Back to list

Desktop App

Ed Tech

Essay's Topic Finder

Unlike other writing apps that focus on sentence structure or grammar correction, Essay stands out for its mission to help writers develop their skills and express creativity through purposeful writing.

To build on this vision, I collaborated closely with Essay’s development team to design an AI-powered topic discovery tool that would assist users in finding meaningful topics to write about and increase their engagement with the app.

May to August 2024

Timeline

Product Designer

Role

Shipped feature

Output

Increase in DAU/MAU by 9%

Impact

May to August 2024

Timeline

Product Designer

Role

Shipped feature

Output

Increase in DAU/MAU by 9%

Impact

THE PROBLEM

One of the biggest challenges in practicing writing is simply finding a topic that feels engaging. Recognizing this, our CEO proposed exploring a feature that would help users discover topics that spark their interest. This direction aligned with Essay’s core value proposition of promoting purposeful, creative writing while also aiming to boost declining user engagement.

THE SOLUTION

We collaborated with an LLM developer to build an AI-powered experience that could learn from each user’s interests and generate personalized writing topics that reflected what they enjoyed writing about. The outcome was a topic finder tool that was:

✅

Simple and intuitive onboarding to help the LLM understand each user’s interests and writing preferences.

✅

Clear, guided messaging to walk users through using the Topic Finder when starting a new essay.

✅

Personalized topic suggestions with opportunities to continuously refine the user’s LLM profile through feedback and optional profile questions.

THE PROCESS

USER TESTING

Unlike with most design projects, I kicked this one off with an existing prototype that was built by an external LLM developer commissioned by Essay’s CEO. Since the demo was a rough version of the feature we were hoping to build, I decided to do an informal user testing session with everyone on the Essay team. My hope was to garner some feedback and first impressions that I could use to guide my designs.

FINDINGS

Purpose of the questions are unclear

While it was understood that answering these questions would lead to something, the messaging wasn’t clear as to the purpose of each question, and so the motivation to continue answering questions would quickly fade.

Purpose of the questions are unclear

Uninspired UI

Although this was a rough prototype, it was clear that the burnout of answering questions needed to be mitigated by a more engaging UI to make answering the questions feel more interactive rather than like filling out a form

Uninspired UI

Separating mandatory from optional

Many people on the team expressed the desire to want to skip certain questions and not being to do so lent to worsening decision fatigue.

Separating mandatory from optional

Many people on the team expressed the desire to want to skip certain questions and not being to do so lent to worsening decision fatigue.

Separating mandatory from optional

Many people on the team expressed the desire to want to skip certain questions and not being to do so lent to worsening decision fatigue.

Separating mandatory from optional

Many people on the team expressed the desire to want to skip certain questions and not being to do so lent to worsening decision fatigue.

Separating mandatory from optional

Many people on the team expressed the desire to want to skip certain questions and not being to do so lent to worsening decision fatigue.

Separating mandatory from optional

Many people on the team expressed the desire to want to skip certain questions and not being to do so lent to worsening decision fatigue.

Tangible results needed

Answering more questions would obviously help the LLM to understand the user better, however, without any tangible results or end goal, it was hard for users to feel engaged with the questions

Tangible results needed

Answering more questions would obviously help the LLM to understand the user better, however, without any tangible results or end goal, it was hard for users to feel engaged with the questions

Tangible results needed

Answering more questions would obviously help the LLM to understand the user better, however, without any tangible results or end goal, it was hard for users to feel engaged with the questions

Tangible results needed

Answering more questions would obviously help the LLM to understand the user better, however, without any tangible results or end goal, it was hard for users to feel engaged with the questions

Tangible results needed

Answering more questions would obviously help the LLM to understand the user better, however, without any tangible results or end goal, it was hard for users to feel engaged with the questions

Tangible results needed

Answering more questions would obviously help the LLM to understand the user better, however, without any tangible results or end goal, it was hard for users to feel engaged with the questions

Design principles

Guide users early and closely

We should be guiding users at the start of the quiz to help them understand the purpose of all the questions.

Utilize some variety

Use visual cues to break up the form’s monotony and clearly indicate when users progress to a new stage of the quiz

Simplicity at its best

We should focus on only the questions that are essential for the LLM’s baseline data. The fewer required, the better the user experience.

Transparency builds trust

Users need to understand the purpose of their answers and see tangible results to stay motivated and engaged.

First iteration

The first step was outline how the topic tool might fit into our existing app. Did we want a separate feature or did we want to incorporate this into our existing tools?

Brainstorm of the various ways we could incorporate a topic generator into the Essay platform.

Brainstorm of the various ways we could incorporate a topic generator into the Essay platform.

Based on the user flows I outlined, the team decided to incorporate the topic finder into our existing tools to make the addition of a new feature feel more seamless.

I then outlined what an ideal user flow would look like, from onboarding and then into the actual app itself and its interactions with various other tools.

Outline of the user flows for the Essay tools that will be impacted by the topic tool.

Then came V1

The first iteration of the onboarding quiz.

The first iteration of the onboarding quiz.

The first iteration of the brainstorm modal. This is where the topic finder lives in the Essay workspace.

The first iteration of the brainstorm modal. This is where the topic finder lives in the Essay workspace.

SECOND ITERATION

Pain points

Confusing buttons and choices

Due to the introduction of new buttons and options to the UI, users became confused due to the similarities in copy and new choices that they had in front of them. Users were more likely to bypass the new buttons entirely until prompted to complete a task.

