Portfolio-Wide Report Batch Generation Automation: One Call for the Whole Client Roster

Generating a single client's report is a small task; generating the same report for every active client in a real agency roster, one modal at a time, every single month, is not — it is the same small task multiplied by however many clients are on the books, with no way to tell, partway through, which ones are already done and which ones still need attention.

Naively looping "generate one report" across an entire client list either runs everything sequentially — which is safe but slow enough to time out on a large roster — or fires everything at once, which risks overwhelming a database and an AI provider that has its own rate limits, causing failures that have nothing to do with any individual report being wrong. And once a batch can partially fail, silence is not an acceptable outcome: a team running reports for thirty clients needs to know, specifically, which ones succeeded, which already existed, and which genuinely failed — not just whether the request as a whole returned an error.

One call accepts either a specific list of client IDs or, when none is given, every active client in the organization, along with the month and year to generate for and whether to also send once generated. The batch is capped at 40 clients per call — an explicit, documented ceiling chosen so a batch always finishes within the request's time budget rather than being cut off partway through — and is additionally rate-limited to 3 batch runs per rolling 5-minute window per organization, so a batch cannot be re-triggered by accident moments after the last one started.

Clients are processed in groups of three at a time, not all at once and not strictly one by one — a deliberate middle ground between finishing quickly and not overloading the database or the AI narrative provider's own rate limit. Each client's result comes back individually: a successful generation, an "already exists" outcome when that period was already generated (worded as informational, not as an error, since a batch of thirty should not look thirty times worse than it is because half the reports were already done), a specific failure message, and — when sending was requested — whether the send itself succeeded, including the number of recipients it actually reached.

An operations lead triggers a batch for the organization's 30 active clients at month-end, with sending turned on. The batch processes them three at a time: 26 generate and send successfully, 3 already had a report for that period from an earlier manual run (reported as "already existed," not as failures), and 1 fails because that client has no report recipients registered yet. The response lists all 30 outcomes individually, so the operations lead fixes the one real gap — adding a recipient for that client — without needing to re-check the other 29.

Month-end reporting for a whole client roster stops being a linear sequence of identical manual steps and becomes a single triggered run with a clear, itemized outcome. A partial failure in a batch of thirty is visible as exactly what it is — one specific client with one specific reason — instead of an ambiguous all-or-nothing result that leaves a team re-checking everyone to find the one that actually needs attention.

A roster larger than 40 active clients cannot be run in a single batch call — the cap is a hard ceiling, and a larger portfolio needs to be split across more than one call (or a subset selected by client ID) rather than run in one shot. A team wanting the batch itself to run automatically on a schedule will not find that here — it is triggered by a call, the same way single-report generation is, not by a cron dispatching it unattended.

Generating reports for a whole client roster one modal at a time multiplies a small task by the size of the roster, with no visibility into progress until every single one has been clicked through. Centriu Atlas instead processes the whole target list — up to 40 clients — in one throttled call and reports each client's specific outcome individually, so a team spends its attention on the one report that genuinely needs it instead of re-confirming the twenty-nine that were already fine.