Bardeen

Prospecting has always been four jobs pretending to be one. Find the companies. Find the people at them. Find a way to contact those people. Work out which of them are worth contacting at all.

Most teams solve these with four subscriptions — a scraper, an enricher such as Clay, a sender, a CRM — and a spreadsheet holding them together, and the spreadsheet is where the process actually breaks — stale, half-enriched, and owned by whoever built it.

Bardeen's proposition is to run all four in one place, starting from the browser you already have open. That is genuinely useful, and it carries an obligation the marketing does not linger on: automating web scraping and contact enrichment at scale makes you responsible for both. This review covers what it does, what it costs, and where that responsibility bites.

What Does Bardeen Automate?

• It scrapes any website — you build a scraper or use a prebuilt template, and it extracts structured data from pages as you browse.
• AI research and enrichment run on top: it researches leads with AI search and finds validated phone numbers and email addresses.
• Qualification is the differentiator — you describe your ideal lead profile in plain language and it scores what it found against that.
• Bardeen Studio is a no-code builder, so a salesperson can chain scraping, enrichment and export without engineering help.
• Compliance is unusually strong for this category: SOC 2 Type II, GDPR, and CASA Tier 2 and 3 certified.

Scraping The Web Is The Engine

Everything Bardeen does starts with extraction. It runs as a browser extension, so it can read pages you can read — including pages behind a login you are entitled to use, which is what separates it from a server-side scraper hitting a site anonymously.

You either build a custom scraper against a page's structure or use one of the premium templates for common sources. The extension approach is the reason a non-technical user can build one: you are pointing at things on a page you are already looking at, rather than writing selectors against HTML you have never seen.

Extracted data lands in CSV, Google Sheets, Airtable or Notion, or onward into HubSpot AI, with broader integrations into CRMs, communication platforms and project tools.

The practical limitation of browser-based scraping is throughput and fragility. It runs at the speed of a browser rather than a server farm, and a site redesign breaks a scraper built against the old layout. Both are acceptable trades for the accessibility you get, but they mean this is a tool for hundreds or thousands of records rather than millions, and one that needs occasional maintenance.

A scraper you can build by pointing at a page is a different product from one you write selectors for. The accessibility is the whole proposition.

Qualification Is Where The AI Earns Its Place

Bardeen lets you describe an ideal lead profile and then qualifies scraped leads against it. This is the step where a language model is genuinely the right tool rather than a fashionable one.

Unlike an outbound platform such as Artisan, filtering on structured fields — company size, industry code, location — has been possible forever and does not need AI. What was never possible was filtering on judgement: whether a company's own description of itself suggests they have the problem you solve, whether a job title implies the authority to buy, whether a recent announcement makes now the right moment.

Those are readable from text and not from a database field, which is exactly what a language model is good at.

The caution is that qualification is a judgement, and a confident wrong judgement wastes the most expensive resource in the process — a salesperson's attention. Sample the output before trusting it: take fifty qualified leads, review them by hand, and count how many you would actually have picked. That hit rate tells you how much to trust the score, and it is the only number that matters.

Is Automated Lead Research Legal?

This deserves a direct answer rather than a disclaimer, because a tool that makes scraping and enrichment effortless does not make them consequence-free, and the responsibility sits with the operator rather than the vendor.

Three separate questions are involved and they have different answers. Whether you may scrape a site is governed by its terms of service and, in some jurisdictions, by computer access law. Whether you may hold someone's business contact details is data protection law. Whether you may email or call them is marketing law, which differs sharply by country.

Under GDPR, business contact data is personal data. Legitimate interest can be a valid basis for B2B outreach, but it requires a documented assessment, a genuine relevance to the recipient's role, and an easy way to opt out — and it does not survive indiscriminate volume. In several jurisdictions unsolicited commercial email to individuals requires prior consent regardless of how the address was obtained.

Bardeen's own compliance posture is genuinely good: SOC 2 Type II, GDPR and CASA Tier 2 and 3 certification says the platform handles data properly. That is a statement about the tool, not about your campaign.

The practical guidance is short. Check the terms of any site you scrape at volume. Keep a record of where each contact came from, because a data subject can ask. Honour opt-outs immediately and permanently. And take advice before running high-volume outreach into the EU or the UK, where the rules are strictest and enforcement is real.

• Read the terms of service of any site you scrape systematically, not just occasionally.
• Record the source of every contact record — provenance is the first thing a regulator or a recipient asks about.
• Treat business contact details as personal data under GDPR, because they are.
• Make opting out trivial, and process it permanently across every list you hold.
• Get advice before high-volume outreach into the EU or UK; consent rules there are stricter than most US-built playbooks assume.

Where Credits Actually Go

Bardeen runs on credits, and the practical difficulty is that different actions consume very different amounts, which makes a monthly figure hard to predict before you have used it.

The rough hierarchy is worth internalising. Scraping a page is cheap. Running an AI research step against a lead is more expensive, because it involves a model call and often a web search. Contact enrichment — finding and validating a phone number or email — is usually the most expensive step, because the vendor is buying or verifying that data from somewhere.

So a workflow that scrapes a thousand companies costs far less than one that scrapes a thousand companies, researches each one and finds two contacts at each. The number of records is not what drives the bill; the number of enrichment steps per record is.

The way to control this is to filter before enriching rather than after. Scrape broadly, qualify on the cheap text you already have, and only spend enrichment credits on the leads that survived. Teams that enrich first and qualify second routinely spend several times more for the same outcome.

We do not publish amounts, since credit rates change. Run one realistic workflow end to end, note the credits consumed, and multiply by your actual monthly prospecting volume — that figure, not the tier label, is what this costs you.

A Test That Tells You Whether It Fits

Automation tools demo beautifully on a clean target site and struggle on the messy one you actually need. Test on yours.

The bar is step four. A qualification hit rate you would not accept from a junior researcher is not one you should feed to a sales team.

1. Pick the site you genuinely prospect from, not the one in the demo. Template coverage is only useful if it covers your source.
2. Build one scraper against fifty real records and check the extracted fields by hand. Field-level errors compound through every later step.
3. Write your ideal customer profile as you would explain it to a new hire, then run qualification against those fifty.
4. Review all fifty yourself and count how many you would actually have chosen. That hit rate is the whole product; below about seven in ten it is costing you attention rather than saving it.
5. Now enrich only the qualified subset, and note the credits that consumed versus enriching all fifty. The difference is your monthly saving from filtering first.
6. Export to the destination you really use and confirm the fields map cleanly, since a broken CRM import undoes the entire workflow.

Buying Bardeen: The Credit Question

Access is a credit-based subscription with tiers, and the credits are the meter that matters rather than seats or automation count.

That structure suits uneven usage, which prospecting genuinely is — a heavy month building a new list and a light one working it. It penalises teams who cannot predict their volume, because the tier has to be chosen before the pattern is known.

The named customer list is a reasonable signal here: Deel, Kearney, JotForm and others are recognisable companies, and vendors do not usually name customers who would object to being named.

Two things to establish before committing. Whether unused credits roll over, since prospecting is bursty and a use-it-or-lose-it month is real waste. And whether enrichment credits are drawn from the same pool as automation credits, because if they are, a single enrichment-heavy campaign can consume a month's allowance.

We do not publish amounts. Model your cost on one real workflow rather than on the tier description.