“Practical experience was the biggest gap in interviewer training. After interviewing AI candidates and reviewing the reports, many participants said they were developing a better feel for how to ask follow-up questions.”
— Kolmar Group HR representative
About
This is not a typical interview-training solution. An AI avatar plays the candidate, while team-lead interviewers ask questions as they would in a real interview and then receive feedback. The person being assessed is the interviewer, rather than the candidate.
Since July 2026, interviewers at five group companies, including Kolmar Holdings, have been using the service to support fairer and more accurate hiring.
Background
Interviewer training teaches questioning techniques, but it has often lacked someone to practice asking those questions with.
An interview ultimately involves making judgments about people. A plausible answer may reveal gaps after a single follow-up question. Learning to notice those gaps takes practice, beyond what a lecture alone can provide.
Kolmar wanted to help interviewers understand candidates more accurately and assess them against job requirements and the group's expectations for its people.
Solution
Interviewers begin by selecting a job posting. AI candidates who have applied for that role then appear. Each has fictional application documents tailored to the posting, allowing the interviewer to review them as they would real applications and choose a candidate. These candidates are newly created fictional people, not real individuals. Selecting the assessment factors to focus on completes the preparation.
When the interview begins, the AI candidate welcomes the interviewer on screen. Questions can be spoken or typed, and the candidate responds immediately. The interviewer can then use follow-up questions to explore experience and competencies. New voices were created for entry-level and experienced candidates. The service also provides time-based closing reminders and prompts to ask about presentation assignments.
After the interview, the interviewer simulates a hiring decision by choosing whether to accept or reject the candidate. Once that decision is submitted, a coaching report provides an overall rating and feedback for the next real interview.
Results
This was the first hiring and HR-training case to combine AI avatars and agents into a corporate learning service addressing the challenges Kolmar Group's HR team faced when running interviewer training.
There had been no one to practice with. Training could easily end with a lecture on questioning techniques. Now interviewers practice with AI candidates prepared for their job postings, asking questions and following up as they would in a real interview. At the kickoff demonstration, the option to choose a candidate's response style received the most positive feedback. Participants said that trying to identify a candidate's characteristics made it feel like real training.
It had been difficult to reflect on the interviewer's decisions. Real interviews rarely offer a chance to check whether questions were effective or judgments were correct. Interviewers can now review each session through a report containing question-level ratings, the hiring-decision result, and suggestions for better questions.
Job postings and assessment criteria differed across affiliates. Interviewers at each group company use their own access codes to practice with their company's postings and receive feedback based on shared assessment factors and report criteria. HR can operate group-wide training through one service and receive usage figures for each company.
Using real candidates' personal information had created an additional burden. Practice does not use actual applicants' information. AI candidates are fictional people created for the job posting, and interview conversations are deleted from the service after the report is generated.
Who This Is For
• HR teams at large companies and corporate groups seeking practice-based interviewer training
• Recruitment and HR development teams managing different postings and assessment criteria across affiliates
• Organizations seeking consistent judgment criteria among interviewers
Frequently Asked Questions
Q. Do interviewers need to enter a lot of information beforehand?
A. No. They select a job posting, candidate, and assessment factors from lists and begin immediately.
Q. How are the AI candidates created?
A. They are fictional candidates created for each job posting. Each comes with education details, a presentation assignment, and a personal statement.
Q. Are interview conversations stored?
A. No. Conversations are used only to generate the report and are then deleted. Interviewers can download and keep the report as a Word file.
Q. How do interviewers ask questions?
A. They speak or type a question, and the AI candidate responds immediately.
Q. What do interviewers receive afterward?
A. A Word report containing question-level ratings, an assessment of the hiring decision, and examples of alternative questions.
![[ESTsoft] ESTsoft provides Seven Eleven with its intelligent AI avatar “AI Mate” to test its usefulness in convenience store shopping!](https://framerusercontent.com/images/N9N0W8oRGGl0nZWld0DSqKJsdqs.webp?width=1440&height=900)
![[ESTsoft] Real-Time AI Avatar Service Perso Interactive Expands Across Public Services, Building Customized AI Services for Local Governments and Public Institutions](https://framerusercontent.com/images/Iie6Ol3RQYNZcXd8NsouLLX7TT8.webp?width=900&height=502)
