Automations that don’t silently die.
Everyone demos a workflow canvas. The canvas is the easy part. What decides whether you trust automation with revenue is what happens on day 40 — when a run is mid-wait and you edit the workflow, when a provider times out, when someone asks why a customer got two emails.
- Runs survive restarts
- Yes
- In-flight runs on edit
- Finish on original version
- Per-run execution trace
- Every step
- Failed step handling
- Retry with backoff, then surface
Durable: a run in progress is a stored fact
Every step a run completes is persisted before the next one starts. A deploy, a restart, or a crashed worker doesn’t lose a run — it picks up from the last completed step. A workflow that waits three days for a reply is genuinely waiting three days, not holding something fragile in memory and hoping nothing changes.
Versioned: editing a workflow doesn’t rewrite history
When you publish a change, runs already in flight continue on the version they started on, and new runs use the new version. This is the failure mode that makes teams stop trusting automation elsewhere: you fix a typo in step two, and a thousand people sitting in a step-six wait are silently moved onto a path that was never designed for them. Each run records which version it executed.
- In-flight runs complete on their original version
- Publishing is explicit — drafts don’t affect live runs
- Every run keeps a step-by-step trace with inputs, outputs, and timings
- Failed steps retry with backoff, then surface instead of vanishing
Debuggable: “why did this happen?” has an answer
Open any record’s timeline and see which automations touched it, when, and what each step decided. Open any run and see the branch it took and the condition value that sent it there. When a customer gets a message they shouldn’t have, you can find the cause in a minute instead of reasoning backwards from an outcome.
Workflow automation, in detail.
How is this different from other visual builders?
The canvas looks similar. The engine underneath is the difference: durable step persistence, explicit versioning for in-flight runs, and a per-run trace. Those three are what determine whether automation still works after six months of edits.
Can automations wait for days or weeks?
Yes. Long waits are a normal step type — a run can wait for a date, a delay, a reply, or a field to change, and it holds that state durably rather than depending on a process staying alive.
What stops an automation loop from messaging someone repeatedly?
Re-entry rules per workflow, plus consent and quiet-hour enforcement at the send layer. Even a misconfigured automation can’t send outside a permitted window or to someone who opted out.
Can I test before going live?
Yes. Run a workflow against a specific record and inspect the full trace without sending anything externally, then publish when the branches do what you expect.
Try it on your own data.
Free workspace, an industry pack installed at setup, and every price published. No mandatory call.
Free trial · no card required · every price published