Write personalised outbound copy at scale
Generate first lines, pain points, angles and subject lines across a list, with a QC pass that drops generic output instead of shipping filler.
3 steps not yet verified live
Tool names and credit costs in this recipe are validated. Some steps work around a platform gap that has not been decisively probed, so treat those as a workaround rather than a documented capability. Check the result before you act on it.
Where the list comes from
A list you already have. This recipe adds copy fields, it does not source.
The steps
- 1
Orient
Confirm the inputs, compute the expected cost, and tell the user what the run will do before anything spends a credit.
- 2
Company context
Extract LinkedIn Company Details (2c) when a company LinkedIn URL exists, richest single source, gives description, size and specialities in one call. Otherwise Custom Research Tool (1c).
- 3
Person context
Full LinkedIn Profile Details (2c), headline, role, tenure, background. Current Employment (1c), confirms they are still there.
- 4
Pull recent signals
Cheapest useful signals for personalization, in order of value per credit: LinkedIn Post Fetcher (0.5c), what the person actually posted. The single best personalization source available, and one of the cheapest.
- 5
Summarize the most relevant signal
WorkaroundNo signal-ranking tool. Several signal columns exist but nothing picks the best one per row. Use AI - Multiple Inputs (0.5c, takes up to 5 inputs, exactly this job).
- 6
Generate the pain point
See the skill file for the full detail of this step.
- 7
First line, angle, subject lines
Cold Email Opener (1c), purpose-built and designed to take an upstream research summary as its sole source. Feed it Step 4's signal_summary, not the raw columns. Cold Email CTA (1c), aligned to the campaign goal.
- 8
QC: remove generic and low-confidence output
0 creditsThis step is the difference between personalization and mail-merge noise.
- 9
Export
WorkaroundApply the keep-filter: first line present, passed QC, confidence not low. Push to HubSpot, Salesforce, Zoho, Attio, Breakcold, Smartlead, or webhook. Instantly and Apollo have no native push; use CSV or webhook.
Where the run should stop
- The QC step removes generic and low confidence output rather than flagging it. That step is the reason the recipe exists.
- Signals are pulled and summarised before any copy is written, so the first line rests on evidence.
What you should end up with
Every retained row carries a pain point, a first line, an angle and subject lines, each traceable to the signal it came from.