!

Overview

TIMELINE
TEAM
ROLE
TOOLS

Feb 2025 - Feb 2026
Async Workflows
Product Designer
Figma, Figma Make

Loom is async video messaging for work – users record their screen to efficiently communicate things that are better shown than told. The result is a shareable video that reduces the need for meetings and enables async work.

Though Loom videos are an effective way to communicate, UXR indicated they can be hard to action on. Creators often record Looms to explain bugs or document processes – yet still have to manually translate that video content into written tickets or step-by-step documentation afterward.

To solve this, Loom in early 2024 launched AI Workflows. Users could now turn their videos into AI-written bug reports, how-to guides, and more. While AI Workflows users were 4x more likely to convert to paid plans, overall adoption remained low. The issue wasn’t usefulness, but discovery and time-to-value. In 2025, I joined the team to move the needle.

Making Workflows Contextual: Smart Prompts

Though AI Workflows was beloved by those who had discovered it, the vast majority of Loom users didn’t know what Workflows could do, and rarely thought to trigger them on their own. Even when Workflows were relevant, they were hidden behind manual actions and generic entry points.

[data-wf-bgvideo-fallback-img] { display: none; } @media (prefers-reduced-motion: reduce) { [data-wf-bgvideo-fallback-img] { position: absolute; z-index: -100; display: inline-block; height: 100%; width: 100%; object-fit: cover; } }

Our first idea for improving AI Workflows was Smart Prompts. Smart Prompts proactively suggest the single most relevant workflow type by using AI to determine the video's intended purpose. If the creator sounds like they're walking through a bug, then as soon as the creator hits "stop record", they're prompted to generate a bug report. Instead of asking users to decide what to do, Loom would now make a confident recommendation at the moment of intent, and enable AI Workflows with a single click.

The following design considerations reframed AI Workflows from an optional tool into a helpful assistant that anticipates next steps:

Show suggestions only when confidence is high to avoid noise
Smart Prompts would only be displayed when our model had high confidence that the video matched one of our document types. At the time of writing, ~52% of all Loom videos display Smart Prompts.
Preview the generated output to build trust
‍I pushed for Smart Prompts to include a preview of the generated output. This served several purposes – to communicate the feature's value at a glance, to show that the document was already available to use, and to build trust with AI, which was still a new way to get value out of Loom at the time.
Allow users to dismiss or recover gracefully from incorrect suggestions
As with any AI feature, the recovery path was critical. We ensured that Smart Prompts remained easy to dismiss and fitted logically within the IA of the full suite of creator actions.

Results

This feature became available in Q2 2025 for our Business + AI and Enterprise users. Amongst these users, we saw a 20% increase in Workflows entry rate after shipping Smart Prompts. Documents generated also increased by a statistically significant amount, but it would be much more impacted by what came next.

Making Workflows Visible: the "Generate" Tab

Even with Smart Prompts, Workflows' only entrypoint was buried in the post-record editing sidebar. Recognizing that we were at a critical stage, I wrote a thorough strategy doc outlining how Workflows was fighting an uphill battle from not getting enough business prioritization. My rationale centered on the need for a dedicated home for Workflows in the IA.

With buy-in from product, I revamped Workflows' presence in Loom's post-record experience:

Introduced a new "Generate" tab in the sidebar
A high-visibility, predictable entrypoint went a long way. By moving Workflows into its own tab, I positioned it as a valuable follow-up step that allows users to turn their video into action.
Designed a new landing page to ease users into the experience
‍A new tab helps get people in the door, but has no staying power on its own. To communicate our value upfront, I led with a visual of turning video into integrated work artifacts. In addition, I chose to surface only the top three document types by usage to keep the experience focused.
Explored loading states and transitions in Figma Make, helping set expectations during async generation
As I ramped onto AI prototyping in the fall of 2025, I found it was especially helpful for motion design. Animations that would be high-effort in Figma could now be accomplished with a prompt or two. I relied on Figma Make to communicate loading states to my engineers and make the experience feel much more alive.

Results

After shipping our new "Generate" tab experience, Workflows usage increased dramatically. The number of videos with workflows engagement rose by 37%, and the number of videos that resulted in generated documents increased by 30%.

Making Workflows Compelling: Developer Context

Smart Prompts and the new "Generate" tab were directed towards widening Workflows' top of funnel and visibility. After launching those improvements, we redirected our focus towards making Workflows deeply compelling for Atlassian's biggest persona: the engineer.

Engineers are the #1 users of Loom, and bug reporting is a frequent but time-consuming part of their workflow. Developers who record Looms to capture bugs on video still have to manually re-enter reproduction steps and bug context into work management software afterward. This gap revealed a clear opportunity: using AI Workflows to turn Loom recordings into bug reports automatically – and enriching them with developer context to provide deeper, more actionable bug reports.

Dev Context was a meaty project that involved end-to-end changes, from our pre-record UI to our post-record experience. But at every step, I emphasized making the user flow as straightforward as possible:

Users enter from the Loom Chrome extension, or from the Jira work item creation step.
‍These locations allowed us to catch users in their usual flow of work.
Clicking into the entrypoint opens the Loom recorder in our new bug reporting mode.
‍I designed two primary affordances into the new mode. First, a short checklist served as a mental guide on what's helpful in a bug report. Secondly, a yellow border highlighting the browser window (similar to Zoom's screen-share border) indicated that there was additional context being captured behind the scenes.
As soon as recording is done, users are taken straight to their auto-generated bug report.
Also known as the previously-outlined "Generate" tab. This deviates from the usual post-record experience, which is more focused on audio and video editing. By displaying the bug report immediately, we make good on the user's upfront intent to capture a bug and allow them to immediately review the generated output to build trust.
Developer context is visible and ready for consumption.
‍This is where the real magic happens. Our new bug reporting mode captures device info, console logs, and network logs in the background while the user is recording, then displays them alongside the generated bug report. These logs are visible to viewers by default, as most bug reports are resolved by a viewer of the video (an engineer), not the creator.

The end result is a vastly improved bug reporting workflow that reduces both the time to create a bug report and the amount of back-and-forth needed for the engineer to actually resolve the bug, all with nearly-zero additional effort required from the creator.

Results

Developer Context went into a public beta experiment in late Feb 2026. Though I was not able to capture the latest experiment data before leaving Atlassian, early signal showed a 61% increase in bug reports generated through Workflows.

Looking Back

My time on Loom Workflows was full of valuable lessons. I had mostly tackled 0 -> 1 projects before joining the team, and I learned to adjust my strategy to improve an existing feature rather than build something new from the ground up. We had the benefit of existing usage data, quick experiments, and willing users for customer calls. Together, it was a strong knowledge foundation for where to go next.

I used to believe that if a feature was useful enough, then people would come. However, Workflows showed that things are not always so simple. At the start, it was Loom's strongest conversion feature, but only 4th or 5th in terms of engagement. The initial signal was there, but it was up to me and my product partner to actually unlock its potential. That led me to my biggest lesson: don't underestimate the power of a well-written one-pager. Many designers suffer from jumping into Figma too soon. A one-pager is, at worst, a design and strategy resource to refer back to. At best, it's changed someone's mind and gotten me the buy-in I need to succeed.

Turning videos into action

A note about this case study

Overview

Making Workflows Contextual: Smart Prompts

Results

Making Workflows Visible: the "Generate" Tab

Results

Making Workflows Compelling: Developer Context

Results

Looking Back