My agent job failed (Error in the task feed)
A job you ran shows Error in the task feed instead of finishing. This page explains what that status means, the common causes, and how to get the work done.
Symptom
Section titled “Symptom”In the Task feed (open it from the header, then View all to see the full list), the job card shows the status Error on the right side. The job’s message turns red and reads “Could not complete” followed by what the agent was doing, for example:
- “Could not complete qualifying” for an account or contact qualification run.
- “Could not complete research” for an account or contact research run.
- “Could not complete search” for a contact finder or email and phone search.
- “Could not complete domain finding” for a domain finder run.
- “Could not complete account plan”, “Could not complete digital twin”, or “Could not complete” a play generation run.
This is different from a job that ran to the end. A finished job shows Finished; a job you stopped yourself shows Stopped; and records the agent intentionally passed over show Skipped (see “How Error differs from Skipped and Stopped” below).
If the job touched several accounts or contacts, some records may have completed while others errored. The progress bar on the card shows Completed, In progress, and In queue at a glance. For the full breakdown, open See report, which lists In queue, In progress, Completed, Skipped, and With errors counts, so an Error result does not always mean nothing got done.
Why this happens
Section titled “Why this happens”A job ends in Error when the agent could not finish the work for the records in the run. The most common causes are:
- Missing input data. The agent needs something it does not have. A research or qualification run on an account with no website cannot gather the evidence it needs; an email or phone search needs a known company and person to look up. Records that lack the required input are the usual reason a run cannot complete.
- An upstream data source did not respond. Research, contact finding, and email and phone enrichment depend on outside data providers. If a provider is unavailable or returns nothing, the affected records error rather than finish.
- An invalid agent setup. If the run relied on an ecosystem whose structure has a logic problem, the agent cannot decide how to proceed. In that case the card shows an information tip next to the status that reads: “Your ecosystem structure has some logic issues. Contact your customer success manager to resolve the issue.”
Open See report from the card’s menu to see which step failed and how many records were affected, so you know whether the cause was missing data on a few records or something broader.
How to fix it
Section titled “How to fix it”- Open the Task feed from the header and find the errored job.
- Open the card’s menu and choose See report to see the per-step breakdown. Note the With errors count and which step it sits under.
- If a tip about ecosystem structure appears next to the status, that setup needs to be corrected before the run can succeed. Contact your customer success manager. The other fixes below will not clear this case on their own.
- If the cause is missing input, fix the underlying record first. Add the company website or domain to the affected accounts, or fill in the missing detail the agent needs, then save. See Account detail page and Contact detail page for where to edit a record, or Run a domain finder when accounts are missing a domain.
- Once the input is in place, run the agent again on the affected records. Follow the same steps you used the first time, for example Run a qualification, Run research, Run the contact finder, or Find an email or phone number. The new run starts a fresh job that appears at the top of the task feed.
- If the records had valid input and the run still errors, an outside data source may have been temporarily unavailable. Wait a short while and run it again. If it keeps failing, ask your customer success manager to have the run’s errored records retried.
You do not pay full price twice for a record that already succeeded. A new run only works the records you point it at, so rerun just the ones that errored where you can.
How Error differs from Skipped and Stopped
Section titled “How Error differs from Skipped and Stopped”These three statuses look similar on the card but mean different things:
- Error means the agent tried and could not complete the work for the affected records. This is the status this page is about, and it is the one you act on.
- Skipped means the agent intentionally passed over a record because it did not need processing, not because anything went wrong. No action is needed.
- Stopped means someone cancelled the job before it finished, using the Stop action on the card. A stopped job did not fail; it was halted on purpose. You can start it again whenever you want.