What does AI lead generation actually involve?
Lead generation is three jobs that are often sold separately. Finding means spotting companies that might buy. Qualifying means deciding which of them deserve a sales conversation. Enriching means attaching what you need to act: the right person, a working email address, the context. Most tools marketed as AI lead generation do one of the three well and leave the other two to you.
AI earns its keep where the input is messy. Forum posts, launch pages, directory listings and company websites are unstructured text; turning them into "this company, this need, this person" used to take a researcher a good part of an hour per lead. That is the work Startories automates, from the first post it reads to a lead that is ready for outreach.
You will also see this work called AI prospecting. The two terms describe one pipeline from different seats: marketing calls the output a lead, sales calls the activity prospecting. When you compare tools, two questions matter more than the label: how much of the chain does the tool cover, and can you check the decisions it makes along the way?
What does a qualified and enriched lead include?
At Startories, a qualified and enriched lead carries six pieces of information. Here is what that looks like for a fictional company:
| Field | What it means | Example |
|---|---|---|
| Signal | The public event that triggered the lead, with a link | Asked in a public post for an alternative to its help desk tool, two days ago |
| Company | The real business behind the signal | A 25-person B2B SaaS that sells scheduling software to dental clinics |
| ICP check | Your criteria, applied one by one | B2B: yes. 10 to 50 people: yes. US-based: yes. Uses a tool you replace: yes |
| Score and reasons | How strong the lead is, and why, in plain words | Strong fit: right size, active switching intent, buyer reachable. Against it: outside your top vertical |
| Decision-maker | The person who owns the problem | The co-founder and CEO, who still runs support at this size |
| Verified email | A business address checked before any contact | Verified. Addresses that fail verification are never contacted |
How does the AI decide what to keep?
Most of what the system reads never becomes a lead, and that is the point. A recommendation thread may hold one real buyer and a dozen students, hobbyists and vendors replying with their own product. A directory may list hundreds of new companies, of which a handful match your market. Each candidate goes through five questions:
Is it a signal?
The post, launch or listing is classified by type (tool search, competitor complaint, hiring, funding and so on) or discarded as noise.
Is there a company behind it?
The signal is matched to a real company. Chatter that cannot be tied to a business is dropped.
Does the company fit?
Your ICP criteria are checked one by one: business model, size, industry, location and anything else you defined. Consumer brands and out-of-scope companies are removed.
How strong is it?
The lead gets a score, and the score comes with written reasons, so you can disagree with it.
Can we reach the right person?
The decision-maker is identified and their business email verified. A lead without a verified address does not go to outreach.
What does a week of AI lead generation look like?
Here is that filter at work on a fictional company over one invented week. The numbers are round and chosen to show how the funnel narrows. They are not a benchmark, and your own ratios will depend on your niche and your sources.
The seller is a 12-person SaaS that sells contract management software to marketing agencies. Its ICP: US marketing or creative agencies with 10 to 80 people that sign client contracts every month and still keep them in shared drives. The buyer is the agency owner or the head of operations. The funnel watches two sources: Reddit, for tool requests and complaints about contracts, and agency directories, for new listings and new offices.
| Stage | Count | What was removed, and why |
|---|---|---|
| Posts and listings read | 2,400 | Nothing yet: this is the raw input |
| Classified as a buying signal | 180 | Jokes, off-topic threads, vendors promoting their own tools |
| Tied to a real company | 95 | Posts with no business named, linked or identifiable |
| Passed the ICP check | 41 | Freelancers, in-house teams, agencies outside the US or above 80 people |
| Decision-maker found, email verified | 33 | Companies where no verified business address could be found |
Three leads from that week, with their reasons
- Kept, strong score. A 22-person agency in Denver posts that a client disputed a scope change and nobody could find the signed version. For it: right size, agency model confirmed on its website, a problem the product solves, owner reachable. Against it: no current tool named, so there is no switching angle.
- Kept, medium score. A 60-person agency appears in a directory with a newly opened second office. For it: inside the size band, and more staff usually means more client contracts. Against it: no stated problem, so the email has to open with a question rather than a fix.
- Rejected. A freelance designer asks for a free contract template. Reasons: one person, no budget, outside the 10 to 80 band. The post matches the topic but not the ICP, which is exactly the kind of result a plain keyword alert would have sent you.
What the shape of the funnel tells you
Every cut has a reason you can inspect, so the funnel doubles as a diagnostic. If the ICP step removes many agencies for size, your band may be too narrow. If the company step removes most signals, the source is full of anonymous chatter, and a directory or a business expansion signal may serve you better than more forum threads. If verification removes a lot, your buyers may sit at companies too small to publish business addresses.
