The problem this solves
Running campaigns generate numbers, and numbers are not diagnosis. A dashboard shows step two's reply rate sagging; it does not say whether the copy is weak, the segment is wrong, the send timing is off, or the list quality degrades past a certain company size. So underperforming campaigns run on untouched - the numbers get glanced at, the conclusion 'could be better' gets reached, and nothing changes because nobody has translated the metrics into a specific, fixable fault.
The honest autopsy is genuinely analytical work: comparing performance across steps and segments, distinguishing a deliverability problem from a copy problem from a list problem, and locating where in the funnel the loss concentrates. It falls between roles - the SDR sees replies, the manager sees totals, nobody owns the cross-section - and it competes with next week's launch for attention, a competition maintenance always loses.
Diagnosis without repair is a report; the loop closes only when the weakest touch gets rewritten and the revision actually enters the tool. That last step is where good intentions go to die, because 'we should rewrite step two' has no owner the moment the meeting ends. A mission that carries the fix all the way into Lemlist is what turns the observation into an outcome.
How the mission runs
- Pull the campaign stats from Lemlist. The agent reads your running Lemlist campaigns step by step: sends, opens, replies, and bounces at each touch, broken down by segment where the campaign structure allows. The raw cross-section that a manual review never quite assembles is the starting material here.
- Locate where the funnel bleeds. The numbers are compared across steps and segments to find where the loss concentrates: a step whose reply rate collapses relative to its opens, a segment that underperforms every step, a bounce pattern pointing at list quality. The diagnosis names a place before it names a cause.
- Distinguish the failure type. Weak opens with healthy delivery point at subject lines; healthy opens with silent replies point at the body or the ask; bounces concentrated in one segment point at the list. Web Research adds context on the underperforming segment where the pattern suggests a positioning mismatch rather than a copy fault.
- Rewrite the weakest touch. The agent rewrites the touch the diagnosis indicts, changing what the evidence says is broken and preserving what works. The rewrite comes with its reasoning attached - what the old version's numbers implied, and what the new version changes in response.
- Stage the revision back into Lemlist. The rewritten touch is staged into the campaign in Lemlist for your review rather than hot-swapped mid-flight. You see the diagnosis, the before and after, and the reasoning in the thread, and you approve the change before any running sequence starts sending it.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A written autopsy in the thread - where the campaign bleeds, by step and segment, with the evidence - plus the weakest touch rewritten with its reasoning, staged into Lemlist behind your approval. The loop from 'the numbers look soft' to 'the fix is live' closes in one mission, and running it on a cadence turns campaign maintenance from a guilt into a routine.
Make it yours
- Run the autopsy as a standing biweekly review, so every active campaign gets an evidence-based tune-up before small decays compound into a dead sequence.
- Ask for two rewrite candidates with different angles and split the campaign's remaining sends between them, turning the repair itself into an in-flight test.
- Point it at a finished campaign before cloning it for a new market, so the sequel starts from the diagnosed lessons rather than the raw template.
Frequently asked questions
Does the agent change my live campaign directly?
The revision is staged for your approval before it replaces anything in a running sequence. In-flight campaigns are commitments with an audience attached, so edits pass through the same review gate as any other consequential write. You approve, then the fix goes live.
What can the diagnosis actually distinguish?
The step-by-segment cross-section separates the classic failure types: delivery problems, attention problems at the subject line, persuasion problems in the body, and list-quality problems by segment. Each conclusion cites the numbers behind it, so you can challenge the diagnosis rather than take it on faith.
When is a campaign worth an autopsy versus a restart?
If the bleed concentrates in one step or segment, repair usually wins - the structure is sound and the fault is local. If every step underperforms uniformly, the diagnosis will say so, and that pattern typically indicts the list or the offer, where a rewrite of one touch cannot save it.