Meet Iris (churn hides in plain sight)

Lesson 8 of 9 · Module 02 · The Agent Team: Catch What GHL Hides

Hugo catches the things that break. Iris catches the things that just sit there.

Clients almost never fire you over one loud failure. They leave quietly. A new contact comes in and nobody onboards them. A subscription gets cancelled and nobody notices until the revenue report. A deal sits untouched in a pipeline for two months and the prospect forgets you exist. None of that throws an error. The account looks fine right up until the cancellation call.

Iris watches every account for three signals:

Onboarding gaps. Contacts created in the last seven days that nobody has tagged as onboarded. The "we'll get to them" pile.
At-risk subscriptions. Anything cancelled or past due. Money already walking out the door.
Stale opportunities. Open deals untouched for fourteen days or more. The pipeline you stopped looking at.

She only reads. Contacts, subscriptions, opportunities, pipelines. No writes, ever.

Today she ran live for the first time, across my own agency. One command, every registered sub-account, about ten seconds. The roll-up line came back:

Accounts: 18   Need attention: 13   Duration: 9949ms

Eleven accounts got fully scanned. Five came back healthy. Six did not. Here is what she found, names withheld.

A client clinic account is carrying 100 stale open opportunities, nearly all idle about 58 days in the same stage of a launch-event pipeline. Two months of silence on an entire pipeline. And 100 is the query cap, so the real number may be higher.

A product account fired all three signals at once. The one that stung: deals at $97 each parked in a Won stage, idle 30 to 54 days. Won means somebody said yes. Then nobody talked to them for a month or more. That is how a yes turns back into a no.

And twenty onboarding contacts flagged across five accounts. Some were real people, days old, that nobody had touched. Some, honestly, were spam. One supplement-brand account's onboarding flags were mostly no-reply addresses and promo senders. The signal works mechanically, but it needs a filter. I am telling you that because this is one real run, not a polished case study.

The other rough edge: seven of the eighteen accounts got skipped because they are not registered for monitoring yet. Iris logged each one as an error and kept sweeping instead of crashing. Good behavior, still a gap to close.

What is yours, and what is coming

Every read Iris did is a question you can already ask the tool in plain English. Contacts created recently. Subscriptions that got cancelled. Opportunities nobody has touched. Iris is those questions asked on a schedule, across every account, rolled into one report.

Same rule as Vera and Hugo: proven one real run at a time, claimed only after it runs. The sweep above is real. The always-on version is what we are building toward. The next lesson is the part you can run today.