Exit criteria and leaving a sequence

What it does

Exit criteria disconnect a record from a sequence automatically part-way through, so the remaining steps stop appearing. In the current product this is a single switch: when a customer replies to any email in the sequence, the record comes off it.

Key facts

The setting is the Exit a sequence toggle on the Sequence Options pane, reached by selecting Exit criteria in the sequence designer.
One trigger only: the customer replies to any email in the sequence. There is no builder for arbitrary exit conditions.
The sequence must be in an inactive or revision state to change it. You can't flip the toggle on a live sequence without going through the edit or deactivate route first.
At least one email activity step must exist in the sequence before the option means anything.
Email engagement must be configured in the organisation, because reply detection is an email engagement capability rather than a sequence one.
When the criteria are met the record is disconnected from the sequence and stops receiving further steps.
Exit criteria are a property of the sequence, so they apply to every record connected to it. There is no per-record variation.
Configuring it needs the administration and customisation roles; using it needs the primary sales roles.
It is generally available, not preview.
If the option isn't visible, the usual causes are licensing, security role, the feature not being turned on, or the organisation running a custom app rather than Sales Hub.

When to use / skip

Turn it on for any sequence whose whole purpose is to get a reply. Prospecting cadences, chase sequences, re-engagement — in all of these, a reply means the outreach worked and continuing to send scheduled follow-ups makes the organisation look like it isn't reading its own inbox. That is a genuine reputational risk, and it's the single most common complaint about cadence tooling generally.

Skip it where the sequence isn't really an outreach campaign. An onboarding checklist, an internal handover cadence, a structured discovery process — a customer replying to one email in those doesn't mean the process is finished, and exiting early leaves the record half-processed with no obvious signal that anything is wrong.

Where you need something more nuanced than "any reply" — exit when the lead is disqualified, exit when the opportunity closes, exit when a field flips — you're not going to get it from this toggle. Use a condition step that branches to Advance to another sequence, or handle disconnection from a Power Automate flow. Set the client's expectations on that early, because "exit criteria" sounds far more configurable than it is.

Configuration decisions

Which sequences are reply-terminating and which aren't, decided per sequence rather than as a blanket rule.
What "finished" means for each cadence if a reply isn't the right signal, and which mechanism will handle it instead — a condition branch, a chained sequence, or an external flow.
Whether records that exit should be picked up by anything afterwards, or whether exiting is genuinely the end of the automated process.
Who is responsible for the remaining open activities on a record that exits, and whether they get cancelled, completed or left alone.
Whether the client needs any record of why a record exited, given the toggle itself doesn't produce a rich audit trail.
How exit interacts with records connected to more than one sequence — exiting one does not touch the others.

Gotchas

The toggle can only be changed while the sequence is inactive or in revision. Consultants who plan to "just turn it on later" find themselves versioning a live sequence in front of users.
Records already connected under a version without exit criteria stay on that version's behaviour. Turning it on is not retrospective for in-flight records.
Without email engagement configured, the toggle can be set and will do nothing. There's no warning tying the two together.
Reply detection is exactly that — a reply. An out-of-office auto-reply is a reply, and it will exit the record. Expect a handful of these and don't let the client discover it during a board demo.
Exiting disconnects the record; it does not tidy up. Open activities the sequence already created remain on the record for someone to deal with.
On a record running several sequences, exit criteria only end the one whose criteria fired. The seller still sees the others in Up next and often assumes the exit didn't work.

Consultant notes

Set this before activation as part of the build standard, not as a later tweak. It saves an awkward versioning exercise on a live cadence.
When the client asks for richer exit criteria, offer the condition-step-plus-chained-sequence pattern in the same breath as saying no. It keeps the conversation constructive.
Warn about out-of-office replies explicitly in UAT notes. It's a small thing that reads as a defect if it turns up unannounced.
Agree the housekeeping rule for leftover activities and, if it matters to reporting, automate it — otherwise the client's open-task counts will drift upwards for years.
Check email engagement is live and healthy in every environment before promising reply-based exit. It's the dependency that quietly doesn't come across in a deployment.

Worth another look if Microsoft extends exit criteria beyond email reply, or adds per-record exit control.