OpenAI and Google Gemini offer a deferred mode: the requests are queued and answered within hours (usually much less) at about half the price. ChronoScan calls it AI Batches. It is the right choice for large volumes that do not need an immediate answer: a night hot folder, an archive migration, a weekly reprocess.
Enabling deferred execution
In the Process and Tasks settings of the job, the row of each AI configuration has an Enable deferred execution option. When a configuration allows deferred mode and the option is not set, ChronoScan asks before processing whether to use it; answering No stops the question for that job (delete the job parameter to see it again).
What happens to the batch
The documents are processed as usual up to the AI stage; the requests of the whole batch are sent as one deferred job to the provider.
The batch is marked as waiting for AI. In Enterprise its status is "deferred AI" in the dashboard and the server polls the provider every few minutes.
When the answers arrive ChronoScan completes the documents (mapping, assistants, scripts) and the batch continues with validation and export.
The AI Batches window lists the deferred jobs with their provider, batch, number of documents and state. Deferred requests use the same prompts and mapping as the synchronous ones, so a job can switch between modes without changes.
The AI cost estimation compares the synchronous and the deferred cost of a batch before processing it.
