Get in touch
Home Overview2-Way Interview Features Trust All productsEight products, one data layer
Get in touch See the live rubric

Products › 2-Way Interview

A real conversation.
Scored like everything else.

Live rounds with your own team, on the same rubric

Your interviewer, the candidate, and the rubric on screen while they talk. The round is recorded, transcribed and scored against the same criteria as the async ones — so a final-round score sits on the same axis as a first-round one.

Book a demo See the live rubric

Your team runs it. The structure runs alongside.

The final round is where
the structure usually collapses.

Everything is structured until a human joins the call. Then it becomes a chat, a gut feeling, and four lines of notes that mean nothing to anyone else a week later. This round keeps the conversation and keeps the record.

  • The rubric is on screen while you are talking. So the criteria you agreed on are in front of the interviewer during the interview, rather than remembered afterwards.
  • Notes land against a criterion, not in a document. Which is the difference between evidence and a paragraph somebody has to interpret later.
  • Recorded and transcribed. The panel is no longer dependent on one person’s recall of what was said, and a disagreement can be settled by reading it back.
  • The same rubric as the avatar and recorded rounds. A live score and an async score are directly comparable, which is what makes a multi-stage process actually add up.
  • More than one interviewer, weighted. Two people on the call score independently against weighted criteria rather than talking each other into a consensus in the corridor.
Live round — 18:24 elapsed
RKR. Kaur · interviewerScoring
TWT. Wong · hiring managerScoring
System design4.1
Trade-offs3.5
Collaboration—
NoteTrade-offs 3.5 — “chose Postgres, could not say why not Mongo”
attached to the criterion · 14:02
Two interviewers scoring independently · weighted, not averaged in the corridor

A conversation that leaves a record.

Five stages, and the human part stays entirely human. The structure sits alongside the conversation rather than in the middle of it.

Scheduled
one slot, not forty
Both join
browser
Rubric on screen
as you talk
Recorded & scored
transcribed
You decide

What the 2-Way round does

The rubric, while you are still talking+
The criteria you agreed on sit on screen during the call, with each one scoreable as the conversation reaches it. That is a small change with a large effect: an interviewer who can see the rubric asks about the rubric. The alternative — scoring from memory twenty minutes after the call ended, between two other meetings — is where most structured interview programmes quietly stop being structured.
Notes that attach to a criterion+
A note is captured against the specific criterion it is about, timestamped to the moment in the call. “Chose Postgres, could not say why not Mongo” sitting under trade-offs at 14:02 is evidence. The same sentence in a shared document a week later is an anecdote.
Recorded and transcribed+
The round produces a recording and a speaker-attributed transcript, so a panel member who could not attend reads it rather than receiving a summary of it — and a disagreement about what somebody actually said is settled by reading, not by recall.
More than one interviewer, weighted+
Two or three people on the call score independently against weighted criteria. The scorecard aggregates them; nobody has to talk anybody round. It is also the mechanism that makes a dissenting score visible rather than absorbed into a group consensus.
Scored alongside the human score+
The AI scores the round against the same rubric, and the human scores sit beside it. The point is not to overrule the interviewer — it is to make a gap between the two visible, because a large one is usually worth a second look at either the candidate or the rubric.
Comparable with the async rounds+
This is the reason it matters that all three video rounds share a rubric. A candidate who took an avatar round at screening and a two-way round at final stage has two scores on one axis, so the pipeline shows a trajectory rather than two unrelated numbers.

The conversation is human.
So is the decision.

Your team runs this round. Three constraints on what sits alongside it.

A human makes the call

It scores and evidences alongside your interviewers. It does not reject anyone, and it does not overrule a human score — it sits next to one so a gap is visible.

One AI gateway, no loose ends

Every model call passes through a single choke point with schema validation, caching, per-tenant rate limits and spend caps. There is no second path to a provider.

Media behind signed links

Recordings and transcripts sit in private storage reached only through short-lived signed URLs, with lifecycle rules that expire media on a schedule.

And plainly. This round does not scale — it needs one of your people for its full length, which is exactly why it belongs at final stage and why the Avatar and Recorded rounds exist upstream of it. Both parties should know the call is being recorded, and consent handling is your obligation in your jurisdiction rather than something a tool can settle for you. We publish no accuracy figure for the AI score, and where it disagrees with your interviewer the correct response is to look again, not to pick the machine.

Conversational AI Avatar InterviewScreening at volume, conducted identically for every candidate.Recorded Video InterviewNo scheduling at all — they record in their own time, you review in yours.

Frequently asked questions

What is TalbotIQ's 2-Way Interview, and how is it different from an asynchronous video interview?+
It is a live round run by your own interviewer with the candidate, with the criteria you agreed on shown on screen while they talk. The round is recorded, transcribed and scored against the same criteria as the asynchronous rounds, so a final-round score sits on the same axis as a first-round one.
Does the AI make the hiring decision in a live interview, or does the interviewer?+
Your team does. The AI scores the round against the same rubric and the human scores sit beside it, but the point is to make a gap between the two visible, not to overrule the interviewer. TalbotIQ publishes no accuracy figure for the AI score, and where it disagrees with your interviewer the stated response is to look again, not to pick the machine.
Can a panel interview be scored by more than one interviewer independently?+
Yes. Two or three people on the call score independently against weighted criteria, and the scorecard aggregates their scores rather than requiring the panel to agree a consensus, so a dissenting score stays visible.
Is the live interview recorded, and can a panel member who missed the call review it?+
Yes. The round produces a recording and a speaker-attributed transcript, so a panel member who could not attend reads it rather than a summary, and a disagreement about what was said is settled by reading it back.
How do interviewers take notes during a structured live interview?+
Notes are captured against the specific rubric criterion they relate to and timestamped to the moment in the call. The page's example is "Chose Postgres, could not say why not Mongo" filed under trade-offs at 14:02.
Where are interview recordings and transcripts stored, and are they kept indefinitely?+
Recordings and transcripts sit in private storage that is reached only through short-lived signed URLs, and lifecycle rules expire the media on a schedule.

Published 9 September 2026 · Updated 9 September 2026 · Written by the TalbotIQ team

Ready to see it?

Tell us how your final rounds run today and who sits on them, and we will show you the same round with a rubric attached.

hello@talbotiq.com
0 / 250
Request demo

Keep the conversation. Keep the record.

Book a 30-minute demo and we will run a live round with the rubric on screen, so you can see what lands on the scorecard.

Book a demo Talk to us

Malaysia-based team · Response within one business day · No obligation