AI limits

Clear usage
no surprises

AI in HereCRM is spent under transparent rules: live chat and background automation have separate budgets, unfinished work is never dropped silently, and the system warns you before you reach the cap.

Three separate budgets

Automation never eats an employee's limit

The usual failure of AI plans: background processing burns the whole daily volume and a person is left without an answer. In HereCRM the budgets are separate, and one cannot take the other's room.

Live chat

Requests an employee makes in the interface: a question to the assistant, a conversation review, a draft reply to a client. This budget is spent by people only.

Background functions

What the system does on its own: ticket summaries, classification, resolution checks, contact profiles. It runs on its own volume.

Batch operations

Bulk work on request: outreach personalisation, processing large lists. Kept separate because it consumes noticeably more than the rest.

Usage rules

What counts
and what happens at the cap

Monthly volume and a daily guard

The plan sells a monthly volume: you spend it across the month as suits you. A separate daily cap protects the shared platform from a one-off spike and does not reduce the monthly volume: unfinished work is carried over and runs once the daily cap resets.

The whole request counts

Usage includes the request text, the model's answer, and the conversation context the model had to read to answer properly — the same way the model provider counts it.

Work is never cut off mid-way

When the daily volume runs out, background tasks move to the next day and run once the limit resets. Nothing is lost and nothing has to be restarted by hand.

A warning before, not after

A notification arrives at 80% and 100% of the daily volume — enough time to pause a bulk operation or discuss more capacity before employees notice.

Costs no limit

Part of the work is free

Search and matching

Semantic search across the base, similar-record hints, and duplicate checks run on vector models and are not part of the plan's AI token volume.

On-premise models

For dedicated workspaces some tasks run on a model on our own server. Those requests spend no plan volume and never leave the perimeter.

Need a different volume

Tell us
about your load

If you run more than a typical process, the AI volume is calculated for your case. We will show current usage across your scenarios and propose a suitable format.