APIANT.ai / Stories / Outreach team
Customer story · Outreach team · HubSpot
An outreach team wanted new funding rounds from the news logged in HubSpot, ready to check before outreach. Its two scans have written 53 round notes since.
An outreach team contacts founders of newly funded startups, and keeps its pipeline in HubSpot. They connected HubSpot and their own Claude key to APIANT and asked their AI client to read the news for them.
The challenge
The team looks for founders of consumer brands that have just raised money, and those rounds are announced in startup funding news in one European market. They wanted each round read out of the news and logged on the company in HubSpot, with the founders and investors actually named, a link to the source and a note to verify before outreach, skipping any round already logged in the last 90 days.
Their existing contacts also had to move out of an outreach workbook and into HubSpot.
- 53funding-round notes written in HubSpot, each with the source link
- 935successful runs across six weeks
- Under 2 hoursfrom connecting HubSpot to a scheduled news scan, tested
- 52contacts moved from their outreach workbook into HubSpot on day one
The journey
-
Day 1
The workbook moves into HubSpot
They connected HubSpot and Claude, then asked for two building blocks: find or create a company (by domain, then by exact name) and find or create a person and link them to the company. The agent used the company finder to import the team's outreach workbook: 52 contacts, each with its company, role and notes, marked as imported so a re-run skips them.
52 of 52
-
Day 1, later
A headline scan, every few hours
The ask: read funding headlines, have Claude pull out any round raised by a named consumer brand, and log it in HubSpot with the founders and investors actually named, plus a note that says "verify before outreach" and links the source. It was tested within a minute of being built and began running on its own a few hours later.
-
Day 2
Filling in the outreach fields
Two more one-off jobs: setting outreach wave, channel and status on the 52 imported contacts, and importing the workbook's remaining tabs. Each ran once and succeeded.
-
Day 5
The daily article scan, split into small parts
A second scan for the full articles. The agent built it as three pieces: one splits each new article into parts, one has Claude read each part and return the rounds as data, and one logs each round in HubSpot. It was tested on 34 article parts in under a minute, and every run succeeded.
-
Days 6 to 44
It keeps running
The daily scan has run 39 times and the headline scan 680 times, every run successful. Together they have written 53 funding-round notes, and skipped rounds already logged in the last 90 days.
What they built
| Automation | What it does | Result |
|---|---|---|
| Company finder | Find a HubSpot company by domain, then name; create only if neither matches | Works, reused by every import and scan |
| Person finder | Find or create a contact and link them to the company with their role | Works |
| Workbook import | 52 contacts with companies, roles and notes | Works, 52 of 52 |
| Outreach fields | Wave, channel and status on the imported contacts | Works, 52 of 52 |
| Headline scan | Every few hours, Claude reads funding headlines and logs new rounds | 680 runs, all succeeded, 9 round notes |
| Daily article scan | Splits each article, Claude extracts rounds, each round logged in HubSpot | 39 daily runs, all succeeded, 44 round notes |
Nine automations in all, counting the scan's two background pieces and the second workbook import.
Why it worked
- 01
They asked in plain English.
The matching rules, the 90-day memory and the splitting were the agent's job; theirs was saying which rounds matter and what a good note looks like.
- 02
Everything was tested before it ran on its own.
Each piece was run against their real HubSpot from their AI client the day it was built.
- 03
Small parts, short runs.
The daily scan splits each article so no single run carries a whole newsletter: the article parts have averaged under five seconds a run, and every round note links back to its source.
More stories
- A gym owner wanted ActiveCampaign to know what Mindbody knows. New clients now arrive hourly through the day, tagged with what they actually do.
- A church needed each Stripe gift in DonorPerfect, on the right donor, with its own codes. By the next day, its charges were arriving as gifts on their own.
- A training studio wanted every new client's first package sale sent from Mindbody to its own systems. The first live sale went through half an hour after signup.
What would you ask for?
Details are taken from the account's own run history and shared without the team's name.