Updated
It reads every reply
Each reply a lead sends is read once and filed as interested, later, not interested, wrong person, out of office, unsubscribe or other. Interested and later count as positive answers, not interested and unsubscribe as negative ones. For the leads who haven't replied, your thumbs up and down count too.
It measures what answers, for you
The brain measures how often your leads answer positively by signal, role, industry, A/B variant and channel. You learn what works for your offer and your market, not an average across other companies.
It waits until the numbers mean something
No conclusion comes from a handful of replies: a group needs at least 20 leads and 3 positive answers before the brain says anything about it. Until then, your agent works as it always has, and the brain grows with what you send.
Your agent acts on it
More of its day where answers come from
Each run gives more of its searches to the signals that bring positive answers.
Scores that remember
A lead's AI score gets a small adjustment from what your replies taught: one point at most.
Messages that build on what worked
New sequences take inspiration from up to three of your own messages that drew an interested reply, with names and companies removed.
You see it, and you correct it
The Brain page says what your agents learned, in plain sentences with the figures behind them, and what they do with it. It lists every reply as it was read: if one was misread, correct it. If a conclusion doesn't hold for your business, set it aside and your agents leave it out.
Your replies stay yours
The brain learns strictly within your account. Nothing it learns from your replies is used for anyone else, and nothing from other customers is used for you.
Coming soon: a marketing agent
We are also building a marketing agent that drafts LinkedIn and X posts in your own voice, from your past posts and what Coolie knows about your business. Coolie never posts for you: you review every draft. Ask us if you would like to try it early.
Example (fictional)
You sell payroll software and run one agent on hiring and funding signals.
Once enough leads have replied, the brain finds that Heads of People at companies hiring in finance answer positively more often than your average, and that funding-based messages draw mostly not-interested replies.
Your agent then gives more of its daily searches to the hiring signal, and new sequences start from the messages that got an interested reply.
Every company, person and result in this example is invented.
Frequently asked questions
What does Coolie's brain learn from?
From your leads' replies, each read and filed as interested, later, not interested, wrong person, out of office or unsubscribe, and from your thumbs up and down on the leads who haven't replied.
Does Coolie use other customers' data for my agent?
No. The brain learns strictly within your account: nothing from your replies is used for anyone else, and nothing from other accounts is used for yours.
How many replies does it need?
A group of leads needs at least 20 leads and 3 positive answers before the brain draws a conclusion about it. Until then, your agent works as it always has.
Can I overrule what it learned?
Yes. The Brain page lists every conclusion and every reply as it was read: correct a reply that was misread, or set a conclusion aside, and your agents leave it out.