Published · 3 min read
Scheduled agent runs now finish and report what they found
AdvisorPPC scheduled agent runs now have a 40 minute time limit, a short wrap-up step and a larger default report size, and a second run is skipped while one is still running.
By the AdvisorPPC Team · Reviewed by Claude
If you schedule an AdvisorPPC agent to review your ad accounts, a run now has a time limit, and it wraps up with a partial result if it reaches it. This update shipped on November 7, 2026. Here is what changed and why.
What was going wrong?
Nothing put a ceiling on how long a scheduled run could take. Every step could wait and retry, so one slow answer could hold up the whole run. The longest case we found was a scheduled run that took 7 hours and 35 minutes. Some runs also stopped mid-report, because the space left for the final write-up was too small.
- Longest scheduled run found before
- 7 h 35 min
- Time limit per run now
- 40 min
- plus up to 5 min to write up
- Room for the final report
- 4x
- larger than before
What happens now?
- Every run has a 40 minute limit. When it is reached, the agent stops looking at new data and gets one short turn, up to 5 minutes, to write down what it found.
- A run that hits the limit still lands as ready, marked as partial, with its findings. It no longer shows up as failed hours later.
- The default size limit for the final report is four times larger than before (16,384 tokens instead of 4,096). Earlier, a long account review could be cut off before its report was complete.
- If a scheduled time comes round while the previous run is still working, the new one is skipped and recorded as "skipped: previous run still working", instead of two runs colliding.
- Small account changes no longer stop a run. Reordering your accounts, or adding or removing an account the run does not use, keeps the run going. A real change to the accounts the run uses still marks it out of date, so it never works on the wrong account.
Does the agent change anything by itself?
No. This update is about runs finishing, not about what they are allowed to do. A scheduled run reads your connected accounts and prepares findings and proposals. Changes still wait for your approval, as described in approve every change and approval gates for AI PPC agents. Account data is read through the platforms' official interfaces, such as the Google Ads API.
What should I expect on the agent card?
The agent strip in chat now says when the last run happened in plain words, for example "Last run 3 days ago", "No run yet" or "Queued", instead of a status code. A partial run reads like any other finished run: open it and you see what the agent covered before the time limit.
What is a good first schedule?
A weekly review works for most small accounts. Pair it with a short routine of your own, such as the Monday Google Ads pass, and use the agent's report to decide what to look at first. If you connect from Claude or ChatGPT, the setup pages are AdvisorPPC in the Claude connector directory and how to add AdvisorPPC in ChatGPT.
Figures come from the AdvisorPPC run settings shipped on November 7, 2026: a 40 minute run limit, a 5 minute wrap-up, and a final report cap raised from 4,096 to 16,384 tokens. The 7 hour 35 minute run is the longest scheduled run recorded before the change. Your runs may be shorter.