How to enrich HubSpot leads with AI research automatically.
A new HubSpot contact lands. Your rep needs 20 minutes of research before the first call — company news, LinkedIn, product signals. Nobody does it. Here's how to automate that research: Claude runs on every new contact, pulls public signals, and drops three conversation openers directly on the HubSpot record.
- Every new lead researched within minutes of hitting HubSpot
- Three specific conversation openers, not generic 'saw your funding' fluff
- Reps open the record and see context in a note — no separate tool
- Reads company website, LinkedIn (public), and press mentions
Why lead enrichment is the highest-ROI sales AI workflow
Every sales team we work with has the same story: a lead lands, a rep opens the record, and there's nothing there. Just a name, an email, and maybe a company. To make the first call worthwhile, the rep needs to do 15-20 minutes of research — look up the company, check recent news, scan the careers page, get a sense of what they do. Most reps do this in a hurry, some don't do it at all, and the resulting first call is generic.
Automating that research is the single highest-ROI use of AI in sales in 2026. It's not glamorous — no autonomous SDR, no 'AI is closing deals for us' story. Just: every new lead lands with a research note already on the record. The rep opens the account, sees the context in 30 seconds, and the first call opens with a specific reference rather than 'so tell me about your company.'
The measurable outcome shows up in two places. First-call conversion (from qualifying call to next stage) rises 20-40% because the rep has real context to draw on. And rep morale improves — nobody enjoys the research phase, so making it automatic frees rep time for the actual selling. The ROI math on this workflow is unusually clean: research time per lead drops from ~15 minutes to ~30 seconds of glance, and first-call quality goes up simultaneously.
The reason this is so effective — and why competitors are slow to catch up — is that the workflow has three distinct pieces that need to be wired together correctly: trigger (new HubSpot contact matching a filter), research (Claude reading the right public signals), and writeback (structured note attached to the contact record). Any one piece done wrong breaks the workflow. Teams that try to build it themselves usually get the trigger right and the research wrong, or vice versa.
How it works
- 1
1. Connect HubSpot
OAuth into your HubSpot workspace. The template only needs the CRM contacts scope and notes scope.
- 2
2. Activate the New Lead Researcher template
Set a filter — e.g., 'contacts with source = demo request'. Claude runs on every match, not every contact ever created.
- 3
3. Watch the first enrichment land in HubSpot
Pilotran runs it against your most recent qualifying contact so you see the note format and content before it starts writing to real records.
What lands on a HubSpot contact
Sample: new HubSpot contact from demo request. Priya Menon, Aster Labs (aster.io). Claude ran research and dropped this note on the contact within 45 seconds of creation.
- You get inbound leads from demo requests, forms, or content downloads
- Your ICP is companies with visible online presence (funded startups, mid-market)
- Your sales team wastes measurable time on lead research
- You want first-call conversion to lift measurably (~20-40% is the realistic range)
- You do purely outbound and manually pick every account — this workflow is trigger-driven
- Your ICP is private-company or low-web-presence (SMB retail, offline services)
- You've already bought Clay and are using it well — this is complementary, not substitute
- Your HubSpot doesn't yet have consistent contact-source tags — set those up first, then wire this
Notes from the build
Why we write structured openers, not a paragraph
Reps skim. A three-line 'here are the openers to try' section drives action; a five-paragraph research memo gets ignored. The full research is still in the note for the deep-reading rep, but the actionable output is upfront.
Why the fit score is a suggestion, not a filter
Auto-filtering based on AI-generated fit scores is a fast way to accidentally reject good leads. The score is on the record for the rep and for reporting; it doesn't gate whether the rep sees the lead.
Yellow flags matter more than green ones
'Growth stage: right stage' is nice. 'Yellow flag: fresh Series A, may pilot before commit' is what a rep actually needs to plan for. The prompt is tuned to surface friction earlier than fit, because that's what changes the sales approach.
Handle the 'no signal' case explicitly
About 5-10% of leads produce thin research because the company has minimal web presence. The template returns 'thin signal — company appears to have limited public presence, may want manual research' rather than fabricating anything. Reps trust the tool more when it admits limits.
New Lead Researcher
For every HubSpot contact created in the last hour, generate a 3-bullet enrichment note and file it against the contact.
See the templateFrequently asked
Does it scrape LinkedIn?
No. Claude reads publicly available signals only — company website, recent news, press releases. LinkedIn scraping violates their ToS and we don't do it.
Can it run on existing contacts, not just new ones?
Yes. Bulk-mode is opt-in — you can point it at a HubSpot list and it enriches every contact in the list.
How is this different from Clay or Apollo enrichment?
Clay and Apollo are data-vendors — they enrich from proprietary datasets (firmographic data, technographics, contact info). This workflow is complementary: it enriches with AI-reasoned context from public web signals, then writes the interpretation as an actionable note on the record. Many teams use both — Clay for facts, this for narrative.
What HubSpot scopes does it need?
Contacts read + write, Companies read + write, Notes write. No sensitive scopes (deal amounts, billing, admin). All revocable in your HubSpot integrations settings.
Does it work on non-B2B leads?
Best signal comes from companies with public web presence — funded startups, established SaaS, ecomm brands with press coverage. B2C leads or private-company leads have less public signal, so the notes are thinner. Still useful, but less differentiated.
How often can it run?
Real-time on new-contact webhook (recommended for demo-request leads), or scheduled bulk-mode for existing lists. Rate limits are set by your HubSpot API tier — HubSpot Pro handles 100 req/10s which is more than enough.
Can it update a lead score too?
Yes — the template can write a custom score field on the contact based on signals it finds (company size, growth stage, tech stack fit). Off by default because scoring is prompt-tuned; turn it on after a week of tuning.