GenAI Works Academy
Gatekeepers are rarely the good guys
The first Academy MVP had one, and it was a registration form. People had to create an account before they could see a single course. They could not check the topic, instructor, time commitment, or outcome first.
The survey responses and interview notes kept pointing to the same issue. I turned that finding into the recommendation to open the catalogue and move registration to enrollment. We also added course categories so people could find the right subject faster.
The other senior designer led the flows and page structure. I led the high-fidelity UI, responsive design, Academy visual system, launch marketing, and Claude Code front end.
We moved from research to launch in two months because the MVP already existed, developers gave us the technical rules for the repository, and the two designers used Claude Code to build the Next.js beta.
Academy launched in late March 2026 with 10 live courses. I designed the campaign and course assets behind the launch. The first 30 days generated a little over $12,000 gross against a $10,000 target.
By August 2026, Academy had 31 live courses and around 37,000 registered student accounts. The Lead of Education and Product Marketing Manager confirmed those figures.
Role | Senior Designer |
|---|---|
Team | Two senior designers, VP of Product, product manager, Lead of Education, four developers, two QA engineers |
Timeline | Late January to late March 2026 |
Build scale | Next.js and React, 15-20 unique pages, 80+ reusable components |
Status | Launched in late March 2026 |
What Academy is
GenAI Works Academy hosts the company's own courses alongside sponsored courses from external instructors.
I worked across the learner product and the system used to recruit, launch, and support instructors. Learners needed enough information to judge whether a course was worth their time. Instructors needed a consistent way to present and promote their material.
An MVP already existed, but the catalogue sat behind registration. Visitors had to sign up before they could see the courses.
What I contributed to research
Research started at the end of January. The team ran around 20 interviews and received more than 100 survey responses from the education group inside the GenAI Works LinkedIn community.
I wrote the questionnaire, analyzed the responses, and turned the findings into product actions. I also joined around 7-8 of the interviews and translated the notes into changes for the catalogue, categories, and learner profile.
Two learner groups appeared often. Juniors and recent graduates were building skills from the beginning. Experienced professionals were adding AI skills to their current careers.
Both groups wanted to judge course relevance before creating an account. They wanted to see the topic, instructor, expected outcome, and time commitment. They also wanted a faster way to find courses that matched their interests.
Decision 1: moving registration to enrollment
My questionnaire analysis made one pattern clear. People wanted to see the catalogue before they registered.
I translated that finding into a product recommendation: open the catalogue and move registration to the point of enrollment. The team adopted the recommendation. Courses became visible and searchable without an account, and registration came after a learner had chosen a course and understood what it offered.
The rejected option kept registration at the entrance and protected the existing account-capture flow. We chose open access because the gate blocked the information people needed to decide. Public course pages could also be indexed, linked, and shared.
I carried the decision through high-fidelity and responsive design across desktop, tablet, and mobile.
Decision 2: making courses easier to find
I proposed and designed the homepage category treatment with the other senior designer. We used clear subjects such as AI, design, and coding so learners could reach relevant courses quickly.
Management questioned whether categories added enough value. The alternative was a mixed catalogue with promotional areas doing more of the filtering work. I used interview findings, survey responses, and a review of competing learning products to defend the category model.
Categories also supported a business goal. GenAI Works wanted its internal courses and free material to stay visible. I designed the controls so those groups could be placed near the front without adding another campaign block to the homepage.
Decision 3: building the instructor publishing system
I owned the visual system used to recruit instructors, package their courses, and promote each release.
The recruitment kit included a 14-slide deck that explained the Academy offer and gave the education team a consistent tool for instructor outreach. Once an instructor joined, I produced the course package: sign-up documents, covers, thumbnails, promotional assets, and course graphics.
I set a photorealistic art direction based on real people and real environments. I excluded CGI scenes, invented product interfaces, and staged stock setups because they made the courses feel less credible.
The original course cards used stylized tags in a position that competed with the thumbnail artwork and made the catalogue feel inconsistent. I moved the badges out of the thumbnail area and redesigned them as a consistent metadata element. This gave the course title, instructor, and artwork a clearer order while keeping course type visible.
The course-card badge rules carried into the thumbnail system and promotion assets. This kept the course identity consistent from the catalogue to the campaign material.
