Short answer: the agent never sends you every outbound message to approve. It picks the prospects, writes and tests its invitation messages on its own.

The one moment you step in is when a prospect replies. There, the agent drafts the reply, and you validate or edit it with a swipe from your phone.

That speed matters: replying within 5 minutes multiplies conversion by 100, compared with a reply sent the next day.

Most people picture an AI prospecting agent firing off messages on a loop, while you spend your day approving every single line before it sends.

It is the opposite. The automated part is the initial outreach. The part where you keep control is the conversation that starts once someone replies.

Who writes the first message, and who writes the reply?

Outbound messages (first LinkedIn invitations and follow-ups) are not approved one by one. That is the whole point of the system.

A message template is validated once, sometimes twice, at setup. After that, the agent runs continuous A/B testing, improves every day, and picks the best message for the right prospect.

It sends 28 LinkedIn invitations a day, the limit every serious tool respects, alongside parallel email follow-up on the same qualified prospects.

The swipe happens in exactly one place: when a prospect replies and a conversation starts.

How does the agent know what to reply to a prospect?

Before the first message even goes out, the agent has already read the prospect's LinkedIn page and recent posts.

When a reply comes in, it draws on real signals (a recent post, a mutual connection, company news) instead of generic template variables.

It is a conversation on a social media chat, not a marketing sequence. It has to sound natural, not like a pitch.

That real personalization is why Formula. measures a 30-60% reply rate depending on sector and offer, against the 7-11% market average on classic mass outreach.

Why a swipe, and not a fully automatic send?

Because a conversation with a future client deserves one last human look, even if that look takes three seconds.

The agent proposes the reply. You read it, adjust anything that needs it, then validate.

It is a deliberate tradeoff between speed and control: you write nothing, but nothing goes out without passing under your eyes.

Why does reply speed matter this much?

A prospect who just replied is in a specific state of attention. The topic is fresh, and they are expecting a follow-up.

Replying within 5 minutes multiplies conversion by 100, compared with a reply sent the next day, once the prospect's interest has cooled.

That is why the agent drafts the reply immediately. The only delay left is your swipe, not the writing.

We break down why this window matters so much in responding to a prospect in under 5 minutes.

A concrete case: what it changes for a B2B SME

A B2B SME (consulting, solar panel installation) used to prospect manually 2 hours a day, with replies handled whenever someone on the team had time.

After the agent went live, every prospect reply gets an immediate draft response, validated with a swipe by whoever is available.

Result: that SME grew from €400,000 to €600,000 in monthly revenue. Not just from more outreach volume, but because no warm reply cools down while waiting anymore.

For the full mechanism, from targeting to reply, see how an AI prospecting agent really works.

Do you still decide, or does the agent do everything?

The right way to see it: the agent prospects on its own. When a prospect replies, you validate the reply with a swipe.

It is neither "AI does everything without you" nor "you must re-read every line." It is a precise split: the agent handles volume and drafting, you keep the final call on every conversation that matters.

To compare this against a classic human sales rep, see human SDR or AI prospecting agent for B2B.

The agent prospects. You collect meetings in your calendar.

Want to see what a drafted reply looks like on your own prospects?

Start the 2-week trial

First replies usually land within the trial. Judge on evidence, not promises.

Frequently asked questions

Do I have to approve every message the agent sends?

No. Outbound messages are not approved one by one. The agent picks the prospects, writes and tests its messages on its own.

The swipe happens at one moment only: when a prospect replies and a conversation starts.

How does the agent personalize its replies to a prospect?

It draws on 2 to 3 real signals (a recent post, a mutual connection, company news) instead of template variables.

That real personalization is why the reply rate reaches 30-60%, against a 7-11% market average.

What happens if I don't swipe right away?

The reply sits ready until you validate it. But replying within 5 minutes multiplies conversion by 100, so it is worth not waiting too long.

Can I edit the reply before it sends?

Yes. The swipe validates the agent's draft, but you can adjust it first if something needs a tweak.