PBX cuts Cedars-Sinai test development time by 83%

Cedars-Sinai is a non-profit, academic healthcare organization that ranks as the #1 hospital in California, and in the top five hospitals across the United States for Gastroenterology, Orthopaedics, and lung surgeries. In 2025, they served millions of patients across California and ran over 3,000 combined research projects and clinical trials. 

Like many healthcare organizations, Cedars-Sinai prioritizes improving digital access and experiences for its patients, but they ran into familiar roadblocks: lack of developer resources, more ideas than can be reasonably tested, and challenges in prioritizing ideas and processes.

Cedars-Sinai joined Kameleoon early in 2025 looking to use Kameleoon's 100% HIPAA-compliant solutions to these problems. Then, in 2025, everything changed when Prompt-based Experimentation moved them drastically closer to that goal.

{{quote-2}}

Scaling with a team of one

Chris Strobl is the Senior Optimization Engineer for Cedars-Sinai, and although he works with designers, product managers, marketing, sales, and engineering teams, he runs Cedars-Sinai’s experiments largely on his own. 

With Kameleoon’s help, Strobl was running three experiments every month for Cedars-Sinai, taking advantage of our integration with Jira and working with designers through Figma. 

His process is one that is familiar for a lot of experimentation leaders:

1. Ideate and vet
2. Send to dev
3. QA
4. Launch
5. Monitor

This process works, but comes with a notable drawback: time required. Hand-coding variants take a long time to produce; achieving a pixel-perfect layout “can take forever,” according to Strobl; and building alternative variants in the QA phase slows down experimentation.

Using Prompt-based Experimentation (PBX) to scale

Kameleoon’s PBX allows Cedars-Sinai to generate experiences from prompts alone, resulting in significant time savings that allow teams to work in parallel and establish strong frameworks while AI takes on more tedious tasks automatically. Strobl uploads mockups to create a solid starting point and rapidly iterates, progressively enhancing variations until they’re ready for standardization.

Cedars-Sinai began exploring PBX: experimenting with prompt styles (long vs. short, descriptive vs. general, single vs. multiple), iterating on test variations, variation complexity, and external tool integrations for visitor and conversions tracking. 

A hybrid approach: AI-empowerment

Using PBX, Cedars-Sinai could drop the time requirement for a simple test by half—from six hours to three, and increase the total annual number of tests run from 36 to 45, a 25% increase.

With so many stakeholders, however, Cedars-Sinai quickly discovered that a hybrid approach worked even better. Their process:

1. Use AI to generate a foundation, covering around 80% of the work
2. Take the foundation to stakeholders and determine a “time:perfection” ratio
3. Based on that ratio, identify how much manual effort vs. prompt effort should go into each variant

This hybrid model, in which tests are initially generated through PBX, refined through additional prompts if needed, and then completed manually, led to even greater savings: a 67% reduction in time to complete simple tests, a 20% reduction for complex tests, and a new average of 60 tests completable per year—a 67% increase across the whole year. 

{{quote-1}}

PBX successfully handled complex technical constraints at Cedars-Sinai, including a single-page app architecture and asynchronous, producing readable and maintainable code (meaningful variable names, small reusable functions). In all, PBX could get tests 80-95% of the way to a production-ready variant, dramatically improving CRO productivity at Cedars-Sinai. 

‍

Want to use PBX to increase your own tests per year average? 

{{cta-block}}