The prospecting allowance is spent
Two separate limits govern prospecting in a customer workspace: an AI budget covering research, and a daily send allowance covering outbound messages. They run out independently and behave differently — research is refused, sends are held. Nothing that was already done is undone, and no approval needs giving twice.
Which limit you have hit#
| Limit | Stops | Behaviour | Clears when |
|---|---|---|---|
| AI budget | Paid model work: research, deeper passes, drafting | Refused, with the refusal recorded | The budget is raised, or the period resets |
| Daily send allowance | Outbound messages | Held; approval is recorded and the send waits | The allowance resets, or the plan changes |
A discovery run that stops partway is usually the first. Outreach that sits approved and unsent is almost always the second. The two are easy to tell apart from the symptom, which is fortunate, because the remedies are different.
What Connect completed, and what carries on#
- Everything already researched keeps its claims, its evidence and its score. An allowance is a limit on new work, not a retraction of old work.
- Reading, filtering, sorting and exporting the list. None of that costs model work.
- Discovery when only the send allowance is exhausted — researching today what you will write tomorrow is ordinary use, and the two limits are genuinely independent.
- Approvals. Approving a held message is recorded immediately; the send is what waits.
- Inbound. Replies still arrive, are still classified, and a stop still stops the sequence. Nothing about a spent allowance makes Connect deaf.
What Connect did not complete, and what waits#
- Held sends
- They go when the allowance resets, in the order they were approved. They are not deleted and do not need re-approving.
- Refused research
- It does not queue itself. A run that was refused resumes when you run it again, and candidates already researched are not paid for twice.
- Sequence steps
- A step that falls due while sending is exhausted is held rather than skipped, so a prospect does not silently miss the message that was planned for them.
- Silent degradation
- Does not happen. Research is refused rather than run at lower quality, and a message is held rather than sent to fewer recipients.
- Address invention
- Still does not happen. A spent allowance changes nothing about what may be written to the record.
What to do#
Open Plan & Usage and read which of the two is exhausted.
Result Both are shown against the plan's ceiling, so this settles the diagnosis before any settings are touched.
If it is sending, check Needs You for what is waiting.
Result The queue shows the held work; nothing there requires action beyond deciding whether any of it should be cancelled instead.
If it is research, narrow the next run before repeating it.
Result A tighter brief and a real location scope reduce spend before a model is called, which is more effective than raising the ceiling.
Where the work genuinely needs more headroom, an administrator changes the plan or the budget.
Result The change applies to the workspace, and held work drains without further approval.
Escalate when Plan & Usage shows headroom on both limits and prospecting is still refusing or holding. That combination is not an allowance problem and belongs with the outreach-blocked diagnosis instead.
Questions#
Did I lose the messages that were held?
No. A held message keeps its text and its approval and goes out unchanged when the allowance clears. What it does not do is arrive on time, which is the cost worth weighing.
Does discovery stop when sending stops?
No — the two allowances are independent. You can research a list while unable to write to any of it, and for a workspace planning a week's outreach that is a reasonable way to work.
Why was research refused instead of just running more slowly?
Because a budget that quietly degrades work is worse than one that refuses it: you would be paying for research whose depth you could not see. The refusal is recorded so the gap in the list has an explanation.