Buying-signal research that runs itself, every morning.
A set of Claude skills — not a one-off agent — that score buying signals, write Signal records back to HubSpot, and stage personalized outreach. Scheduled on Claude Routines, so the work happens whether or not anyone kicks it off.
I’ll run the signal-scoring skill on this week’s 12 target accounts — pulling fresh signals, scoring each by buying intent, and writing a structured Signal record to each company in HubSpot.
Here’s how this morning’s signals scored by buying intent:
Writing the top signal to HubSpot:
Done — 9 Signal records written to HubSpot, 4 flagged hot. The two strongest (Cascade, Meridian) are staged for outreach review in the next step.
The problem
Signal research is the kind of high-value work that never gets done consistently. Someone has to remember to check for new buying signals — facility news, new hires, intent data — score them, log them where the team will see, and turn the good ones into outreach. In practice it happened in bursts, or not at all, and the signals that did get found rarely made it back into HubSpot in a usable shape.
What we built
- A library of reusable Claude skills — signal research, signal scoring, outreach generation, and email staging into HubSpot — each invokable on demand by the team or by another skill
- A scoring model that ranks each signal by buying intent and writes a structured Signal record back to the HubSpot company, so the whole team sees the same prioritized view
- Outreach generation that drafts personalized, on-brand emails from the signal context and stages them in HubSpot for review
- The whole suite scheduled on Claude Routines — research and scoring run unattended on a cadence, with results waiting in HubSpot each morning
How it works
The skills run on a schedule through Claude Routines: pull the target accounts from HubSpot, research fresh signals across public and intent sources, score each one, and write a Signal record back to the company in HubSpot. The strongest signals get drafted into staged outreach. The team opens HubSpot to a ranked, current view of who to contact and why — work that used to depend on someone remembering to do it.
The outcome
- Signal research happens on a cadence instead of in sporadic bursts
- Every signal lands in HubSpot as a structured, scored record the whole team can act on
- Reusable skills, not a brittle one-off — the same building blocks power on-demand and scheduled runs
- Outreach is staged from real signals, not generic templates
Stack
Have a build like this on your list?
Every engagement ships custom AI — apps, agents, and skills — built for your sales motion and wired into your HubSpot and the rest of your stack. You own all of it.