Use AI for better cold calling

Mohamed ChahinSeptember 11, 20265 min readWatch tutorial

See a real Ultra-generated cold call opener and voicemail, the prospect research behind them, and how to add calling alongside LinkedIn and email in Twain.

A useful B2B cold calling script gives you a specific reason to call this person and a question they can answer. The research should explain the connection between their work and your offer, so you can get to that question without giving a product tour first.

We've added cold calling as a third channel in Twain. You can now combine LinkedIn, email, and calls in one campaign, or create a campaign just for calls. The cold call and voicemail steps are written scripts for the people making the calls.

In this short walkthrough, I add both steps to a campaign using my friend Mark Colgan's profile. Both scripts were generated by Twain's Ultra mode and are shown here unedited.

Add calling to a campaign, inspect the generated opener and voicemail, and edit their frameworks.

Add calls to a new or existing campaign

There are two places to add calling:

  • When creating a campaign, select Call in the channel options. Choose it on its own, with LinkedIn or email, or with both.
  • Inside an existing campaign, add a step and choose Cold call or Voicemail. You can add them to a sequence that already has messages in another channel.

The campaign in the video has a LinkedIn follow-up, a cold call opener, and a voicemail. That lets you review the written and spoken parts of the outreach together.

Twain prepares the scripts. Your rep makes the call, listens to the answer, and decides how to continue the conversation.

Use the research to choose an opening

For Mark, Twain's Research surfaced a post about a cold email framework that pulls buying signals and website proof points from LinkedIn URLs. It also found his work with Claude and MCP workflows.

Those details give the caller something more useful than a job title. The connection is between the research systems Mark builds for clients and the research Twain does before drafting outreach. The script can start from that overlap and ask whether there is a problem worth discussing.

Part of Mark's generated Research report. The opener uses the cold email framework signal shown here.

You can inspect the supporting material in Research. High and Ultra also show Absent signals, which identify gaps you should not fill with assumptions. Thesis explains the prospect's situation and fit; Ultra's Strategy sets out how to approach the outreach.

A generated cold call opener

Here is the saved opener for Mark. The labels on the left describe the structure; the text on the right is what the caller would say.

Cold call opener · generated with Ultra
Greeting
Hey Mark, this is [Your Name] from Twain.
Reason
I emailed you last week, quick follow-up. Saw your post on that cold email framework pulling buying signals off LinkedIn URLs, really liked it. Sounds like you're building that research layer yourself for client setups.
Hypothesis
So keeping drafts accurate across all those accounts is probably the slow part, right?
Objection
You likely don't want another tool next to Claude. Ours runs inside Claude through the MCP server.
Call to action
Got a couple minutes to talk through whether that's worth fixing now?

The Reason section refers to a specific piece of Mark's work. The Hypothesis turns the proposed problem into a question. The Objection section connects the offer to the Claude workflow mentioned in the research, and the closing question asks whether there is time to discuss it.

Use the structure to prepare for the conversation. Once the prospect answers, follow what they say rather than continuing through every section regardless of the response.

A voicemail using the same context

The voicemail keeps the reference to Mark's cold email framework and explains why Twain could be relevant to that work. It finishes with a way to respond, because the prospect is not there to answer a question.

Voicemail · generated with Ultra
Greeting
Hey Mark, this is [Your Name] from Twain. Just tried to reach you, sorry I missed you.
Hook
I emailed last week too. Saw your post on that cold email framework pulling buying signals off LinkedIn URLs, and figured you're probably building that research layer by hand for each client.
Value
Our research agent grabs those live signals and drafts the copy inside Claude through the MCP server, so your setup stays as is.
Sign-off
Again, [Your Name] from Twain. You can reach me back on this number.

Read the opener and voicemail together. They should give the same reason for reaching out, while accounting for whether there is someone on the other end of the call. If your actual next step is an email or a LinkedIn message, make sure the wording agrees with that sequence.

Edit the structure and check the speaking time

Call and voicemail steps show an estimate of speaking time in place of the word count used for written messages. Read the script aloud as well. Saying a name, leaving a pause, and giving someone time to respond all affect how the call feels.

You can edit the framework to change the structure of either step. For example, your team might want to ask permission to continue before explaining the reason for the call, or use a different closing question. Use the framework for that shared structure, then edit individual scripts when a detail needs to change for one prospect.

The labels make the structure visible without becoming part of what the rep says. Keep that distinction when preparing the script for your calling workflow.

Try it on one lead

Start with one lead. Read the opener and voicemail aloud, replace [Your Name], and check the callback instruction. If your team uses a different opening or closing question, change the framework before generating the rest of the campaign.

You'll have the research, the written outreach, and the calling scripts in the same place when it's time to reach out.