Rainmaker.
How Outbound Works · 5 min read

Copilot vs autopilot: the AI outbound distinction that actually matters

Most AI SDR tools default to autopilot. That single product decision is the source of most AI outbound spam.

ApprovalBatch is Rainmaker's clearest single-concept differentiator and the one buyers most need explained before they trust any AI outbound tool.

There is one question that sorts every AI outbound tool into two camps: what is the default?

Does the product draft and send - unless you turn on an approval gate? Or does it draft and hold - unless you explicitly opt into autopilot?

That single default setting is the difference between a copilot and a cannon. And it has more impact on your domain reputation, your brand, and your meetings than any other feature in the product.

Why defaults are the real product decision

Most products offer both modes. Most tools have an approval gate somewhere in the settings. The question is not whether the gate exists - it is what happens when a new user signs up, connects their account, and lets the product run without reading the manual.

For autopilot-by-default tools, the answer is: it sends.

Most users never change defaults. This is well-documented in software behavior and it is not a criticism of users - it is a description of how people actually interact with products. If the default is autopilot, most deployments are autopilot. Your company's name goes on emails you did not review.

Gojiberry defaults to autopilot. Explee is autopilot only - there is no approval gate at all. In both cases, the product decision is made: speed over control.

Rainmaker defaults to ApprovalBatch. The agent drafts, assembles a batch, and holds it for your review.

Nothing sends until a human approves. Autopilot requires an explicit opt-in - a deliberate choice, not a changed setting you might have missed.

This is not a feature. It is an architectural stance.

What ApprovalBatch actually does

ApprovalBatch is Rainmaker's default operating state. Before any email sends, the system:

Builds leads against your BuyerPersona - scoring each against pains, triggers, and disqualifiers, with reasoning shown Researches each lead and drafts an email where every factual claim links to a real source Assembles the leads and drafts into an approval batch Holds the batch for your review You see every lead, the reasoning for why it was included, and the full draft - including the sources behind each factual claim. You approve the batch, reject individual leads, or edit specific drafts before anything moves.

Only after your approval does the batch send.

This is what "copilot" means in practice. The agent does the research and drafting work. You make the final call. The send does not happen without a human in the loop.

The case for approval-required as a default

Three reasons we made approval the default rather than the option:

Your name is on the emails. Cold outreach goes out under your company name, from a domain associated with your brand. An email you did not write, about facts you have not verified, represents you to someone you may want to do business with. That is a meaningful reputational exposure.

Defaults are what ships. If we built Rainmaker with autopilot as the default and an approval option buried in settings, we would effectively be building an autopilot product. The users who would benefit most from the approval gate are the users who never find it. Approval-first inverts this: autopilot users who want speed choose it deliberately; everyone else gets a system they can trust.

The upside of autopilot is smaller than it appears. The volume argument for autopilot is that it sends more emails faster. But more sends against a loose BuyerPersona, without a citation check, to a list you have not reviewed, degrades faster than it scales. Domain reputation problems compound. Reply rates fall. The speed advantage disappears in the noise of poor targeting.

Rainmaker is built on the premise that precision outperforms volume. Approval-required is the mechanism that enforces precision by default.

What autopilot-by-default actually costs

The cost is not obvious upfront. The tool sends emails. Some get replies. It looks like it's working.

The costs accumulate in the background:

Domain reputation. Autopilot sends to leads that have not been reviewed for quality. Some are wrong - they are lookalike matches that are not real buyers. Low reply rates and spam complaints from irrelevant recipients degrade your sending domain over time. Once the domain is burned, deliverability drops for everyone on the list, including the leads who would have replied.

Brand exposure. AI-generated factual claims without source verification have a non-trivial error rate. An email that gets a company detail wrong - especially when sent at scale and at speed - creates negative impressions at the moment of first contact. Reversing that impression is harder than not creating it.

Trust deficit. The people you are most trying to reach - senior buyers, people who get a lot of cold outreach - are the most sensitive to the signals that distinguish thoughtful research from a mass blast. An email that reads as generated, with no evidence of sourcing, is deleted or marked as spam. An email where the research is visible and the claims are verifiable behaves differently.

The approval gate is not friction. It is the mechanism that makes the precision advantage real.

How the copilot model changes the pipeline review

One benefit of ApprovalBatch that matters for teams, not just solo founders: it changes what the pipeline review conversation looks like.

In an autopilot system, the pipeline review is a damage assessment. Which emails sent? What replied? Why did the campaign underperform?

In a copilot system, the pipeline review is a targeting and quality conversation. Which leads in the batch were right? Which did the persona score too high? What did the drafts get right, and what did they miss?

The batch is a structured artifact - a list of scored leads with reasoning, and drafts with sourced claims. That artifact is reviewable and improvable. The feedback from a batch review makes the next batch better. This is the loop that builds a compounding outbound motion, not a spray-and-pray cycle.

Questions this post answers

In short.

What is the difference between a copilot and an autopilot AI SDR?

A copilot AI SDR drafts leads and emails, then holds them for human approval before anything sends. An autopilot AI SDR finds leads and sends emails without a required human review step - the human can opt in to approval, but sends happen by default. The practical difference is what happens when a user deploys the product without reading the settings: a copilot holds the batch; an autopilot sends.

Does Rainmaker offer an autopilot mode?

Yes, but it requires an explicit opt-in. The default state of Rainmaker is ApprovalBatch - the agent drafts, you approve, it sends. Users who want autopilot can enable it deliberately. We made approval the default because defaults are what most deployments run on, and most users should not be on autopilot unless they have made a conscious choice to hand off the final approval step.

Why does it matter if the AI sends emails without my approval?

Because the emails go out under your name. A claim the AI gets wrong, a lead that is not in-market, a message that reads as generated rather than researched - all of these are associated with your company and your domain when they send without review. The approval step is not about distrust of the AI; it is about maintaining accountability for outreach that represents you.

How does ApprovalBatch work in Rainmaker?

The agent researches and drafts a batch of emails, each with a scored lead and sourced claims. The batch is held in your approval queue. You review each lead (with the reasoning for why it was included), edit or reject drafts, and approve the batch. Nothing sends until you confirm. You can approve in bulk or individually; the batch format is designed for quick review, not one- at-a-time deliberation.

See the receipts on your own domain.

Paste your URL and read the persona and the sourced drafts it writes for your market. No card for the 14-day trial.

You might also read