Watch for the moment they start looking

Something you can point to that says an account's priorities just moved toward what you sell: a job post naming your problem, a funding round, a new leader in the seat that owns the budget, repeat visits to your pricing page, a public request for recommendations. Matching your ICP doesn't count on its own. That's fit, and a message is only worth sending when an account has fit and a reason to talk now.

Usually two at once. A job post naming your problem at a company that also raised money this quarter beats either one alone. Some signals point the other way. A company that just hired for the role your product replaces has probably decided against you. Heavy content engagement with no pricing-page visit, ever, tends to be someone studying the category. Buyers eventually check the price. And likes and follows on their own are about the weakest signal there is.

Intent data usually means third-party research data: a publisher network reports that a company is reading more than usual about a topic, at company level, refreshed weekly. Buying signals are events you can point at (a job post, a round, a thread) tied to a named company and often a named person. This page is about signals, and we don't sell third-party intent data. If you already buy it, export it into the workspace and the agent treats it as one more source.

Yes, if you want it to. The agent works through your own LinkedIn account, signed in once in its browser, so it only sees what you'd see yourself: profiles, posts, and who comments on and likes them. No bought database, no fake profiles. You set the pace and stay in control of the account. Where LinkedIn's official API covers the job, like posting to your profile or the company pages you manage, it uses that instead. Job posts through TheirStack, funding through Crunchbase, company-level visits through Leadfeeder, Reddit threads and your CRM fill in the rest.

At company level, through Leadfeeder, which matches visiting companies and shows the pages they read. Person-level tools promise names, but one operator measured a popular one matching only 20 to 30 percent of visitors. A company that keeps coming back to pricing tells you more than a single named visitor, and the agent weighs it that way.

Don't expect a flood. One sender found that 73 percent of their best cold email replies started from a signal. Another, messaging people who'd followed a competitor, reported roughly one meeting per 145 people contacted. The point is that a message about something that really happened gets read, at a time when average cold email reply rates keep falling.

A signal tool stops at the name. You still build the ICP filter, the CRM check and the follow-up yourself, usually across two or three more subscriptions. And when the signal vendor changes its product or shuts down, everything built on its API stops with it. Here one AI agent watches, filters, checks the CRM and drafts the opener, and the rules it follows live in your own workspace.

No, but they fit together. Ten accounts every morning ranks your whole market into a daily shortlist. Signal watch is the watching and judging underneath it, and it can alert you the moment a strong signal lands instead of waiting for the morning list. You can run both on the same agent.

If your market is a few hundred accounts, you can watch them by hand. If you need third-party intent data across tens of thousands of companies, buy that from a data provider. And if you want outreach going out with nobody reading it, look elsewhere. Every drafted message here waits for a person.