Skill.re
AI-Assisted Interview Preparation, Scheduling, and Synthesis
The interview stage is where recruiting becomes genuinely human. Conversation, judgment, relationship-building. AI's role here is more constrained than in earlier stages. Some AI applications are genuinely useful (scheduling). Others are problematic (analyzing personality from video). This lesson clarifies what works and what crosses ethical lines.
Interview Scheduling: One of AI's Success Stories
Finding a time that works for a candidate and multiple interviewers is genuinely tedious. AI handles this well.
How Scheduling AI Works
The system integrates with calendars (Outlook, Google Calendar). It reads availability, finds overlapping times, sends invitations, handles timezone conversion, manages follow-up if times don't work.
This is rule-based logic. No magic. But valuable magic.
Why This Works
- Eliminates back-and-forth coordination
- Respects people's calendars accurately
- Removes scheduling as a friction point
- Candidate experience improves (faster scheduling, fewer delays)
- Genuinely saves recruiter and hiring manager time
Limitations (Small But Real)
- Assumes calendars are accurate (they often aren't)
- Can't account for context (someone might unblock time for an important candidate)
- Might schedule at inconvenient times without understanding context
- Doesn't handle unusual timezone cases smoothly
Verdict: Use AI scheduling. This is genuinely useful, low-risk, and improves candidate experience. One of the few places where "AI" is actually solving a real problem well.
Interview Preparation: Generating Interview Guides
Some AI tools help prepare interviews. "What should I assess in a data scientist interview?" The AI suggests dimensions and questions.
What Works
- Generating starting frameworks
- Suggesting relevant assessment dimensions (technical skills, communication, problem-solving, etc.)
- Creating question templates
- Structuring interview guides
The Reality
The AI is synthesizing interview guides from thousands of job descriptions and interview templates it's learned from. It's not thinking about your specific role, your team, your unique needs. It's pattern-matching.
Good use: Use the AI output as a starting point. Then customize. Your actual role requirements matter more than the AI's generic framework.
Bad use: Using the AI output as your interview guide without customizing. Your interviewers will ask generic questions about generic competencies instead of questions specific to your role.
Video Interview Analysis: Where AI Gets Problematic
Some platforms claim to analyze video interviews using AI. They generate scores for communication, confidence, honesty, personality traits.
How These Tools Work
They use computer vision to track: facial expressions, eye contact, head movements, hand gestures. They use speech recognition to analyze: tone, pace, fillers ("um," "uh"), vocabulary, sentiment. Then they generate scores.
The Problem
This is pseudoscience. There's no reliable correlation between facial expressions or speech patterns and job performance, honesty, or personality traits. A tool that claims "this candidate is confident" based on eye contact is making unfounded inferences.
And it's deeply biased. Cultural differences in eye contact, communication style, hand gestures are misinterpreted as personality traits. Someone from a culture with less direct eye contact gets marked "lacks confidence." Someone who speaks slowly and deliberately gets marked "uncertain." Someone with an accent gets marked differently than native speakers.
The Legal Risk
Video analysis tools can create EEOC liability. If the tool systematically scores candidates from certain demographics differently, you have disparate impact. The FTC has already warned about interview analysis tools making discriminatory inferences.
Recommendation: Avoid video analysis tools that generate personality or confidence scores. They're unreliable, biased, and legally risky. If you must record interviews, use the video for reference only. Make hiring decisions based on what candidates say and do, not on algorithmic inferences about their psychology.
Interview Synthesis and Note-Taking
Some platforms transcribe interviews and generate summaries. AI extracts key points, creates highlights, structures notes.
What Works
- Transcribing audio/video accurately (speech recognition has improved dramatically)
- Extracting key quotes and topics
- Structuring interview notes for later review
- Making searchable records
Limitations
- Hallucinations in transcription (AI misheard words sometimes)
- Misses context and nuance when summarizing
- Can misrepresent what was actually said if you rely only on the summary
Best Practice
Use AI for transcription and note-taking. But always verify transcripts against the actual interview recording before relying on them for hiring decisions. Don't trust summaries alone. Read the full transcript. Listen to key sections.