Confusing buttons and choices

Unsure of purpose of the quiz

Most of our user testing participants could not discover the explanation modals for the purpose of the quiz. Due to this confusion, many participants decided to skip the quiz entirely and go straight to the Essay homepage.
Participants that did complete the quiz expressed feeling fatigue and disinterest by the time they were able to complete the quiz – mostly due to the amount of questions and choices they needed to make before even hitting the home screen.

Unsure of purpose of the quiz

Most of our user testing participants could not discover the explanation modals for the purpose of the quiz. Due to this confusion, many participants decided to skip the quiz entirely and go straight to the Essay homepage.
Participants that did complete the quiz expressed feeling fatigue and disinterest by the time they were able to complete the quiz – mostly due to the amount of questions and choices they needed to make before even hitting the home screen.

Unsure of purpose of the quiz

Most of our user testing participants could not discover the explanation modals for the purpose of the quiz. Due to this confusion, many participants decided to skip the quiz entirely and go straight to the Essay homepage.
Participants that did complete the quiz expressed feeling fatigue and disinterest by the time they were able to complete the quiz – mostly due to the amount of questions and choices they needed to make before even hitting the home screen.

Unsure of purpose of the quiz

Most of our user testing participants could not discover the explanation modals for the purpose of the quiz. Due to this confusion, many participants decided to skip the quiz entirely and go straight to the Essay homepage.
Participants that did complete the quiz expressed feeling fatigue and disinterest by the time they were able to complete the quiz – mostly due to the amount of questions and choices they needed to make before even hitting the home screen.

Unsure of purpose of the quiz

Most of our user testing participants could not discover the explanation modals for the purpose of the quiz. Due to this confusion, many participants decided to skip the quiz entirely and go straight to the Essay homepage.
Participants that did complete the quiz expressed feeling fatigue and disinterest by the time they were able to complete the quiz – mostly due to the amount of questions and choices they needed to make before even hitting the home screen.

Unsure of purpose of the quiz

Most of our user testing participants could not discover the explanation modals for the purpose of the quiz. Due to this confusion, many participants decided to skip the quiz entirely and go straight to the Essay homepage.
Participants that did complete the quiz expressed feeling fatigue and disinterest by the time they were able to complete the quiz – mostly due to the amount of questions and choices they needed to make before even hitting the home screen.

Limited understanding of how to use the tool

Through a combination of unclear copy and proper sectioning of the tool finder tool’s modal, most of the participants were unable to understand how to use the tool.

Limited understanding of how to use the tool

Through a combination of unclear copy and proper sectioning of the tool finder tool’s modal, most of the participants were unable to understand how to use the tool.

Limited understanding of how to use the tool

Through a combination of unclear copy and proper sectioning of the tool finder tool’s modal, most of the participants were unable to understand how to use the tool.

Limited understanding of how to use the tool

Through a combination of unclear copy and proper sectioning of the tool finder tool’s modal, most of the participants were unable to understand how to use the tool.

Limited understanding of how to use the tool

Through a combination of unclear copy and proper sectioning of the tool finder tool’s modal, most of the participants were unable to understand how to use the tool.

Limited understanding of how to use the tool

Through a combination of unclear copy and proper sectioning of the tool finder tool’s modal, most of the participants were unable to understand how to use the tool.

Improvements

Before

Dropdown menu meant more clicks before users could get to the topic generating modal.

After

Streamlining the user flow simplified the options onto the modal itself and reduced # of clicks.

Before

Onboarding quiz text was confusing and provided very little context for users.

After

Modals were used before each section to grab the attention of the user and provide more context.

Before

Users were presented with both choices in the brainstorm modal with no context, this caused some confusion between the two different choices.

After

Using a progressive disclosure modal allowed users to see more context and directly see the impact of their decisions.

third ITERATION

Although the second iteration helped to address some of the confusion surrounding the purpose of the onboarding quiz, I felt like it didn’t address one of the main concerns of users expressing the desire to drop off during the onboarding quiz.

After some brainstorming with the development team, we decided to do an overhaul of the onboarding quiz to simplify the copy, the user flows, and get users to the home screen in a more direct way.

✅

Simplified the copy of the interest selector

✅

Removed the RIASEC categories as they were adding to the confusion the copy of the interest selector

✅

Simplified the user flow for selecting interests and adding interests to mimic existing onboarding personalization flows from other similar apps

✅

Showing suggested topics right after users select an interest helped to bridge the disconnect between the purpose of the onboarding questionnaire and the topic generation

✅

Breaking the monotony of the quiz via a more visually engaging UI

✅

Minimizing the visual clutter by removing any sort of tab selection option

✅

Streamlining the flow to make it easier for users to provide the LLM with more guidance and provide feedback on generated topics

KEY LEARNINGS

For this project it really was a learning experience of balancing the user experience and the business desires. Since each topic that is generated using the LLM comes at a cost, it was in the company’s best interest to gather as much data as possible on the user in order to generate well-informed topic suggestions. However, the user testing showed that users found answering questions tedious and boring and could potentially increase the bounce rate. As such I had to get comfortable asking a lot technical questions and collaborating with the development team to brainstorm an onboarding flow that would gather enough information for the LLM while not being tedious for the user.

NEXT STEPS

After many brainstorming sessions and back and forth decisions made by both the development team and myself, we were finally able to release the new feature into Essay in September 2025.

Since release, we’ve seen an increase in DAU/MAU from 16% to 25%, and many users have expressed how helpful the tool has been in finding interesting writing topics. For the real third iteration, we will continue to monitor user interactions for any pain points and feedback.

After many brainstorming sessions and back and forth decisions made by both the development team and myself, we were finally able to release the new feature into Essay in September 2025.