What happens when high school students use an AI career coach in class? In spring 2026, CareerVillage and Leanlab Education ran a pilot study to find out.
Personalized, one-on-one career support is something a single teacher or counselor can't give every student. CareerVillage and Leanlab Education explored whether Coach, CareerVillage's AI career coaching tool, could help fill that gap in real high school classrooms, and what it would take to use it well. Five teachers put it to the test.
Students did more to prepare for their futures: they researched careers, worked on resumes, practiced interviews and talked with adults about next steps.
Testing Coach in real high school classrooms
Beginning in April 2026, teachers at three schools used Coach with one class each, in grades 9 through 12. Most were classes where post-secondary topics already came up, or advisory periods.
The research questions this study set out to answer were:
Do students show improvement in career outcomes, such as confidence and behavior?
Do educators report increased career-advising capacity after using Coach?
The five teachers who finished the study assigned one or two Coach activities per week for four to seven weeks, and 186 students completed surveys before and after the study. Students answered 13 questions about career confidence and 16 about how often they did career-readiness tasks. Teachers took surveys before, midway and after, and sat for an interview about how it went.
The study used both survey and interview data, and it had no comparison group, meaning the outcomes are best read as early, directional evidence rather than proof that Coach caused the change.
Students took more career-prep actions
Students' average career-readiness behavior score rose from 2.0 to 2.3 on a 0 to 4 scale (a statistically significant, small-to-medium effect). The increase held up after accounting for class size, differences between teachers and each student's starting score. Nearly every student improved on at least one survey item, which means students were practicing career-readiness tasks more often. Results varied by group:
Grade: 10th graders drove the increase in behavior scores, followed by 11th and 12th graders. 9th graders showed no significant increase. Confidence did not vary by grade.
Starting point: Students who were least active at the start gained the most. Students with a clear plan for after high school, especially a trade or internship, also benefited most.
In self-reports, about 54% of students used Coach weekly or more, 26% used it once or twice, and 11% never used it.
Coach was strongest on concrete tasks and personal questions
Teachers said Coach worked best on concrete, action-oriented tasks: building and revising resumes, practicing interviews, researching school, GPA and certification requirements, and exploring careers from strengths to subjects to pathways. It did less to change how students think about themselves. They also pointed to two things a single counselor can't easily give every student: personalized, conversational one-on-one support, and a place to ask questions students are afraid to ask a teacher. Students who struggle with traditional research, including students with IEPs, could talk with Coach instead and finish their projects.
Several students widened their options. One student turned a single idea into several careers and backup plans, and some moved from college to trade school after exploring their options. A 12th grade teacher described what happened for seniors who were torn between paths:
Those who were kind of up in the air about maybe 3 different options, they left more secure about what option they wanted to pursue, and what routes they can [use to] get there.
Teachers did more advising, with mixed effects on confidence
All five teachers reported doing more career-related activities with students while using Coach. Whether that translated into feeling more capable was less clear. In interviews, 2 of 5 teachers (40%) described a clear increase in their career-advising capacity.
Teachers' average self-efficacy score, based on 10 items, stayed flat. That is partly because all five teachers already scored above neutral before the study. Two of the teachers' scores decreased, both of whom also reported poor onboarding or a poor fit with their class.
Setup and fit made the difference. Coach worked better when it fit easily into a class, and worse when onboarding was rough or the subject was a poor match. This points to a limitation of an independent evaluation like this one. When the tool is evaluated in isolation, it misses a critical factor: the Coach team’s commitment to empowering partner organizations with custom support. To maximize impact, there needs to be a dedicated partner success function to that supports onboarding and implementation tailored to the needs of each classroom or program.
The teachers also used what students discussed with Coach to start classroom conversations they might not otherwise have had. One 11th grade teacher said Coach lightened the load:
“There are lots of things in Coach which can be done by the students themselves, and require less class time... that reduced a kind of workload for me.”
The same teacher said Coach left them "in a better position to help my students in achieving their goals."
Two moments from the classroom
Two stories from teacher interviews show what this looked like in practice.
