Overview

The AI Companion Dashboard is a native, read-only dashboard that reports on the AI Interview Companion. It answers three questions in one place: how widely the Companion is being used, how reliably each Companion session completes, and whether interviews that used the Companion end up with better feedback outcomes than interviews that didn't. Everything on it is reporting over activity that already happened — opening the dashboard changes nothing about how interviews are scheduled or how the Companion behaves.

Why it matters

• You can see adoption as it happens. The share of scheduled interviews that had the Companion enabled is shown as a single number and as a weekly trend, so a rollout that's stalling is visible while there's still time to act on it.
• You can check whether the Companion is earning its place. Feedback submission rate and turnaround time are reported side by side for interviews with and without the Companion, so the comparison is built into the dashboard rather than assembled by hand.
• You can tell where a session failed. The funnel separates sessions that were scheduled, joined, recorded, and produced feedback, so a shortfall points at a stage instead of leaving you to guess.

How it works

1. Open Insights and find the AI Companion dashboard in the left rail, under the Agentic Platform group that also holds the AI Interviewer dashboard. The rail's Home tab lists your dashboards, and dashboards you've bookmarked appear under Bookmarks. The dashboard has a single tab, Overview, holding all thirteen widgets. If it isn't listed, it hasn't been turned on for your organization yet — contact your Eightfold representative to request it.
2. Set Date range, the filter that applies across the dashboard. Every widget respects it, but they don't all date from the same event: adoption widgets filter on the interview's meeting time, feedback widgets filter on when the feedback record was created, and the opt-out and funnel widgets filter on when the Companion was scheduled to join. A single range can therefore cover slightly different populations across the three groups, which matters most at the edges of a narrow range.
3. Read the adoption group first. Total Interviews Scheduled counts every interview in range, with or without the Companion. Total Companion Interviews Scheduled counts those scheduled with the AI Companion enabled. Companion Adoption Rate % is the second divided by the first, and Weekly Trend in Companion Adoption Rate % plots that same figure week by week.
4. Check Candidate Opt Out Rate % alongside adoption. It's the share of Companion sessions where the candidate opted out after the interview was scheduled. Read it as a trust signal — a rising line is worth investigating even while adoption climbs, because it counts people actively declining the Companion.
5. Work down the AI Companion Interview Funnel to find reliability problems. Its four stages are Scheduled to Join, Joined, Recorded, and Feedback Generated, each a subset of the one before. A large gap between two adjacent stages localises the failure: sessions lost between Joined and Recorded point at the meeting itself, while sessions lost between Recorded and Feedback Generated point at feedback generation.
6. Read Feedback Outcomes by Companion Usage for the comparison that matters most. It's a table split into two cohorts, labelled Companion and No Companion, and for each it reports total feedbacks, feedbacks submitted, feedback submission percentage, and average time to submit in hours. Weekly Trend in Feedback Submission % and Weekly Trend in Feedback TAT plot those last two over time, keeping both cohorts on the chart so any divergence is visible.
7. Use the three Companion-only feedback widgets to size the drafting workload. Total Profile Feedbacks Filled by Companion counts feedback the Companion drafted after an interview. Profile Feedbacks Submitted Once Filled by Companion counts how many of those the interviewer went on to submit. Direct Ship Rate % reports the share of submitted Companion-drafted feedback that scored as needing no changes.
8. Treat Interviewer Time Saved by Companion as an estimate, not a measurement. It multiplies the number of Companion-drafted feedback records by a flat ten minutes each. It's a consistent way to express drafting volume in time, and it moves only when that volume moves — it doesn't observe how long anyone actually spent.
9. To take the underlying rows away, open the widget's options menu and select Export Raw Data. In Select columns to export, choose the columns you want, then select Export. The file is prepared in the background — you can access it in notifications.
10. Check for a staleness banner before you act on a number. If the data behind the dashboard hasn't refreshed recently, a banner reads ‘A dataset used by this dashboard has not been successfully refreshed in the past 48 hours’. Treat the figures as out of date until it clears. A widget reading No data available means no matching activity in your date range, which usually means the range is too narrow or predates your Companion rollout.

FAQs

Q: Why don't these interview totals match my other interview reports?

A: Three sets of interviews are left out of every count here. Slots that were cancelled, declined, or deleted are removed so withdrawn interviews don't inflate the figures. Interviews with no video conferencing tool are excluded, because the Companion can't attend one. And interviews run by the AI Interviewer are excluded, so the numbers describe human-led interviews only. Expect these totals to sit below a raw count of scheduled interviews.

Q: If a candidate opts out, does that interview still count as a Companion interview?

A: Yes. Adoption is measured at scheduling, and an opt-out happens afterwards, so the interview still counts in Total Companion Interviews Scheduled and in Companion Adoption Rate %. That's why Candidate Opt Out Rate % is worth reading next to adoption rather than instead of it — adoption tells you what was intended, the opt-out rate tells you what candidates did about it.

Q: Does Direct Ship Rate % cover everything the Companion drafted?

A: No. It's calculated only over submitted Companion-drafted feedback that carries a quality score. Feedback without one is left out of both the top and the bottom of the calculation, so the percentage describes the scored subset, not every draft. Read it alongside Total Profile Feedbacks Filled by Companion to see how much drafting sits behind it.

Q: Is feedback turnaround time an average or a median?

A: An average, reported in whole hours. Averages are pulled upward by a small number of very late submissions, so a single slow month can move the number more than the typical experience changed. The weekly trend is the more reliable read — a sustained move there means something real, where a single period's jump often doesn't.

Q: Will everyone in my organization see the same numbers?

A: No. The dashboard respects your existing analytics access, so each person sees the activity they're already entitled to see. A leader with organization-wide access and a recruiter working a handful of positions will open the same dashboard and read different totals, which is worth remembering before comparing figures with a colleague.