# How to check whether your AI SDR's meetings actually happened

> Match every meeting an AI sales agent counted as booked to your own calendar and CRM, and check who booked it, who the prospect was, whether it was cancelled and who attended.

Published 2026-11-12 · Reviewed 2026-11-12. Canonical: https://attify.app/guides/check-ai-sdr-meetings-happened

Take the list of meetings your AI sales agent counted as booked, find each one in your own calendar and CRM, and check four things: that the AI booked it rather than one of your reps, that the person was a new prospect, that it was not cancelled or moved indefinitely, and that someone from the prospect's side attended. Then read a sample of the conversations that led to the meetings to confirm each was a sales meeting with a potential buyer.

The checks are those of the [Attify Audited Meeting Standard](https://attify.app/standards/sales-meetings) (v0.6), all of them counted, as for a vendor that claims held meetings. A vendor that claims meetings booked is audited only on who booked it, a new prospect, a rep stepping in and the conversation read, with its cancellations and no-shows shown beside the claim ([the claimed unit](https://attify.app/standards/sales-meetings#the-claimed-unit)).

## 1. Get the vendor's list, with ids

Ask the vendor for every meeting it counts as booked in the period, with the meeting's id in your calendar, CRM or scheduling tool. AI SDR vendors count at booking or earlier (a booking event, a status set by hand, a reply label, any meeting within a window after a touch; see [how vendors define a meeting](https://attify.app/compare/ai-sdr-meeting-definitions)), so the list is the claim, not the result.

Use a period long enough to settle: 90 days of first scheduled times, ending at least 7 days before you check.

## 2. Check each meeting in your own systems

| Check | A meeting fails when | Where to look |
|---|---|---|
| Not booked by the AI | The calendar or CRM shows a person on your team created it, or the prospect booked through your own page before any AI message | Calendar event creator; HubSpot record source; Salesforce created-by |
| Not a new prospect | Their company was already a customer, they had another meeting with your team in the previous 60 days, they are the vendor's own people, or it is a second claim for the same person | CRM lifecycle stage and deals; calendar history |
| Person stepped in | A rep wrote to or called the prospect between the AI's first message and the booking, or rebooked it after a cancellation | CRM activity timeline |
| Cancelled | It was cancelled, or not held within 14 days of its first scheduled time | Calendar event status; CRM outcome |
| No-show | Nobody from the prospect's side attended | Video-call participants (at least 10 minutes); HubSpot meeting outcome (Completed, No show); Calendly or Cal.com no-show marks |

HubSpot's meeting outcome has default values Scheduled, Completed, Rescheduled, No show and Canceled ([HubSpot](https://knowledge.hubspot.com/properties/hubspots-default-activity-properties)). Google Meet's conference records list each participant's join and leave times ([Google](https://developers.google.com/workspace/meet/api/reference/rest/v2/conferenceRecords.participants)).

Count each meeting at the first check it fails.

## 3. Read a sample of the conversations

For a random sample of the meetings that passed, read the conversation that led to the booking. Fail meetings with job seekers, recruiters, vendors pitching back, students, investors, journalists or competitors, existing customers asking for support, and anyone who declined or was paid to attend.

## 4. Compute and compare

- **Audited share** = meetings passing every check ÷ meetings counted as booked.
- **Your own team's baseline**: put your reps' meetings with outside people in the same period through the cancellation and no-show checks. Some of any gap is the prospects, not the AI.
- **Qualified**: of audited meetings, how many led to a new opportunity within 30 days of the meeting. Report it separately; it depends on your sellers too. Human SDR teams set quota on held meetings and then converted opportunities ([The Bridge Group, 2025](https://www.bridgegroupinc.com/research/2025-sdr-models-metrics-report-the-bridge-group)).

### Worked example

Synthetic example. 120 meetings counted as booked over 90 days.

| Reason a meeting failed (first check) | Meetings |
|---|---|
| A rep booked it | 6 |
| Not a new prospect (existing customer, recent meeting, duplicate) | 10 |
| A rep stepped in before the booking | 8 |
| Cancelled or not held within 14 days | 15 |
| No-show | 11 |
| Not a sales meeting, from a sample read of the remaining 70 (10.0%) | 7 |
| **Passed every check** | **63** |

Audited share 63 ÷ 120 = 52.5% of the meetings counted as booked. If the vendor bills $250 per booked meeting, spend is $30,000 and each audited meeting cost $476.19.

## Do it with Attify

Connect your CRM (HubSpot, Salesforce), calendar (Google Calendar, Microsoft 365) and scheduling tool (Calendly, Cal.com) on the [buyer dashboard](https://attify.app/buyers) to get the audited share every month, your own team's baseline beside it, and the [cost per audited meeting](https://attify.app/tools/cost-per-audited-outcome).

## Questions

### Our pilot booked 47 meetings and 4 became opportunities. Is that a failure?

First check how many of the 47 were held, with new prospects, without a rep stepping in. If 30 were held, 4 opportunities is 13.3% of held meetings. Compare that with your own reps' meetings with similar prospects in the same period, not with a vendor's benchmark.

### How do I report AI-booked meetings separately in HubSpot?

Use the record source or the creating user on each meeting: an AI agent writes as its own user, app or integration. Where the AI books through a rep's calendar, the vendor's list of booked meetings, with ids, is the only way to tell.

### The vendor counts reschedules as new meetings. How do I dedupe?

Treat a meeting moved to a new time as one meeting, and a cancelled meeting rebooked with the same person within 14 days as the same meeting. Count one per person per 60 days.

### What if nobody logs meeting outcomes?

Video-call participant records (Google Meet, Microsoft Teams) show who joined. Without them, and without a team that logs outcomes, attendance cannot be tested, and any held rate is a guess.

## Sources

- Attify: [Attify Audited Meeting Standard v0.6](https://attify.app/standards/sales-meetings), read 2026-11-09
- HubSpot: [HubSpot's default activity properties](https://knowledge.hubspot.com/properties/hubspots-default-activity-properties), read 2026-11-08
- Google: [conferenceRecords.participants (Google Meet REST API v2)](https://developers.google.com/workspace/meet/api/reference/rest/v2/conferenceRecords.participants), read 2026-11-08
- Google: [Events resource (Google Calendar API v3)](https://developers.google.com/workspace/calendar/api/v3/reference/events), read 2026-11-08
- The Bridge Group: [SDR Metrics: 2025 Edition](https://www.bridgegroupinc.com/research/2025-sdr-models-metrics-report-the-bridge-group), read 2026-11-08