Every instructor and every published course used the system. By August 2026, I had produced visual packages for all 31 live courses. That full adoption made the visual pack part of the publishing process rather than an optional set of templates.
Expanding the first release
Research showed that learners wanted to track progress after enrollment.
The team rejected a full learning management system for the first release. I designed the high-fidelity UI for a focused profile view that showed enrolled courses and current progress while keeping the scope controlled.
The education team also increased the amount of free material. I designed the in-house free courses and guides as an entry point into the wider catalogue.
Building the beta with Claude Code
The existing MVP, developer guidelines, and Claude Code made the two-month delivery possible.
Developers first gave us the technical rules and repository constraints. The two designers then worked in a shared repository and used Claude Code with Next.js to turn the approved Figma work into a functioning front end.
I built desktop, tablet, and mobile layouts across roughly 15-20 unique pages and more than 80 reusable components. The beta gave developers and QA engineers a working product to review before production.
I shipped that beta without checking each breakpoint as its own state. At tablet widths, it rendered a compressed desktop layout instead of the intended tablet design. The generated breakpoint logic was wrong, and I had not caught it before review. Several controls became too small for reliable touch input.
Engineering caught the issue during beta review. I handed over the intended tablet designs and related code. The developers corrected the breakpoint behavior manually before production and used the beta as the base for the released front end.
I checked the corrected tablet controls myself against a 44 by 44 CSS pixel minimum after the developer fix and confirmed that the tap targets met it.
I now treat each breakpoint as its own test state. I would add those checks throughout implementation instead of waiting for the full beta review.
Results
Academy launched in late March 2026 with 10 live courses.
During the first 30 days after launch, Academy generated a little over $12,000 in gross revenue before expenses against a $10,000 target. This was the main audience-building period. The launch used the company's existing audience and owned channels, with no paid acquisition.
I was the main designer for the campaign graphics, course thumbnails, promotion assets, and instructor packages behind that launch. The revenue was a team result supported by the visual and marketing system I led.
As of August 2026, Academy had 31 live courses and around 37,000 registered student accounts. The Lead of Education and Product Marketing Manager confirmed those figures.
What the research missed
We studied what people wanted to learn and how they chose a course. We had less evidence about what they were willing to pay.
That risk did not materialize during the first 30 days, when revenue beat the launch target. It still left the team without a strong research base for later pricing or subscription decisions. Pricing sat outside my design scope, but the missing evidence was relevant to the product work.
The sample came from a motivated education community. We needed more input from visitors who viewed a course and decided against paying.
What I would change
For the next pricing round, I would recruit non-buyers from both learner groups. I would ask about budget, past course purchases, and the point where the price became too high. I would compare per-course pricing with a subscription using conversion, revenue per visitor, and refund rate.
For the build, I would review tablet as a separate layout state and add breakpoint checks to each implementation review before beta.
GenAI Works Academy
Gatekeepers are rarely the good guys
The first Academy MVP had one, and it was a registration form. People had to create an account before they could see a single course. They could not check the topic, instructor, time commitment, or outcome first.
The survey responses and interview notes kept pointing to the same issue. I turned that finding into the recommendation to open the catalogue and move registration to enrollment. We also added course categories so people could find the right subject faster.
The other senior designer led the flows and page structure. I led the high-fidelity UI, responsive design, Academy visual system, launch marketing, and Claude Code front end.
We moved from research to launch in two months because the MVP already existed, developers gave us the technical rules for the repository, and the two designers used Claude Code to build the Next.js beta.
Academy launched in late March 2026 with 10 live courses. I designed the campaign and course assets behind the launch. The first 30 days generated a little over $12,000 gross against a $10,000 target.
By August 2026, Academy had 31 live courses and around 37,000 registered student accounts. The Lead of Education and Product Marketing Manager confirmed those figures.
Role | Senior Designer |
|---|---|
Team | Two senior designers, VP of Product, product manager, Lead of Education, four developers, two QA engineers |
Timeline | Late January to late March 2026 |
Build scale | Next.js and React, 15-20 unique pages, 80+ reusable components |
Status | Launched in late March 2026 |
What Academy is
GenAI Works Academy hosts the company's own courses alongside sponsored courses from external instructors.
