Nobody wants to waste time on prospects that aren't likely to buy. So how do we know who is? What data is a good proxy for it, and how do I get that data into my GTM agent of choice (Claude, Codex, Grok Bot, Hermes)?
Signals that are a good indicator are open roles by department, recent funding, news, newly adopted technology, live ads and new executive hires. They sit in different places: a funding database, a job board, a news feed, an ad library. I ran it in Delegate Agents, and the same prompt works from any agent connected to Delegate. I picked Ramp, the finance software company, because it moves fast and a lot is public.
Pull buying signals for Ramp (ramp.com): open roles by department, recent funding, news from the last 30 days, technologies they recently started using, ads they are running, and new executive hires. One line per signal with a source. Keep the whole job under $2.
About three minutes later I had a brief with a source under every line. It used seven services and the receipt said up to $0.248.
| Signal | What it found | Service it used |
|---|---|---|
| Open roles | 110 roles counted across 15 departments | PredictLeads, page extraction |
| Funding | $1.85B at a $60B valuation, October 15 | Company news |
| News | 7 dated items in 30 days, the funding among them | Company news, web search |
| Technology | 2 tools adopted, 4 newly mentioned in job posts | PredictLeads, web search |
| Ads | 2 LinkedIn campaigns running | LinkedIn and Meta ad libraries |
| Executive hires | None confirmed | Company pages |
1. Open roles by department
Hiring tells you where the money is going. Sales led with 34 open roles, then engineering with 18 and people and talent with 12. Compliance had 6, including a chief compliance officer opening.
The agent flagged its own number: the departments added up to 110, while the job board claimed 171. It called the counts "provisional, not a reconciled census."
2. Recent funding
On October 15 Ramp announced $1.85 billion in new funding at a $60 billion valuation. The agent kept one number apart: $5 billion is all the equity Ramp has raised so far, not this round.
3. News from the last 30 days
Seven dated items with links. Besides the funding and a growth update, there was an acquisition and a new banking view on October 20, an application for a national trust bank charter, a product for nonprofits on October 21, a larger New York office, and an integration with Waymo for ride receipts. It noted that a charter application is not an approval.
4. Technologies they recently started using
This is the hardest signal to get right, and the agent said so. It found two tools Ramp uses for security and internal support, from a vendor case study and a Forbes piece, and marked both "start date unverified." It also listed four technologies newly mentioned in a job post, with the warning that a job post is not proof they started using them.
5. Ads they are running
Two LinkedIn campaigns, one on finance automation and one on approval workflows, each with its landing page. It searched Meta too and threw out other advertisers with similar names.
6. New executive hires
None confirmed. One founder joined through the October acquisition, but in a product role, so the agent didn't count it.
- signals, each with a source
- 6
- services used
- 7
- from question to answer
- about 3 min
- at most, for the whole brief
- $0.248
What it cost, line by line
Most of the 25 cents went on PredictLeads' company signals, $0.176 for four kinds of signals, job openings among them. Company news came to $0.036 over two calls. The two ad library searches were $0.00144 each. The rest was web search and reading pages, a cent or two each.
Run it on your own list
- Before a first call, run it on that one company and read it in the car.
- For a target list, ask for the same six signals on 20 companies and sort by who raised or is hiring the people you sell to.
- Schedule it weekly and ask only for what changed since last time.
Questions people ask
Where does the hiring data come from?
From the company's own job board, read through PredictLeads and page extraction. When the counts don't reconcile, the agent says so.
Is "technologies they started using" reliable?
It's the weakest of the six. Treat it as a lead to check, and look at the dates the agent marks as unverified.
Can it do this for private companies?
Yes, as long as there's something public: a job board, news, ads. Very small companies return less, and the brief tells you which signals came back empty.
Open Delegate Agents and paste the prompt with a company you're about to call. Which of the six would change what you say to them?