How KindWork Turned Interview Readiness into a Data-Driven Signal with Yoodli AI Roleplays

TL;DR

KindWork, a free workforce development program operated by Brooklyn Workforce Innovations, prepares low-income young adults in New York City for careers in customer experience and customer success. With only five weeks per cohort and a four-person program team, KindWork needed a way for fellows to practice interviews quickly, repeatedly, and against KindWork’s own frameworks rather than generic expectations. Using Yoodli’s customizable rubrics and AI roleplays, KindWork has enrolled roughly 90 learners across five cycles since spring 2025, with each fellow completing three to four interview rounds. Coaches now track interview readiness as an objective signal: in the most recent cycle, every fellow scoring 68 or higher on Yoodli reached an employment outcome, and 10 of 16 graduates either improved over time or remained consistently strong.

Background

KindWork has operated since 2019 with a clear mission: create an accessible pathway into technology through customer-facing roles that offer strong wages and career mobility, and prepare graduates to work alongside AI rather than compete against it. The program serves approximately 72 learners per fiscal year across four cycles, with 16 to 20+ fellows per cohort, and maintains an 85% graduation rate with in-sector placement rates in the mid-70s, competitive numbers for workforce development in tech.

Career readiness, not just technical training, is what fellows say they are most anxious about entering the program. And the interview process itself has changed: entry-level customer support candidates now regularly face AI interviewers as part of hiring, with three, four, or more rounds even for entry-level roles.

The Challenge

Before Yoodli, interview practice happened live with a coach or a peer. That worked, but with a five-week program, fellows needed to improve fast, and a four-person team could only run so many mock interviews. As Stephanie Tkach put it, running mock interviews for 25 people makes it hard to ensure every fellow gets feedback that is complete, actionable, and documented well enough to drive improvement the next time.

Other interview platforms didn’t fit. Their AI interviewed in a static way (ask a question, hear the answer, move on), which left fellows feeling uncomfortable and unauthentic. Questions were pitched at consulting or sales candidates, not the customer support roles KindWork trains for. Worst of all, generic platforms gave feedback that contradicted what KindWork taught in class, creating a disconnect that made them unusable.

The Solution

KindWork built its own interview frameworks directly into Yoodli and trained the AI to grade fellows against those frameworks, not a broad industry standard. Fellows practice roleplays introduced gradually across the program, starting with foundational questions like “tell me about yourself” and “why do you want to work here,” in a low-stakes environment at home after class. They can attempt a roleplay multiple times to hit a defined passing score, review recordings of past attempts, and improve through repetition. Each fellow completes three to four interview rounds, often with many attempts per round.

Critically, KindWork runs a human-in-the-loop coaching model. Coaches sit down with fellows after they receive Yoodli feedback to make sure they understand it and to add a human touch, like developing a story further. Scores, feedback, and improvement are tracked on the platform over time as the team’s way of assessing whether a fellow is ready to begin the job search in earnest.

“Yoodli has put us in the position to have AI help amplify our coaching, but not replace us. That’s really important for our team in wanting to maintain that human touch with our participants.” 

– Shani Watler, KindWork Program Director

Why Yoodli

• Customizable rubrics that match a bespoke curriculum. KindWork is one of very few customer support workforce development programs in the country. Stephanie defined what a one and a five look like based on the program’s own rubrics, used Yoodli’s built-in rewrite and preview features to refine how the AI assessed responses, and tested each roleplay herself before rolling it out.
• Easy iteration. The platform makes it simple to go back into roleplays and edit frameworks as the curriculum evolves or something feels off.
• Direct collaboration. Stephanie worked with Yoodli co-founder and Chief Customer Officer Esha Joshi to tailor the grading experience: “It felt like we were working on that together to help our fellows achieve the lives they dream of.”
• Practice that feels real. Dynamic roleplays replaced the static Q&A experience that had made fellows hesitant to practice with AI at all.

The Results

• ~90 learners enrolled across five cycles in roughly 15 months, with every cohort since spring 2025 using Yoodli
• Deep, repeated practice: the current cohort of 18 fellows has initiated 675 total practice sessions, averaging 37.5 practices per fellow as they refine their delivery
• A readiness signal coaches can act on: in the most recent cycle, fellows scoring 68 or higher on Yoodli practices all reached employment outcomes, while scores in the low 50s flagged early who needed more coaching time
• Measurable skill progression: 10 of 16 graduates in the last cycle either improved over time or remained consistently strong
• Significantly more feedback from the same four-person team: assignment-level feedback increased across the fellowship, down to the specific framework a fellow needs to work on
• Data-driven graduation standards: Fellows advance based on demonstrated skill growth, moving beyond intuition to objective performance metrics.
• Confidence that outlasts the program: alumni continue using Yoodli in their job search, and even fellows placed outside the tech sector report that the practice prepared them for interviews while still achieving wage progression

“We don’t want people to spend five weeks and graduate without clear evidence that they’re ready. We’ve been able to significantly increase the amount of assignment-level feedback we’re giving our fellows, and really scale our coaching as a result.” 

– Stephanie Tkach, KindWork Head of Career Services & Industry Partnerships

“Yoodli pushed me to practice in a way that felt challenging but useful, and the feedback helped me improve my responses over time. The interview simulation made the process feel real and prepared me to interview for the tech role I’m in now.”

–  Elise D., KindWork Alum, Winter 2026 Cohort

What’s Next

Small class sizes are core to KindWork’s model, and that won’t change. But Yoodli has helped the program stay nimble as screening and selection processes in tech grow more complex, and it could support growth in other ways over time, like potentially offering classes more frequently. Shani is already encouraging peer programs across Brooklyn Workforce Innovations, in and outside of tech, to experiment: “Give it a shot and see how you can make this tool help make your coaching more robust and more meaningful for your participants.” 

Want to see what Yoodli can do for your team?

Contact: sales@yoodli.ai  |  yoodli.ai