I worked across the learner product and the system used to recruit, launch, and support instructors. Learners needed enough information to judge whether a course was worth their time. Instructors needed a consistent way to present and promote their material.
An MVP already existed, but the catalogue sat behind registration. Visitors had to sign up before they could see the courses.
What I contributed to research
Research started at the end of January. The team ran around 20 interviews and received more than 100 survey responses from the education group inside the GenAI Works LinkedIn community.
I wrote the questionnaire, analyzed the responses, and turned the findings into product actions. I also joined around 7-8 of the interviews and translated the notes into changes for the catalogue, categories, and learner profile.
Two learner groups appeared often. Juniors and recent graduates were building skills from the beginning. Experienced professionals were adding AI skills to their current careers.
Both groups wanted to judge course relevance before creating an account. They wanted to see the topic, instructor, expected outcome, and time commitment. They also wanted a faster way to find courses that matched their interests.
Decision 1: moving registration to enrollment
My questionnaire analysis made one pattern clear. People wanted to see the catalogue before they registered.
I translated that finding into a product recommendation: open the catalogue and move registration to the point of enrollment. The team adopted the recommendation. Courses became visible and searchable without an account, and registration came after a learner had chosen a course and understood what it offered.
The rejected option kept registration at the entrance and protected the existing account-capture flow. We chose open access because the gate blocked the information people needed to decide. Public course pages could also be indexed, linked, and shared.
I carried the decision through high-fidelity and responsive design across desktop, tablet, and mobile.
Decision 2: making courses easier to find
I proposed and designed the homepage category treatment with the other senior designer. We used clear subjects such as AI, design, and coding so learners could reach relevant courses quickly.
Management questioned whether categories added enough value. The alternative was a mixed catalogue with promotional areas doing more of the filtering work. I used interview findings, survey responses, and a review of competing learning products to defend the category model.
Categories also supported a business goal. GenAI Works wanted its internal courses and free material to stay visible. I designed the controls so those groups could be placed near the front without adding another campaign block to the homepage.
Decision 3: building the instructor publishing system
I owned the visual system used to recruit instructors, package their courses, and promote each release.
The recruitment kit included a 14-slide deck that explained the Academy offer and gave the education team a consistent tool for instructor outreach. Once an instructor joined, I produced the course package: sign-up documents, covers, thumbnails, promotional assets, and course graphics.
I set a photorealistic art direction based on real people and real environments. I excluded CGI scenes, invented product interfaces, and staged stock setups because they made the courses feel less credible.
The original course cards used stylized tags in a position that competed with the thumbnail artwork and made the catalogue feel inconsistent. I moved the badges out of the thumbnail area and redesigned them as a consistent metadata element. This gave the course title, instructor, and artwork a clearer order while keeping course type visible.
The course-card badge rules carried into the thumbnail system and promotion assets. This kept the course identity consistent from the catalogue to the campaign material.
Every instructor and every published course used the system. By August 2026, I had produced visual packages for all 31 live courses. That full adoption made the visual pack part of the publishing process rather than an optional set of templates.
Expanding the first release
Research showed that learners wanted to track progress after enrollment.
The team rejected a full learning management system for the first release. I designed the high-fidelity UI for a focused profile view that showed enrolled courses and current progress while keeping the scope controlled.
The education team also increased the amount of free material. I designed the in-house free courses and guides as an entry point into the wider catalogue.
Building the beta with Claude Code
The existing MVP, developer guidelines, and Claude Code made the two-month delivery possible.
Developers first gave us the technical rules and repository constraints. The two designers then worked in a shared repository and used Claude Code with Next.js to turn the approved Figma work into a functioning front end.
I built desktop, tablet, and mobile layouts across roughly 15-20 unique pages and more than 80 reusable components. The beta gave developers and QA engineers a working product to review before production.
I shipped that beta without checking each breakpoint as its own state. At tablet widths, it rendered a compressed desktop layout instead of the intended tablet design. The generated breakpoint logic was wrong, and I had not caught it before review. Several controls became too small for reliable touch input.
Engineering caught the issue during beta review. I handed over the intended tablet designs and related code. The developers corrected the breakpoint behavior manually before production and used the beta as the base for the released front end.
I checked the corrected tablet controls myself against a 44 by 44 CSS pixel minimum after the developer fix and confirmed that the tap targets met it.
