Replies answer the actual message

Reply drafting had three separate, unrelated problems that could leave a reply feeling disconnected from what a prospect actually wrote, or leave a contact stuck for no good reason at all. Reply drafts now receive the exact inbound thread they are actually answering, a temporary research failure retries without charging you anything at all, and a malformed AI response retries immediately instead of stalling the whole campaign for a full day.

What changed

When a prospect takes the time to reply to your outreach, the AI needs the actual content of that inbound message to write back a relevant response. In some cases, the reply draft was not being given the exact thread it was responding to, which meant the AI could end up writing a reply that did not clearly connect to what the prospect had actually said, sometimes answering a slightly different question than the one that was actually asked.

Separately, research sometimes fails for temporary, transient reasons, like a site being briefly unreachable rather than genuinely lacking information the AI could use. Previously, this kind of temporary failure could still result in a charge, even though no usable research came out of it, which meant you could end up paying for an attempt that produced nothing you could actually use in a draft.

Third, when the AI's response came back malformed, meaning it did not parse into something the system could use, the retry did not always happen right away. Instead, a campaign could end up stalled for a full day before trying again, even though a malformed response is usually resolved by simply trying once more within a minute or two, not waiting until the next scheduled cycle.

Each of these three problems compounds the others in practice. A reply that misses the actual thread context is a quality problem. A research charge for a failed attempt is a fairness problem. A day-long stall over a malformed response is a speed problem. Together, they meant a genuinely engaged prospect, the kind who replies quickly and expects a quick response back, could end up waiting on a campaign that had quietly stalled behind the scenes for reasons that had nothing to do with the quality of the conversation itself.

Reply drafts now receive the exact inbound thread they are responding to, so the AI is writing with the actual prospect message in front of it, not a partial or disconnected version that could lead it to answer the wrong thing. Temporary research failures now retry without a charge, so you are not paying for an attempt that produced nothing usable. And malformed AI responses now trigger an immediate retry instead of waiting up to a day, so a hiccup in one response does not stall a campaign that should otherwise be running normally.

These three fixes were verified together end to end, following an actual reply through thread handling, research, and drafting rather than checking each piece in isolation. That matters because a reply pipeline is only as good as its weakest link, and confirming the full path rather than just the individual fixes was the only way to be confident a prospect's reply now reliably produces a fast, relevant response.

How to use it

If you noticed reply drafts in the past that felt off topic or disconnected from what a prospect actually wrote, that should no longer happen, since the reply now has the correct thread to work from every time.

If you saw campaigns that seemed to pause research unexpectedly, or contacts that appeared stuck for a day without explanation, those delays should be gone. A temporary research hiccup or a malformed response should now resolve itself quickly rather than leaving a contact stranded in a queue overnight.

If you review billing for research charges, you can now expect that a charge only appears when research actually produced something usable. A charge without a corresponding usable result on a contact you are reviewing is worth flagging, since that should no longer happen under normal operation. Keep an eye on that pattern over the next few billing cycles as a simple, ongoing sanity check.

Why it matters

A reply that does not clearly respond to what a prospect actually said undermines the entire point of replying at all: showing that you read and understood their message. Getting charged for research that failed for a temporary reason is simply unfair, and it erodes trust in the billing itself. And a day-long stall over something as fixable as a malformed response wastes time in a channel where responsiveness matters, since a slow reply can cost you the momentum of a prospect who was actually engaged. Fixing all three keeps reply handling working the way it should: fast, connected to the actual conversation, and fair on cost. A prospect who takes the time to reply is telling you they are paying attention, and the reply pipeline should treat that moment with the same speed and accuracy they are showing you.

Reply quality also compounds across a longer relationship with a prospect. A first reply that clearly understood what they wrote sets the tone for every message after it, while a reply that felt slightly off can make a prospect more guarded even if the rest of the conversation goes smoothly. Getting the first reply right is worth the specific fixes described here, and it is the kind of detail a prospect notices even when they could not say exactly why a conversation felt easy or felt effortful.

These fixes apply automatically to every reply your campaigns handle from now on. There is nothing to configure and nothing to re-check on your side beyond watching whether replies now feel more connected and campaigns keep moving instead of quietly stalling behind the scenes, day after day, without you needing to intervene or dig through a queue to find out why a contact quietly stopped moving through the sequence for reasons that turned out to be entirely fixable.

Availability

These fixes apply automatically to every reply your campaigns handle now, across every workspace. There's no setting to flip and no configuration required on your side.