FlowsBook a callSign inCreate free accountSign up
AI employees and automation

Reading the run history of an AI employee: what each run tells you

Every time an AI employee works it leaves a run: passed, failed, waiting or stopped, with what each block did. How to read the history and spot a problem before it costs you.

Part of the guide AI employees

By the xMap team · · 5 min read

An AI employee's run history is its work diary: one line for every time it worked, saying whether the run passed, failed, is waiting or was stopped, what started it, what it found and what it sent. Read it from the top for health (a row of passed runs, then the first failure), then open single runs for detail: which block did what, how many items moved between blocks, and whether data was live or sample. Most problems show in the history days before anyone notices missing emails.

This is for anyone who has deployed an AI employee in xMap Flows and wants to know it is doing its job: owners, team leads and the person who built it. It explains what a run is, the four outcomes, what to look for in one run, a made-up example from Singapore, and the habits that keep a team of employees healthy.

What a run is

A run is one go through the flow, from its trigger to its last block. A daily employee makes one run a day; one that starts when a new row lands in a sheet makes a run per new row; a chat page makes one per message. Each run is kept with what started it:

  • On its own: the schedule, a webhook call, new rows, an app event or a change it was watching for.
  • By hand: Run once from the editor, which is a real run that sends.
  • A test: Test run, which fetches data but sends nothing. Tests are marked, so a history full of tests is not a working employee.
  • A retry: a failed run tried again, linked to the one it retried.

The four outcomes

OutcomeWhat it meansWhat to do
PassedEvery block that needed to run did its jobNothing, unless it found nothing for days
FailedA block could not finish and the run stopped thereOpen it, read the reason, use Explain and fix or retry
WaitingPaused on purpose at a Wait or an Approval, or carrying on a long run in partsAnswer the approval, or let it carry on
StoppedSomeone pressed Stop, an approval was declined, or a branch ended on purposeUsually nothing: it is not a failure

How to read each run's status.

The employee's card on the Flows page shows its recent runs as dots, failures in red, so a pattern stands out without opening anything.

Reading one run

Open a run and its first line sums it up in words, for example how many it found, how many were new, how many scored as strong fits and whether it emailed. A run with results shows them first; the blocks are under How it ran. A failed run shows the blocks first, open, so you see where it stopped.

  1. Find the block that failed, if any. Its reason is written in words: a missing setting, a service that did not answer, an account that needs signing in again.
  2. Click a block to see its data. Every passed block shows the fields it passed on and its first items, so you can see where a list of 50 places became 3.
  3. Check live or sample. Each data step says whether it used live data. A step marked sample means a service was not set up or not available, and the results are examples.
  4. Look for blocks marked Not needed. In a run started by one trigger, the flow's other triggers do nothing. That is normal.

A made-up example from Singapore

This story is invented. A cleaning company in Singapore runs an employee that emails new offices every morning. For three weeks its history shows a passed run each weekday with eight to fifteen new offices. Then four mornings in a row it passes with zero.

Opening the latest run, the team clicks through the blocks: the data block found 60 places as usual, xMap LLM scored them, and a Filter kept only scores above 80. Someone had edited the instructions to xMap LLM the week before, and the new wording made it stricter, so no place reached 80. The version history shows the change. They restore the old instructions, press Run once, and the next run brings eleven offices again. Nothing had failed; the history simply showed the moment its output changed.

Habits that keep a team healthy

  • Look at the history weekly, not only when an email is missing. The weekly report card of each employee sums up runs, failures and what was found.
  • Use Explain and fix on a failed run. xMap AI reads the error and the flow, says what went wrong and offers a change you can undo.
  • Retry from the failed block. A retry reuses what earlier blocks already fetched and sent, so it neither pays again nor sends twice.
  • Keep failure emails on. You get one when an employee starts failing and one when it works again, not one per run.

Doing it with an AI employee

A run history is only as useful as the employee is easy to read. Give each block a clear name, keep a Filter or Sort and limit where you can see what it cut, and end with Email so every passed run leaves a message you can match to its run. A daily employee built as Schedule, Nearby places, Only new, xMap LLM, Filter and Email leaves a history where each step's count tells the story. To check the whole team at once, Team findings in the template A weekly brief from your AI employees sums up what every employee found this week against last; New places every morning, only new is a simple employee to learn on.

Questions

How long is run history kept?

Runs are kept for a year. After 90 days the detailed data between blocks is slimmed, but the outcome and summary stay.

Why does a run say passed when nothing was sent?

A run passes when every block did its job. If nothing new was found, there was nothing to send. Check the counts block by block if that is unusual.

What is the difference between a test run and Run once?

A test run fetches data and shows what would be sent, but sends nothing. Run once is a real run: it sends and its results are remembered.

Should I worry about a stopped run?

Usually not. Stopped means someone pressed Stop, an approval was declined or the flow ended a branch on purpose.

Templates that do this

IndustriesB2B sales

Read next

More on ai employees and automation · Every article · Templates

Put it to work

An AI employee in xMap Flows does this on live data, on its own timetable. Start from a template or describe it in plain words.