What the first email to the strong lead could say
"When a client disputes a scope change, the hard part is usually finding which version they signed. We keep every agency contract searchable by client and date, with the signed copy one click away. Would a short walkthrough using one of your own contracts be useful?" It speaks to the problem the agency described without quoting the post, and the ask is small. Because you can approve emails before they go out, you decide how close to the source each message gets.
How many leads per month can you expect?
Volume depends on the plan and on your market. These are approximate monthly volumes of qualified and enriched leads, meaning leads that went through every check above, not raw contacts.
That unit is worth noticing when you compare prices. Many databases and enrichment tools bill in credits, per exported row or per revealed email, so the cost of one usable lead depends on how many rows you had to buy to find it. A count of leads that already passed qualification is easier to compare, provided you check in the first weeks that those leads really match your ICP.
| Plan | Price per month | Funnels and sources | Leads per month (approx.) |
|---|---|---|---|
| Starter | $99 ($1 for the first 3 days of your first project) | 1 funnel, 1 ICP, 1 intent source | ~300 |
| Growth | $499 | 3 funnels, multiple intent sources | ~2,500 |
| Scale | $999 | More funnels, all sources | ~6,000 |
Reach companies with a reason to buy this week
Startories finds the buying signal, verifies the decision-maker and runs the outreach until they book a call.
Which kind of B2B lead generation tool do you need?
Lead databases such as Apollo and ZoomInfo give you search and filters over a very large set of companies and contacts. They are the right tool when you already know which accounts you want, need broad coverage to size a market, or run a calling team that needs phone numbers. Apollo also bundles its own sequencing, and ZoomInfo is built for larger revenue teams.
What a database cannot tell you is timing. Every company that matches your filters looks the same in a search result, whether it is shopping this week or happy with its vendor for three more years. And the same records are open to every competitor with the same subscription. Signal-based AI lead generation produces a shorter list, but each entry comes with a reason to write now.
Databases are one of six kinds of tools sold as B2B lead generation tools. The categories overlap and many teams combine two, so start from the job you are missing: names, timing, research, reach or sending.
| Type | What it does | Example | Best when |
|---|---|---|---|
| Contact database | Search and filters over a large set of companies and contacts | Apollo, ZoomInfo | You know your target accounts or need phone numbers |
| Enrichment workflow | Chains data providers and AI research steps in a table you design | Clay | Someone on your team likes building and maintaining workflows |
| Third-party intent data | Flags accounts researching a topic more than usual across a publisher network | Bombora | You run account-based marketing on a defined account list |
| Website visitor identification | Shows which companies visit your own website | Warmly | Your site already gets meaningful B2B traffic |
| Cold email platform | Sends from and warms up inboxes for lists you bring | Instantly, Smartlead, lemlist | You have a list and a person who owns targeting |
| Signal-based engine | Finds public buying signals, qualifies, enriches and runs the outreach | Startories | You want timing and the whole chain in one place |
Why not just scrape the leads yourself?
Scraping looks cheap: a script or an automation tool, a list of sites, a few hours. In practice you inherit four problems. Scrapers break when a page changes. Many sites forbid automated collection in their terms. Raw output is unqualified, so someone still has to read every row. And guessed or unverified addresses bounce, which is one of the fastest ways to damage a sending domain.
Startories uses official APIs and compliant sources where they exist, keeps a link to the original signal for every lead, and only contacts verified business emails. If you enjoy building your own enrichment and want control over every column, a workflow tool will suit you better; just budget the hours to maintain it once the novelty wears off.
| Lead database | Scraping | AI lead generation (Startories) | |
|---|---|---|---|
| What you get | Contacts that match your filters | Raw rows from the pages you target | Qualified companies with a decision-maker |
| Reason to reach out now | No | Only if you build it | Yes: the signal and its link |
| Email verification | Provided, method varies by vendor | Your job | Included; only verified emails are contacted |
| Maintenance | Low | High: scripts break | Low |
| Outreach included | Sometimes, as a sequencing feature | No | Yes |
| Best for | Known accounts, market sizing, calling teams | One-off research by technical teams | A steady pipeline of companies with timing |
What should you ask before buying an AI lead generation tool?
Demos of lead generation tools tend to look alike: a search box, a list, a confident score. These questions get past the demo. They apply to any vendor, including us.
- Where exactly do the leads come from? "Proprietary data" is not an answer. Ask for the sources, and whether they are read through official APIs or scraped against a site's terms.
- How old is a lead when I see it? A database record can be months old, while a signal only helps while the need is open. Ask to see the date of the event behind each lead.
