The problem this solves
Every outreach list arrives dirty. Whatever its source - an export, a purchase, a scrape, an old campaign - some fraction of the addresses are dead, some are catch-alls, and a few are spam traps. Sending to that list as-is spends your domain reputation to find out which is which, and reputation is the one asset a cold-email program cannot quickly rebuild. Bounces compound: each one makes every future send a little less likely to reach an inbox.
Verification alone gets you to safe; it does not get you to good. A clean list with a generic opener still performs like a generic campaign. The last mile - a first line per contact that reflects their actual company - is the work teams skip under deadline, because writing a hundred context-aware openers is a day of research nobody budgeted. So launches go out clean but flat, and the list's potential is spent on a template.
The result is a false choice between speed and quality: launch now with generic copy, or launch next week with personalization. The list-prep step is squarely mechanical-plus-research work, which is exactly the shape of work worth handing to an agent. The agent runs the verification and the research in one pass, and the launch happens today with both boxes checked.
How the mission runs
- Ingest the list from the thread. You paste the list directly into the mission thread in whatever shape it exists - CSV, spreadsheet rows, even a rough copy-paste. The agent normalizes names, companies, and addresses into consistent fields before any verification begins, so the output is cleaner than the input in structure as well as content.
- Verify every address. Email Verification runs on each contact and classifies the result: deliverable, risky, or undeliverable. Catch-all domains and unknowns are treated as risky rather than optimistically included, because the entire point of the preflight is that your send list contains only addresses that will land.
- Split the list honestly. The output separates into a deliverable set and a quarantined set with the verification result recorded per row. You see exactly what fraction of the original list was healthy, which is also a quality report on wherever the list came from and worth knowing before you buy or export from that source again.
- Rewrite the opener per deliverable contact. For each deliverable contact, Web Research looks at the company - what it does, its positioning, anything current - and the agent rewrites the first line to reflect that context. Every send-ready row gets an opener that reads as written for the recipient, because it was.
- Hand back the send-ready CSV. The finished CSV lands in the thread: verified contacts, their rewritten openers, and clean columns ready to load into whatever sends your mail. The quarantined rows travel alongside in their own file, so nothing from the original list is silently lost.
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
Two files back in the thread: a send-ready CSV of verified-deliverable contacts, each with a company-aware opening line written for them, and a quarantine file of risky and undeliverable rows with the verification verdict on each. Your launch list is safe for your domain and sharper than the template it replaces, and you know precisely what the cleaning cost you in list size.
Make it yours
- Run verification-only when you just need the safety pass fast, and add the opener rewrite on the second pass before launch.
- Give the agent your value proposition and have the openers bridge from company context to your pitch in the same line.
- Make it a standing gate: every list from any source passes through this mission before it is allowed near a sending tool.
Frequently asked questions
What happens to catch-all and unknown results?
They are quarantined with their verdict rather than mixed into the send set. Catch-alls accept everything at the server level, so deliverability cannot be confirmed; treating them as risky keeps your bounce rate honest. You can always choose to work quarantined rows through other channels.
How much of a typical list survives cleaning?
It varies entirely with the list's source and age, which is why the mission reports the split explicitly instead of quoting an expected rate. Older and purchased lists lose more; recent first-party exports lose less. The report tells you the truth about your specific list.
Are the rewritten openers based on real information?
Each opener comes from Web Research on that contact's company at run time - their actual positioning and current public footprint. Nothing is invented to fill space, and where research comes back thin, the agent writes a plainer line instead of fabricating specificity.