Alacrio can send email now. It reads your inbox, notices a thread that has gone quiet, writes the follow-up, and puts it in front of you. Once you approve it, the message goes to your client from your own account, threaded into the original conversation.
That is a large amount of trust to ask for. So here is the other half of the product: the things it will not do, and why each one is written down as a rule instead of left to judgment.
None of the five below are settings. There is no toggle, no plan that removes them, and no confidence score high enough to skip past them.
Every message is proposed, never dispatched. The system writes a draft and then waits. You approve it from the dashboard or straight from Slack, and until you do, nothing leaves your mailbox.
This one costs us something. A fully automatic version would demo better and would save you the ten seconds of reading. Ten seconds is also the cheapest insurance you will ever buy on a client relationship, so the trade is easy.
The moment your client answers, the follow-up sequence for that thread stops. The reply itself does the pausing, so it holds whether or not you ever open the dashboard or think about it again.
The failure we were guarding against is one everyone has been on the receiving end of. You answer someone, and their software chases you anyway three days later, because a human somewhere never got around to telling it you had replied.
When a follow-up gets no reply, the system learns nothing from it. That looks like wasted signal, and discarding it was deliberate.
If silence were treated as a bad outcome, the reasoning layer would gradually conclude that follow-ups do not work. It would raise its own thresholds, propose fewer of them, and eventually go quiet. A slow client would train the product into uselessness. So silence is recorded as a fact about the thread and never as a verdict on the writing. Only you can tell it a draft was wrong.
A follow-up loop closes fourteen days after it opens. A loop paused because your client replied still closes on day fourteen. There is no mechanism to extend one, and we removed a feature that would have added it.
A loop that never closes is a loop that eventually nags. We would rather the system forget something than quietly accumulate obligations you never agreed to.
Reply to the thread from Gmail and the system sees it, counts it as the follow-up it was about to suggest, and moves its own clock forward. You will not get an alert about a thread you already answered.
This sounds obvious. It is also the most irritating way this category of product fails, and getting it right meant treating your own sent mail as a real event rather than background noise.
Rules like these are easy to hold when the product is small and quiet. They get tested the first time someone asks for an exception, and the honest answer is that we would probably say yes to any one of them in isolation.
Writing them down first is how we avoid deciding them again under pressure, one customer request at a time. If you ever catch Alacrio breaking one of the five, that is a bug, and we would like to hear about it at hello@alacrio.com.