- Can I see why each lead was chosen? If the tool cannot show the criteria it applied, you cannot correct the targeting when it drifts.
- What does "verified" mean? Ask how emails are checked, what happens to addresses that fail, and whether the tool ever sends to a guessed address.
- What counts as a lead in the price? A row, a revealed email, a credit, or a company that passed qualification? The unit decides your real cost.
- Who sends, and from which domain? Outreach from your main domain puts your everyday email at risk. Ask about separate domains, warm-up, daily caps and what happens when bounces rise. Google's sender guidelines show what inbox providers expect of bulk senders.
- What happens when someone says no? An unsubscribe should apply to every campaign, not just the one that received it.
- Who checks the AI? Ask whether you can approve emails before they go out, and whether a person reviews the targeting in the early weeks.
How do you know AI lead generation is paying off?
Judge the leads first, then the replies, then the meetings. Five numbers cover most of it, and each one comes from your own campaign data rather than a vendor dashboard:
| Metric | How to compute it | What it tells you |
|---|---|---|
| Lead acceptance rate | Leads you would keep ÷ leads you reviewed | Whether the ICP and scoring match your own judgment |
| Verified contact rate | Leads with a verified decision-maker email ÷ qualified companies | Whether your buyers can be reached by email at all |
| Reply rate by signal type | Replies ÷ leads contacted, split by signal | Which signals deserve more volume |
| Positive reply rate | Positive and interested replies ÷ leads contacted | Whether the angle and the offer land |
| Cost per meeting | Total monthly cost ÷ meetings booked | Whether the channel pays for itself |
A worked cost-per-meeting example
Replace every assumption here with your own. Suppose you are on Growth at $499 a month, you spend four hours a week reviewing leads and approving emails, and you value that time at $75 an hour, or about $1,300 a month. Your total monthly cost is roughly $1,800. If that month produces six meetings, each one cost about $300. If one meeting in four becomes a customer worth $4,800 in its first year, each meeting is worth about $1,200 to you.
The six meetings are an assumption chosen to show the arithmetic, not a forecast. The point is to know your break-even before you start, so one slow week does not make you quit and one lucky week does not make you overspend. Notice too that your own time is the largest cost in the example, which is why approval mode should loosen once a funnel proves itself.
Mistakes that distort the numbers
- Judging a funnel on its first week, before the inboxes are warm and the ICP is tuned.
- Mixing two signal types in one campaign, so you cannot tell which one produced the replies.
- Counting opens. Mail apps increasingly load images automatically, so open rates say little about interest.
- Loosening the ICP to hit a volume target, which raises the lead count and lowers every number after it.
Who is it for, and how do you start?
Startories is built for B2B companies with a definable buyer and deals worth at least several hundred dollars: SaaS companies, IT service providers and MSPs, recruitment agencies, consultants and other B2B services. It is not built for consumer lead generation, and it is a weak fit if your sales depend on phone numbers for a calling team.
Write down your ideal customer profile first; the ICP template takes about an hour, and our guide on how to find B2B leads shows the manual version of this process if you want to see what is being automated. Then pick one signal that matches your offer, such as software recommendation requests if buyers ask peers for tools in your category, and start on Starter with a 3-day full-access trial for $1.
The leads flow straight into outreach, which the AI SDR handles; if you would rather have our team run it, see done-for-you lead generation. Compare the three plans on the pricing page.
Frequently asked questions
What is AI lead generation?
AI lead generation uses AI to find companies that may buy, decide which ones fit your ideal customer profile and enrich them with a decision-maker and a verified email. Startories starts from a buying signal, so each lead also comes with a reason to reach out now.
What does qualified and enriched mean?
The lead was matched to a real company, checked against your ICP and scored with written reasons. It also has an identified decision-maker, a verified business email and a link to the public signal that triggered it.
How many leads will I get each month?
Roughly 250 on Starter, 2,500 on Growth and 6,000 on Scale. Very narrow niches can produce fewer, because volume depends on how many companies in your market show a signal in a given month.
Do I still need a lead database or an email finder?
No. Signal detection, company matching, decision-maker search, email verification and outreach are included. A database can still make sense if you run account-based sales on a fixed list of named enterprise accounts.
Where do the leads come from?
From public sources: Reddit, X, Product Hunt, startup and industry directories, and search results built for your niche. Official APIs and compliant sources are used where available, and every lead keeps a link to the original signal.
Is AI lead generation the same as an AI SDR?
They overlap. AI lead generation produces the qualified, enriched lead; an AI SDR takes that lead and runs the outreach, follow-ups and replies. In Startories both are one pipeline, so a lead moves from qualification to a personalized first email without an export.