I now treat each breakpoint as its own test state. I would add those checks throughout implementation instead of waiting for the full beta review.
Results
Academy launched in late March 2026 with 10 live courses.
During the first 30 days after launch, Academy generated a little over $12,000 in gross revenue before expenses against a $10,000 target. This was the main audience-building period. The launch used the company's existing audience and owned channels, with no paid acquisition.
I was the main designer for the campaign graphics, course thumbnails, promotion assets, and instructor packages behind that launch. The revenue was a team result supported by the visual and marketing system I led.
As of August 2026, Academy had 31 live courses and around 37,000 registered student accounts. The Lead of Education and Product Marketing Manager confirmed those figures.
What the research missed
We studied what people wanted to learn and how they chose a course. We had less evidence about what they were willing to pay.
That risk did not materialize during the first 30 days, when revenue beat the launch target. It still left the team without a strong research base for later pricing or subscription decisions. Pricing sat outside my design scope, but the missing evidence was relevant to the product work.
The sample came from a motivated education community. We needed more input from visitors who viewed a course and decided against paying.
What I would change
For the next pricing round, I would recruit non-buyers from both learner groups. I would ask about budget, past course purchases, and the point where the price became too high. I would compare per-course pricing with a subscription using conversion, revenue per visitor, and refund rate.
For the build, I would review tablet as a separate layout state and add breakpoint checks to each implementation review before beta.
GenAI Works Academy
Gatekeepers are rarely the good guys
The first Academy MVP had one, and it was a registration form. People had to create an account before they could see a single course. They could not check the topic, instructor, time commitment, or outcome first.
The survey responses and interview notes kept pointing to the same issue. I turned that finding into the recommendation to open the catalogue and move registration to enrollment. We also added course categories so people could find the right subject faster.
The other senior designer led the flows and page structure. I led the high-fidelity UI, responsive design, Academy visual system, launch marketing, and Claude Code front end.
We moved from research to launch in two months because the MVP already existed, developers gave us the technical rules for the repository, and the two designers used Claude Code to build the Next.js beta.
Academy launched in late March 2026 with 10 live courses. I designed the campaign and course assets behind the launch. The first 30 days generated a little over $12,000 gross against a $10,000 target.
By August 2026, Academy had 31 live courses and around 37,000 registered student accounts. The Lead of Education and Product Marketing Manager confirmed those figures.
Role | Senior Designer |
|---|---|
Team | Two senior designers, VP of Product, product manager, Lead of Education, four developers, two QA engineers |
Timeline | Late January to late March 2026 |
Build scale | Next.js and React, 15-20 unique pages, 80+ reusable components |
Status | Launched in late March 2026 |
What Academy is
GenAI Works Academy hosts the company's own courses alongside sponsored courses from external instructors.
I worked across the learner product and the system used to recruit, launch, and support instructors. Learners needed enough information to judge whether a course was worth their time. Instructors needed a consistent way to present and promote their material.
An MVP already existed, but the catalogue sat behind registration. Visitors had to sign up before they could see the courses.
What I contributed to research
Research started at the end of January. The team ran around 20 interviews and received more than 100 survey responses from the education group inside the GenAI Works LinkedIn community.
I wrote the questionnaire, analyzed the responses, and turned the findings into product actions. I also joined around 7-8 of the interviews and translated the notes into changes for the catalogue, categories, and learner profile.
Two learner groups appeared often. Juniors and recent graduates were building skills from the beginning. Experienced professionals were adding AI skills to their current careers.
Both groups wanted to judge course relevance before creating an account. They wanted to see the topic, instructor, expected outcome, and time commitment. They also wanted a faster way to find courses that matched their interests.
Decision 1: moving registration to enrollment
My questionnaire analysis made one pattern clear. People wanted to see the catalogue before they registered.
I translated that finding into a product recommendation: open the catalogue and move registration to the point of enrollment. The team adopted the recommendation. Courses became visible and searchable without an account, and registration came after a learner had chosen a course and understood what it offered.
The rejected option kept registration at the entrance and protected the existing account-capture flow. We chose open access because the gate blocked the information people needed to decide. Public course pages could also be indexed, linked, and shared.
I carried the decision through high-fidelity and responsive design across desktop, tablet, and mobile.
