Learn
Can an AI agent run a process that takes weeks?
What breaks when a process waits on a person for days, and what a platform needs to survive it.
The short answer
Yes, but only if the platform treats waiting as a normal state rather than an error. Most automation assumes a process starts and finishes in one session. Real enterprise work does not.
Why enterprise work waits
A process waits on a vendor who has not returned a quote, on an approver who is on leave, on a quarter close, on a legal review that has to queue behind other reviews. The waiting is not an exception to the process; it is most of the elapsed time in it.
The failure modes
- Sessions time out while the process is legitimately idle.
- Context is lost, so whoever picks the work up has to reconstruct what happened.
- State is rebuilt by hand, which introduces differences between the run and the record.
- The run restarts from the beginning, repeating work that already completed.
Work that happens while nobody is watching
Waiting is not the same as doing nothing. A long-running WorkStream can keep working through the gap, which is where most of the time savings come from.
- On a schedule. A Flow runs at a set interval and puts an agent to work. It can check commodity prices on key inventory items and signal a buy recommendation to the procurement manager, or place the order itself when the Playbook has pre-authorized that action within its limits.
- On an inbound event. A WorkStream waiting on RFQ responses listens for the replies. When a quote arrives, the agent reads it, normalizes the line items, and compares it against the quotes already received.
- Ready when the person returns. The buyer opens the WorkStream and finds the comparison prepared and the exceptions flagged, instead of opening an inbox and rebuilding a spreadsheet.
Every one of those background actions runs under the same governance as a step a person triggers, and lands in the Execution Ledger the same way.
Where VeroTX sits
WorkStreams hold state across days or weeks and resume with full context, so a step that waits on a person picks up exactly where it left off. Failed runs resume from the point of failure rather than from the start. Every run records the Playbook version it started on.
Where to next
Read deeper
Product and platform pages on the same topic, drawn from the VeroTX knowledge base.
Search everything in one place in the Research Center, or ask Veroli, which answers from the same knowledge base.