A student changed his plan for after high school. A 9th grade teacher described a student who started the study set on going to college. After talking with Coach, he decided he wanted to attend a trade school and could explain why. She said:
“[A student told me] ‘At first I wanted to go to college, but now I really want to go to a trade school'... just seeing that they understand how their career connects to the real world, as far as next steps for where they need to be.”
Coach helped this student understand a fuller picture of his options to help guide smart decisions toward his future. Ninth graders' behavior scores did not rise significantly, but shifts like this one, in how students think about their future, weren’t fully captured by the measured behavior changes. It’s a signal worth watching and considering for a future study design.
Coach flagged a student's question to his teacher. A student asked Coach about prosthetics. Coach did not recognize the question as related to the course, so it sent the message to the student's teacher. She asked the student what he meant, since she first wondered whether he was thinking of mortuary work. He told her he wanted to fit people with prosthetic limbs. In 10 years of teaching, she had never heard a student name that career, and it led to a class-wide conversation.
In this case, Coach flagging a student’s question to a teacher is an example of Coach augmenting student-to-teacher relationships building. Coach did not try to be the expert in the room. By routing the question back to a human, it gave the teacher the opening to see her student more fully and start a deeper conversation, where everyone learned something.
Challenges rolling out Coach
The friction from the study came from getting started, fitting Coach into the school day and keeping students engaged.
Keeping onboarding quick. Account setup and joining a class with an invite code took 5 to 10 minutes for students who struggled. A few never got on because of district email policies blocking sign-in. With limited class time, every minute spent on setup is a minute not spent on career exploration, so sorting out access with each school before launch matters.
→ Pairing the tool with onboarding support. Teachers’ experiences hinged on set up and fit. The lack of support that Coach’s partner success team typically provides can make successful onboarding and program implementation more challenging.
Engaging students who aren't already motivated. Participation concentrated among students who were already motivated, and students who were checked out stayed checked out. Reaching those students will take more than access to the tool.
→ Starting earlier in the school year. Starting in the spring reduced how seriously students took it, especially juniors and seniors. Beginning in the fall or winter would give students more time and more reason to engage.
Managing AI fatigue. Some students reported being tired of being asked to try new AI tools, and they skimmed or clicked through long AI responses. Coach's opening questions allowed for personalization from the start, but for some high school students who aren’t sure about their futures, open ended questions felt nerve-wracking.
→Finding the balance between early personalization and a comfortable first experience is key.
Students who used Coach did more to prepare for life after high school, especially students who were doing little beforehand. That is a promising early signal, and we want to understand it better. A larger study across many classrooms, one that compares students who use Coach with students who don't, could show whether Coach is behind these changes. It could also explore whether Coach helps students feel more prepared to make decisions about their next steps, whether it helps students at schools with little or no counseling support, and how it affects the confidence of the educators who advise them.
Takeaways for schools
Teachers told us that setup, timing and fit shaped how well Coach worked. For any school weighing an AI career coaching tool, a few questions are worth asking, along with one thing the field should watch:
When and where will it start? Rolling out in the spring dampened seriousness, especially for juniors and seniors. Starting in the fall or winter, in advisory or career classes where it fits best, gives it more room to work.
Is it tied to what teachers already teach? Most teachers assigned Coach as individual work. Teachers who built it into existing assignments and added a class discussion saw the most benefit, and clear task-completion goals like do-nows and exit tickets would give others more ways in.
Who will it reach? Improvement clustered among students who were already motivated, so schools will need easier ways in for the students who tune out.
Where does the human stay in the loop? AI tools should hand moments back to teachers, make student interests visible to them, and fill gaps for students who have no one to ask rather than replace relationships that already exist.
The Coach pilot is one data point, not the final word. Five classrooms can't tell us whether Coach works at scale, but they suggest that the interesting question for the field is not just whether to use AI for career navigation. It is how to implement it so that it helps students and the teachers who support them.
Join the conversation
We are continuing to study how AI career coaching can support students and educators. If you are a school leader, educator or funder thinking about the same questions, we would love to connect.