Decision 2: making courses easier to find
I proposed and designed the homepage category treatment with the other senior designer. We used clear subjects such as AI, design, and coding so learners could reach relevant courses quickly.
Management questioned whether categories added enough value. The alternative was a mixed catalogue with promotional areas doing more of the filtering work. I used interview findings, survey responses, and a review of competing learning products to defend the category model.
Categories also supported a business goal. GenAI Works wanted its internal courses and free material to stay visible. I designed the controls so those groups could be placed near the front without adding another campaign block to the homepage.
Decision 3: building the instructor publishing system
I owned the visual system used to recruit instructors, package their courses, and promote each release.
The recruitment kit included a 14-slide deck that explained the Academy offer and gave the education team a consistent tool for instructor outreach. Once an instructor joined, I produced the course package: sign-up documents, covers, thumbnails, promotional assets, and course graphics.
I set a photorealistic art direction based on real people and real environments. I excluded CGI scenes, invented product interfaces, and staged stock setups because they made the courses feel less credible.
The original course cards used stylized tags in a position that competed with the thumbnail artwork and made the catalogue feel inconsistent. I moved the badges out of the thumbnail area and redesigned them as a consistent metadata element. This gave the course title, instructor, and artwork a clearer order while keeping course type visible.
The course-card badge rules carried into the thumbnail system and promotion assets. This kept the course identity consistent from the catalogue to the campaign material.
Every instructor and every published course used the system. By August 2026, I had produced visual packages for all 31 live courses. That full adoption made the visual pack part of the publishing process rather than an optional set of templates.
Expanding the first release
Research showed that learners wanted to track progress after enrollment.
The team rejected a full learning management system for the first release. I designed the high-fidelity UI for a focused profile view that showed enrolled courses and current progress while keeping the scope controlled.
The education team also increased the amount of free material. I designed the in-house free courses and guides as an entry point into the wider catalogue.
Building the beta with Claude Code
The existing MVP, developer guidelines, and Claude Code made the two-month delivery possible.
Developers first gave us the technical rules and repository constraints. The two designers then worked in a shared repository and used Claude Code with Next.js to turn the approved Figma work into a functioning front end.
I built desktop, tablet, and mobile layouts across roughly 15-20 unique pages and more than 80 reusable components. The beta gave developers and QA engineers a working product to review before production.
I shipped that beta without checking each breakpoint as its own state. At tablet widths, it rendered a compressed desktop layout instead of the intended tablet design. The generated breakpoint logic was wrong, and I had not caught it before review. Several controls became too small for reliable touch input.
Engineering caught the issue during beta review. I handed over the intended tablet designs and related code. The developers corrected the breakpoint behavior manually before production and used the beta as the base for the released front end.
I checked the corrected tablet controls myself against a 44 by 44 CSS pixel minimum after the developer fix and confirmed that the tap targets met it.
I now treat each breakpoint as its own test state. I would add those checks throughout implementation instead of waiting for the full beta review.
Results
Academy launched in late March 2026 with 10 live courses.
During the first 30 days after launch, Academy generated a little over $12,000 in gross revenue before expenses against a $10,000 target. This was the main audience-building period. The launch used the company's existing audience and owned channels, with no paid acquisition.
I was the main designer for the campaign graphics, course thumbnails, promotion assets, and instructor packages behind that launch. The revenue was a team result supported by the visual and marketing system I led.
As of August 2026, Academy had 31 live courses and around 37,000 registered student accounts. The Lead of Education and Product Marketing Manager confirmed those figures.
What the research missed
We studied what people wanted to learn and how they chose a course. We had less evidence about what they were willing to pay.
That risk did not materialize during the first 30 days, when revenue beat the launch target. It still left the team without a strong research base for later pricing or subscription decisions. Pricing sat outside my design scope, but the missing evidence was relevant to the product work.
The sample came from a motivated education community. We needed more input from visitors who viewed a course and decided against paying.
What I would change
For the next pricing round, I would recruit non-buyers from both learner groups. I would ask about budget, past course purchases, and the point where the price became too high. I would compare per-course pricing with a subscription using conversion, revenue per visitor, and refund rate.
For the build, I would review tablet as a separate layout state and add breakpoint checks to each implementation review before beta.