Every prospect is researched before the first email is written

Research before the first line is written

Most cold email fails in its first sentence. It opens by describing the recipient's own company, which tells them nothing new and signals that a tool wrote it.

Smart research runs before the writing step so the draft has real material to work from: what the company has published recently, what the person's role covers, and what is publicly visible about both. If there is little worth saying, you learn that before you send, not after.

Research runs for every lead queued for outreach, whether it arrived from Lead mining, a list, a CSV import, or by hand.

How a research pass runs

The pass follows the same steps every time, and you can follow its progress on the lead record.

1. A lead is queued for outreach in a sequence.
2. A research query is built from four fields on the lead: name, title, company, and company domain.
3. A third-party research provider searches public web sources and returns a set of notes.
4. The notes are saved on the lead and shown in the app.
5. The writing step runs, using those notes together with the workspace writing controls as its inputs.

What ends up on the lead

Research notes are stored on the lead record and appear in the app alongside the contact and company fields. You can open any lead and read exactly what the draft was based on.

Thin results are shown too. When a company has almost no public footprint, the lead carries a short note rather than an invented one. That gives you a clear decision point: send a plainer email, add your own context through custom instructions, or drop the lead before it costs a send.

From research to personalized copy

The writer receives the notes, the lead's fields, and the workspace controls, and produces one draft per email step. Copy is generated by a third-party language-model provider.

The notes shape the opening and the relevance of the message, not its structure. Your tone, call to action, and subject style still decide how the email reads. Personalization depth decides how heavily a given touch leans on the research, from a single specific reference to a lightly personalized follow-up.

Every draft lands in the review queue by default. You read the note, read the draft built on it, and approve, edit, or discard. Approving a draft sends it.

Writing controls

Writing controls live on the workspace, so every sequence in it inherits the same voice.

• Tone: how formal or direct the copy should be.
• Call to action: what each email asks for, from a short reply to a meeting.
• Subject style: the pattern subject lines follow across the workspace.
• Personalization depth: how much of each email is built from the research notes.
• Custom instructions: free-text guidance the writer follows, such as what to lead with or what to avoid.

The outbound filter

Every product-written email passes through an outbound filter before it is queued. The filter removes the tells that mark a message as machine-written, including em dashes and stock openers, and strips tool-fingerprint headers, so the message arrives as ordinary mail from your mailbox.

The filter applies only to copy the product wrote. Replies you type by hand in the unified inbox are sent verbatim, with no rewriting and no filtering.

Scope

A few boundaries define what research can and cannot do.

• It reads the public web only. It does not access private systems, internal documents, email, or anything behind a login.
• A research note is what the web says about a person and a company, not verified fact. Review it before approving the email built on it.
• Quality tracks public footprint. A company with a blog and press coverage produces detailed notes. A small company with a one-page site produces very little.
• It is a one-time pass before contact, not account monitoring. See Sequences for how follow-ups and Reply detection work after the first send.
• Copy is generated by a third-party language-model provider on terms under which your data is not used to train their models. Human review is the default; nothing sends without approval until autopilot is switched on for a sequence.

Credits

Research is metered on the same credit allowance as every other action in the platform.