How do you know if Chloe is actually working, or just making noise?

Chloe can make a hundred calls a day and still not move your pipeline an inch. Activity isn't the same thing as working.

It's easy to glance at a dashboard, see a big call count, and assume things are going well. The real question is narrower and less flattering: are the right leads getting qualified correctly, and are the wrong ones getting filtered out instead of booked anyway, just at a faster pace than before?

This is worth sitting with for a minute, because it's uncomfortable in a useful way. A dashboard full of green numbers can hide a script that's quietly booking the wrong meetings at volume, and the only way to catch that is to look past the summary numbers and into the actual substance of what's happening on the calls.

Most founders default to checking the metric that's easiest to see first — total calls, or total meetings booked — because that's what the dashboard surfaces without any extra effort. Neither one tells you whether the underlying qualification logic is sound, and both can look perfectly healthy while something underneath is quietly and steadily going wrong.

What's the first thing to actually check?

Pull a sample of the meetings Chloe booked and see how many were actually qualified when a human looked closer at the transcript and the answers given. If reps are showing up to calls that shouldn't have been booked in the first place, the qualification logic — not the AI itself — needs work.

What does a healthy disposition breakdown look like?

Not every call should end in a meeting, and that's actually a good sign rather than a problem. If nearly everyone Chloe talks to gets booked, she's probably not qualifying — she's just scheduling anyone who answers the phone. A believable mix of qualified, not-now, and no is a far better sign than a suspiciously high booking rate that looks great on a dashboard but doesn't hold up under scrutiny.

How do you check the quality of the conversations themselves?

Listen to actual calls, not just summaries. Summaries can smooth over a call that was actually confusing or repetitive for the lead, rounding off the awkward parts into something that reads cleanly on paper. The transcript tells you what really happened, including the parts a summary would naturally leave out or simplify.

What should you track week over week?

The ratio of qualified to booked, and how often reps flag a meeting as a bad fit after the fact. A rising rate of rep complaints about bad meetings is the clearest early signal that something in the script or criteria has drifted out of alignment with reality.

Put this on a recurring weekly check rather than something you look at only when a problem surfaces on its own. Small drifts are easy to fix in a single sitting. Drifts that have compounded silently for two months are a much bigger project than they ever needed to be.

Who should actually be doing this checking?

The same person who owned the initial setup, ideally, since they already understand the intent behind the script and why certain decisions were made the way they were. Rotating this responsibility without a proper handoff tends to lose the context that makes the checking meaningful in the first place, and problems slip through as a result.

What's the real sign it's working?

Reps trust the meetings on their calendar without double-checking them first before every call. When that trust exists on a team, you know Chloe's actually qualifying, not just calling and hoping something sticks.