Roche CX Lab Recruitment Automation

An MVP for Roche CX Lab that explores how agentic AI can support participant recruitment for usability studies. Built in collaboration with IBM and HSLU, the project turns a fragmented manual process into a structured workflow concept for intake, participant matching, scheduling, communication, attendance tracking and report follow-up.

Build an MVP that shows how the CX Lab participant recruitment process could be automated without losing control. The concept should reduce the happy-path workflow from around 5 hours to around 5 minutes, keep participant handling structured, support better matching and make the experience smoother for researchers, recruiters and participants.

The value is not only speed. In MVP testing, the happy path moved from around 5 hours to around 5 minutes. The larger value is the controlled operating model: structured intake, clear workflow states, automated recruitment support and human checks where automation should not decide alone.

A real Roche use case from the Agentic AI challenge

This project came from an Agentic AI challenge with IBM, HSLU and Roche. The client was Roche CX Lab, an internal team that helps product teams validate designs with real user feedback. A big part of that work is finding and managing the right test participants.

The existing recruitment process worked, but it required a lot of manual coordination across different tools. In MVP testing, the happy path was reduced from around 5 hours to around 5 minutes by turning intake, recruitment routing, participant matching, scheduling, communication and follow-up into one controlled flow.

Built as an MVP, not presented as production software

This case study describes an MVP and validated concept. It is not presented as a live production product. The goal was to prove the workflow design, the agent responsibilities and the integration pattern.

The 5-hour to 5-minute reduction refers to the tested happy path inside the MVP. Edge cases, missing information, no-match situations and attendance confirmation still require controlled pauses or human involvement.

Supervisor-led agentic workflow

From project request to final report follow-up

The pain point was coordination, not only time

The recruitment process crosses many small steps: collecting requirements, searching the participant pool, preparing outreach, managing screener responses, checking calendars, sending invitations, replacing declines, tracking attendance and sending follow-up messages.

In MVP testing, the happy path moved from around 5 hours to around 5 minutes. The bigger improvement is that those steps become one visible workflow instead of a chain of manual reminders across tools.

Targeted recruitment or broad outreach

What the MVP automates

Automation stops where judgment is needed

The MVP was not designed to remove humans completely. It automates repetitive coordination, but it still pauses when the request is incomplete, no reliable match is found, scheduling needs a decision or attendance must be confirmed.

That balance matters in a research environment. Agents handle routine work. Humans stay responsible for context, quality and accountability.

What the MVP demonstrated