Claude for high-intent B2B leads?

wordweaver

I love automating lead gen, but I hate paying for bloated databases that are half stale. So I tried building a workflow with Claude to scrape public sources for intent signals instead of forking out for another subscription.

Honestly? Mixed results. Claude's decent at parsing messy web content and spotting patterns-like job changes, funding news, or tech stack mentions. But verifying contacts is where it all goes to hell. You'll get a list of names and companies, then spend half your day confirming emails and titles.

Someone in another thread pointed me to Exa AI-apparently that's the real goldmine for intent data. Another said Claude works perfectly if you configure the prompts right. But configure how? I'm not sinking hours into prompt engineering just to get a handful of leads.

Anyone actually cracked this? What broke for you?

trendspotter

Honestly, you're describing Exa AI. It's a search engine built for intent - gorgeous interface, zero noise, just leads that actually want what you're selling.

storyweaver

Oh, perfect? That's one way to put it. I'd say it's dangerously good if you know how to prompt it. Tried feeding it a handful of ICP signals and a few scrappy competitor lists - it spat back leads that were actually relevant instead of the usual "CEO of a company that sells paperclips" nonsense.

Still have to do the legwork on intent signals, obviously. But it's a hell of a lot less painful than wrestling with some of the paid "AI" tools that promise the moon and deliver a soggy spreadsheet. Just don't expect magic without a bit of prompt engineering elbow grease.

bigcoffee

Using AI to scrape LinkedIn data is a fast track to getting your account banned. Their terms of service and privacy policy are explicit about automated data collection - it's not a grey area. A colleague of mine lost their entire B2B prospecting pipeline that way, and rebuilding that channel took months.

If you're experimenting with Claude for lead gen, I'd keep it to compliant workflows: manual research paired with a solid CRM enrichment tool. Anything that touches automated extraction is a risk you don't need when LinkedIn is a core acquisition channel.

dailygrind

Great question. I split this into two distinct jobs rather than trusting a single LLM call to both identify intent and validate the contact.

Job one: use Claude purely for summarising public intent signals. Feed it scraped trigger events - funding announcements, leadership changes, SOC 2 certifications, job postings with specific titles. Claude is solid at distilling noisy data into a signal score. But it's terrible at deterministic truth.

Job two: identity resolution stays manual or semi-automated with an enrichment step. That's the brittle part. Before anyone hits an outbound list, I run three checks:

Company fits ICP - verified via Crunchbase or LinkedIn Sales Navigator.
Role matches target persona - cross-referenced with title patterns and seniority.
Recent trigger is real - confirmed from a primary source, not a hallucinated summary.
Source credibility - e.g., GSC data > scraped directory.

Skipping that step means you'll burn CAC on junk leads that look hot in Claude's output but fail on LTV. Keep the LLM as a signal aggregator, not a decision maker.

metricsmuse

Oh yeah, I've run into the same thing. In my experience, LLMs make up contacts about a third of the time - sometimes even more. Honestly, just use them to surface signals, then run your own verification separately. Saves a ton of headache.

wordweaver

Oh, you think configuring Claude is like setting up a spreadsheet, do you? It's more like teaching a sarcastic toddler to read your buyer persona docs. Give it a clear prompt with intent signals and industry jargon, then watch it either nail it or recommend you sell ice to Antarctica. What's your actual niche? B2B SaaS? Manufacturing? I've got opinions.

bigcoffee

Still breaching LinkedIn's Terms of Use and Privacy Policy. No workaround changes that - only compliance does.

Read the User Agreement yourself. The automated scraping language is pretty clear.
Every "hack" I've seen gets patched within a quarter. Then you're back to square one with flagged accounts.
If you're running a B2B lead gen operation, ask yourself: what's the cost of a banned profile vs. the cost of a compliant tool? Numbers don't lie.

Stop trying to game the system. Find a method that doesn't trigger their detection and stick with it.