Short answer: keep prospecting manually until three things are true at the same time: your average deal size is above roughly $1,000, your offer has converted consistently for 3 to 6 months, and you can describe your ideal client in one precise sentence.
Below that bar, an AI agent will not save a weak offer, it will just get you rejected faster, at scale. Formula. itself ran a full year of 100% manual acquisition before automating anything.
Why did Formula. spend a full year prospecting manually first?
Before there was an agent, there was Laura and Gabriel, LinkedIn open on one tab and an inbox on the other.
Every message we sent that first year was written by hand, one prospect at a time, on LinkedIn and email. No automation, no sequences, no shortcuts.
That year was not wasted time before the "real" product. It was the product's foundation.
We learned which opening line got ignored and which one got a reply. We heard the same objection five times before we understood what it actually meant.
We rewrote our pitch a dozen times until the same message stopped needing a rewrite. Only then did we know exactly who we were talking to, and exactly what to say to them.
That is the raw material an AI agent needs to work from. Without it, we would have automated our own guesswork.
What actually changes when an AI agent takes over your outbound?
This is the part people get wrong, so it is worth being precise.
An AI prospecting agent like Formula. prospects on its own. It builds the target list, writes every outbound message, and handles the follow-ups, with no human validation on the way out.
You do not approve messages before they are sent. That would defeat the point of stepping out of the loop.
The only moment you step back in is when a prospect actually replies. The agent drafts a response, and you validate it with a swipe, from your phone, in a few seconds, before it goes out.
That single gesture is your entire role in the conversation. Everything before it, the targeting and the writing, happens without you.
Which is exactly why the input matters so much: an agent that prospects alone needs a validated offer and a precise ICP (Ideal Customer Profile) to aim at, or it will simply repeat your mistakes at high volume.
What are the 3 conditions before an agent becomes worth it?
We use the same rule internally that we now apply to every account that signs up.
An agent becomes profitable, and relevant, once three conditions are met together:
- Average deal size above roughly $1,000. Below that, the economics of an agent rarely make sense against what each closed deal is worth.
- Offer validated manually for 3 to 6 months. Enough time to know your message converts on its own, without constant rewriting.
- A precise persona (ICP). A clear, specific description of who buys from you, not a broad guess like "small businesses."
Miss any one of these and an agent does not accelerate your growth. It accelerates your rejections.
"An agent doesn't fix a weak offer. It just helps you fail faster, at scale," says Laura Terriou, co-founder of Formula.
If all three boxes are checked, the math flips. That is the point where a case study like our AI prospecting case study becomes relevant to your situation, not before.
What happens if you automate too early?
Salesforce found that sales reps spend less than 30% of their time actually selling (Salesforce, 2023).
That statistic is usually cited to argue for automating everything, immediately. We read it differently.
The other 70% of a rep's time is where the real learning happens: talking to prospects, absorbing objections, adjusting the pitch on the fly.
Skip that phase and hand an untested offer to an agent, and you have not removed the friction. You have industrialized it.
On LinkedIn specifically, where more than 1 billion professionals are reachable (LinkedIn), that scale cuts both ways. Reach the wrong persona with the wrong message and you burn goodwill fast, across a platform where your name follows you.
The invitation limit itself, roughly 28 per day, is set by LinkedIn for everyone (LinkedIn Help). An agent respects it. It cannot make a bad message land, no matter how well it paces the sends.
Manual, too early, or after validation: how do the 3 paths compare?
| Criteria | 100% manual prospecting | Automating too early | Automating after manual validation |
|---|---|---|---|
| Cost | Time only, no tooling spend | Subscription cost plus wasted outreach on an unproven message | Subscription cost, offset by every reply landing on a working offer |
| Risk | Slow, but every lesson is learned first-hand | An unclear offer gets rejected faster, at scale, on channels tied to your name | Low: targeting and message are already proven manually |
| Outcome | A validated offer and a precise ICP, ready to hand off | Volume without conversion, and a bruised sender reputation | The same validated message, reaching far more prospects than one person can by hand |
How do you know you're ready to switch to an agent right now?
Run through the same 3 conditions honestly, on your own numbers.
If your average deal is above roughly $1,000, if you have been closing that same offer by hand for 3 to 6 months, and if you can name your ICP in one sentence, you are not guessing anymore. You are ready to scale what already works.
That is the exact profile behind the anonymized case we see most often: consulting firms going from around €20k to €55k in monthly revenue after switching on their agent, while saving roughly 2 hours a day previously spent on manual prospecting.
Once you are in that position, Formula. prospects for you on LinkedIn and email at the same time, reaching 840 qualified prospects a month across both channels, with 30 to 60% of them replying depending on your sector and offer, against a 7 to 11% market average for classic outreach.
Reply in under 5 minutes and that conversion multiplies by 100 compared to answering the next day, which is what the mobile swipe is built for.
If you are not there yet, that is not a failure. It is information. Two related reads worth your time: how to define your ICP before an AI prospecting agent and how long until an AI prospecting agent pays for itself.
And if you are prospecting alone, without a team, start with is an AI prospecting agent worth it for a solo consultant.
Frequently asked questions about manual prospecting and automation timing
How long should I prospect manually before considering an AI agent?
Between 3 and 6 months of regular manual prospecting on the same offer.
That window is what it takes to know your message converts, to hear the real objections, and to describe your ideal client precisely instead of guessing. Formula. itself ran a full year manually before automating anything.
What counts as a validated offer?
An offer is validated when you can close it repeatedly, by hand, without rewriting your pitch every week.
Concretely: a consistent average deal size above roughly $1,000, a message that gets replies without heavy tweaking, and a clear pattern in who says yes and who says no.
What is an ICP and why does it matter before automating?
ICP stands for Ideal Customer Profile: the precise description of who buys from you and why.
An AI agent needs this to target correctly. Feed it a vague persona and it will reach the wrong people at scale, faster than you ever could by hand.
Can automating too early actually hurt more than doing nothing?
Yes. An agent does not fix a weak offer, an unclear message, or a fuzzy persona.
It reproduces whatever you feed it at volume, so an untested pitch just gets rejected faster and at a bigger scale, on channels where your reputation matters.
Does Formula. check whether I am ready before I sign up?
We apply the same 3-condition filter we used on ourselves: average deal size above roughly $1,000, an offer already validated manually for 3 to 6 months, and a precise ICP.
If you are not there yet, we will tell you to keep prospecting manually first.
Offer validated, persona clear, and ready to hand your outbound to an agent?
Start the 2-week trialSee pricing, starting at €97/month.