Automation agent
Escalation / Exception Handler
The circuit-breaker for autonomous playbooks - catches a stalled sequence instead of letting it loop.
What does Escalation / Exception Handler do?
Escalation Handler is the agent that exists so a fully autonomous playbook can't quietly get stuck. It reviews execution history across every running workflow, and when a sequence has stalled - waiting on a condition that will never resolve, or hitting a step with no clear next action - it raises an internal task instead of looping or failing silently. It's never customer-facing: nothing it does reaches an account contact directly.
What signals does it watch?
It watches the health of automation itself, not account risk or expansion signals - a distinct job from every other agent in the library.
- A playbook step that has been waiting past its expected duration with no state change
- A condition check that keeps returning the same unresolved result across multiple passes
- A workflow run that has no further defined step to advance to
- Repeated failures attempting the same action, which would otherwise retry indefinitely
What actions does it take?
Because it exists purely as a safety net, its own action set is deliberately narrow.
- Raises an internal task on the account and the Tasks dashboard, describing exactly where and why the sequence stalled
- Stops the stalled playbook from re-attempting the same failed step indefinitely
- Runs at the fully-autonomous tier - there's no customer-facing action here to gate behind approval
- Leaves a full record in the account's Agent Runs history, so the stall is diagnosable after the fact
How does it fit into your workflows?
Escalation Handler doesn't belong to one seeded workflow - it works across every workflow that runs at the fully-autonomous tier, which is precisely where a silent stall would otherwise go unnoticed the longest.
What does this look like in practice?
A longtail account is on a fully-autonomous Onboarding Stall workflow. The playbook is waiting on a condition - "has the account completed integration setup" - that never resolves, because the account stalled before that step entirely. Days pass with the sequence idling. Escalation Handler catches the stall, raises a task explaining exactly which condition never cleared and how long it's been waiting, and stops the workflow from continuing to poll. A CSM picks up the task, reaches out directly, and manually advances the account once the real blocker - a missing technical contact - is resolved.
Frequently asked questions about Escalation / Exception Handler
- Does Escalation Handler ever contact a customer?
- No - by design it's never customer-facing. Every action it takes is internal: a task for your team, never an email or message to an account contact.
- How long does a step have to stall before it gets flagged?
- That depends on the step's own expected duration, which is set when the workflow is built - Escalation Handler compares actual elapsed time against that expectation rather than using one fixed timeout for everything.
- What's the difference between this and a normal signal?
- A normal signal describes something happening on the account - risk, expansion, a stalled onboarding. Escalation Handler's task describes something wrong with the automation itself, so your team knows to look at the workflow, not just the account.
- Can I turn it off?
- You could, but it's the one agent we'd actively discourage disabling - without it, a stalled fully-autonomous sequence has no built-in way to surface itself, and would simply sit idle until someone happened to notice.
See Escalation / Exception Handler on your own accounts.
Book a 30-minute demo, or connect your stack and try it free for a week.