AI lead generation: three data sources, 900+ leads

Magnifying glass focuses on pins highlighting travel destinations on a world map.
Photo by Lara Jameson on Pexels

AI lead generation means using software, with some AI in the loop, to find businesses that fit your customer profile, check they exist and can be reached, and hand you a list worth calling. I've just finished the first stage of a pipeline that does this for a waste management company I work with. It took three data sources to find over 900 leads. It doesn't take much.

Finding leads is doable. It doesn't need a big budget, and it can change how a business grows. Here's what each source did, what went wrong, and what I'd check before paying anyone for leads.

How to generate leads: start from customers you already have

Most advice on how to generate leads starts with "define your audience." That's too vague to act on. This pipeline started from something narrower: the business's own sales records. Which kinds of customer have already paid, and what do they have in common? A lead that looks like one of them is a repeat of something that has already worked. A lead that doesn't is a guess.

How to find leads: three sources

1. Google's business listings. Automated searches against the Google Places service, run by category and area, did most of the finding.

2. The Australian Business Register. Its free web service confirms a business is real and registered, and gives me a registration number to de-duplicate on.

3. NSW Government mine data. Google has no category for mining, and searching "mining company" mostly returned head offices in the Sydney CBD, which is no use when you're after the mine sites themselves. The NSW Geoscience service has a free public data feed of every operating mine in the state: 52 sites, with names and coordinates, no scraping and no login needed. Look for a public register like this in your own industry before you pay for anything.

After de-duplicating across the sources, the total is over 900 leads. I also matched the list against the business's customer records, so the sales team doesn't cold-call someone who already buys from them.

I knew I was on the right track when the leads the pipeline found matched customers the business already had. The search was turning up businesses that already paid them, which is the strongest sign your customer profile is right. At last count, 15 existing customers had turned up in the list.

The whole thing runs on n8n and a Postgres database. There's no lead-list subscription.

What went wrong

Three things, and each is a reason to distrust a list you didn't build.

The noise got through. A business directory returns airports, train stations, car parks and an air force base alongside the businesses. Filtering by listing type caught most of it. A second batch still let some through, so it needed a name-based pass on top.

The register confirms less than you'd hope. Of 768 listings checked, only 32 came back with a corroborated registration number. The rest failed the name match, and I'd rather have an honest "couldn't confirm" than a wrong match. I also went in assuming the register would tell me each business's industry. It doesn't. It confirms a business exists and is trading, and that's all.

Matching names is where automation lies to you. Two listings can look like the same business and be different ones. Any match on a similar name needs a strict rule and a manual look at the edge cases before you trust it.

Where AI fits in lead generation tools

Most of what I built is ordinary automation: search, store, de-duplicate, filter, match. That's plain code, a database and a few scheduled jobs, and it's the cheapest and most predictable way to do those steps. AI comes in where the data runs out. Three places I'd use it:

  • Driving the scraper. A conventional scraper breaks when a website changes layout. An AI agent can read a business's website itself and pull the details the data sources don't have, like the services they offer or the size of their fleet.
  • Filling in incomplete leads. Directory listings often arrive with half the picture. AI can decide the missing metadata, such as what kind of business it is and whether it fits your customer profile, from whatever text is available.
  • Anywhere else. AI could do every step in the pipeline, if you're happy to pay the token cost.

That last point is the real constraint. Every step you hand to AI costs tokens on every lead, and plain code costs almost nothing. So I use plain code for the steps that are rule-based, and AI for the steps that need judgement or reading. I'd be wary of any lead generation tool that sells "AI" as the product. What you want is a list you can check, built by steps you can inspect, for customers you already know pay.

Get leads for my business: what I'd do this week

  1. Write down your three best customers. What do they have in common?
  2. Pick one source that lists businesses like them: a business directory, a Google Maps search, an industry body's member list.
  3. Pull 50 names by hand. Check each exists and has a working phone number or email.
  4. Contact them, and record who replied.
  5. Only then decide what to automate.

Fifty checked names beat 5,000 unchecked ones. Automate after the manual version works.

Lead generation strategies that don't need a budget

Ask the customers you already have who else they'd send you to. Search the places owners already ask for help: Reddit threads, local Facebook groups, industry forums. And write down what you actually know, in specific first-hand terms, because that's what search and AI answers cite.

If you want one built

This is the kind of work I do: automation, whether that's AI or plain code, a database and some scheduled jobs. I pick the cheapest thing that works, build it with you, and hand it over so you own it. If you want a lead pipeline like this for your business, get in touch.

What I don't know yet

Leads aren't customers. None of these has converted yet. I'm not going to guess a conversion rate. The next stage is outreach, and I'll publish the reply and booking numbers, including the ones that don't look good.

FAQ

Is AI lead generation worth it for a small business? It's worth it when it saves you the manual search and you can inspect what it found. It isn't worth it if it hands you an unchecked list. Do 50 names by hand first; if that produces customers, automate it.

How do I get leads for my business without paying for a list? Start from your best existing customers, find a public source that lists businesses like them, and check each name yourself. The pipeline above is that process, automated.

Should I buy leads? Sometimes. One Australian real estate lead seller advertises packs from $69 a lead. Before you pay anyone, check whether the leads are exclusive to you, whether your area is reserved or a competitor can buy the same one, and how the leads are verified.

How many leads do I need? It depends on how many convert, and I don't have conversion numbers yet. Track the reply and booking rate on your first 50 and work backwards from that.

Share LinkedIn X Facebook Email

Recognise this in your own business?

A Systems Health Check maps what's actually costing you time — no commitment beyond that.

Book a Health Check