Assessment Tools: When They Work, When They Don't
Some platforms include skills assessments: "This candidate's problem-solving score is 75/100." Or: "Communication assessment: strong."
Technical Skills Assessment
These can work: A coding challenge evaluated by an AI system (does the code run? does it handle edge cases?). A data analysis task where AI checks the SQL or Python. These are objective, verifiable assessments.
These don't work: AI assessing "quality of thinking" or "elegance of solution." That requires human judgment.
Soft Skills Assessment
These generally don't work: AI assessing communication, leadership, teamwork from an interview. An AI can identify what topics the candidate discussed. But assessing whether they're a good communicator requires context, judgment, and understanding of your actual role needs.
The Temptation and the Reality
There's a temptation to trust AI assessment scores because they look objective. "75/100" sounds more rigorous than "seemed pretty good." But the rigor is an illusion. The assessment might be wrong. And you've outsourced judgment to a system that can't actually assess what matters.
Impact on Candidate Experience
How does interview-stage AI affect candidates?
Positive
- Faster scheduling improves experience
- Clear communication about process builds trust
- Structured interviews are fair and professional
Negative
- Video analysis tools create anxiety and distrust (candidates don't know they're being psychoanalyzed)
- Automated rejection based on AI assessment feels impersonal
- Lack of transparency about AI use erodes trust
Transparency matters: If you use AI in interviews, candidates should know. "We use automated scheduling" is fine. "We analyze your facial expressions to assess your personality" is not—it's invasive and unreliable. Consider whether using the tool is worth the candidate trust cost.
Key Takeaway
Key Takeaway
Interview-stage AI varies wildly in usefulness. Scheduling? Great. Use it. Preparation frameworks? Fine, use as a starting point. Video analysis for personality assessment? Problematic, unreliable, and legally risky. Avoid. Interview synthesis and transcription? Use, but verify against source material. Assessment scores? Technical skills can work; soft skills assessment doesn't. Remember: the interview is where recruiting becomes human. Use AI to reduce friction and handle logistics. Don't use it to replace human judgment about communication, judgment, and fit.
Frequently Asked Questions
Is it ever acceptable to use video interview analysis tools for assessment?
Not if they're assessing personality, confidence, or soft skills from facial expressions and tone. These tools are pseudoscientific and biased. However, using video for reference—re-watching key moments, transcription, extracting what the candidate actually said—is fine. The analysis should come from a human interviewer who understood the context, not from an algorithm reading microexpressions.
Can we use AI-generated interview guides as-is, or should we always customize them?
Always customize. AI-generated guides are generic frameworks based on average hiring patterns, not your specific role, team, or needs. Use them as starting points, but have a hiring manager (or someone who deeply understands the role) edit and customize. Replace generic competency examples with examples specific to your actual work. Your interview guide should reflect what you're really hiring for, not what average companies hire for.
If we're using interview transcription AI, how do we handle errors?
Make transcription verification a standard practice. Have someone (preferably the interviewer) review the transcript against the recording for key topics and quotes. Errors to watch for: misheard technical terms, names, company names, and numbers. These matter in hiring decisions. For hiring decisions, rely on the actual recording and transcript combined, not the AI summary alone. If you reference something from the transcript in your notes, verify it was actually said.
Should we tell candidates we're using AI in the interview process?
Yes, for transparency and candidate trust. "We use automated scheduling to coordinate interview times" is transparent and reasonable. "We analyze your speech patterns and facial expressions to assess your personality" is something candidates should know about—and frankly, you should reconsider using. Transparency doesn't just affect candidate trust; it affects how you think about your own tools. If you wouldn't want to tell candidates you're using a tool, that's a signal you shouldn't be using it.
Can we use AI to identify candidates who might not accept an offer?
Tools that claim to predict offer acceptance are unreliable. Candidates' job satisfaction and openness change over time—an AI can't predict these from an interview. Rather than trying to predict acceptance, focus on genuine engagement: understanding what the candidate actually cares about, addressing their concerns, negotiating offers that reflect market value. The best predictor of acceptance is making a genuine offer that meets the candidate's actual needs, not an algorithm's guess about their psychology.