Clay: the best prospecting engine we have tested, and the bill that comes with it

Clay, an AI-powered prospecting tool, delivered strong results in a live test, correctly identifying CEOs for ten companies and finding valid work emails for nine of ten leads. However, its high cost and steep learning curve may limit its appeal to serious outbound teams, with paid plans starting at $167 per month and credits burning quickly on expensive enrichments.

The Proof · AI tool Clay: the best prospecting engine we have tested, and the bill that comes with it - Who it's for - Founders and marketers running real outbound volume who are comfortable with a spreadsheet and a credit meter. If you only need a few leads a month, Apollo or Instantly is cheaper and simpler. - Real cost - Free tier: 2,000 credits a month, with phone enrichment blocked. Paid plans were restructured in March 2026 to roughly $167 to $185 a month Launch and $446 to $495 a month Growth , plus custom Enterprise. Credits are metered from about $0.05 each, and the expensive enrichments burn fast: a mobile number runs about 11 credits a row. We found 50 real leads, ran the six-provider email waterfall, and drove Clay's AI agent across ten companies on the live tool. The AI is the real thing and the coverage is high. The catch is the price and the learning curve. What's good - Claygent is the standout. Asked for the current CEO of ten companies on a live test, it went ten for ten, handled the co-CEO cases, and for Activision no CEO since the Microsoft deal it returned "Unknown" rather than inventing a name, showing its reasoning, its confidence, and the exact pages it visited. - The email waterfall is the real product. Clay chained six providers per row and found a valid work email for nine of ten real SMB leads, falling through to the fifth provider on one and stopping the moment it had a match so it did not pay every provider. - The AI lowers the floor. Sculptor turned a plain-English target into nine structured filters over a 338-million-person database, and the column builder wrote the prompt and picked the model for us. - Finding and importing leads is free. You only spend credits on enrichment, and every run quotes its credit cost before you commit. - It feeds your whole stack. Clay exports to CSV, CRMs, and sequencers, runs email campaigns natively, and exposes its enrichments to ChatGPT and Claude over MCP. Where it breaks - The price is real. Paid plans run about $167 to $495 a month after the March 2026 restructure, and the good enrichments burn credits fast: a mobile number is a fourteen-provider waterfall at about 11 credits a row, so phoning 50 leads is roughly 570 credits. - The Free tier is a demo, not a workhorse. It gives 2,000 credits a month and blocks phone enrichment outright, with a flat "your subscription does not allow this." - The learning curve is steep. The surface area is enormous tables, waterfalls, signals, functions, sandbox, MCP , and we fumbled the sandbox toggle ourselves. The most common complaint in reviews is the time it takes to get productive. - The marketing oversells coverage. Clay implies it can triple your match rate; independent tests put the real uplift closer to a doubling. - It is a power tool, not a magic button. It will not fix a bad list or a weak offer, and the openers it personalized for us were specific but templated, still needing a human edit before send. How we tested We logged into Clay on the Free tier and drove the real product, not the demo reel. We built a lead list from scratch, ran the email waterfall on real people, tried the phone waterfall, pushed Clay’s AI agent across ten companies, wrote personalized openers, and watched the credit meter the whole way. Everything below is what the tool returned, with the receipts. The whole exercise cost about 30 of the 2,000 monthly credits. Claygent is the standout Clay’s AI research agent, Claygent, is the thing that separates it from a database with a spreadsheet bolted on. We gave it a deliberately awkward job: find the current CEO of ten companies, names only. It went ten for ten. It got the easy ones Tim Cook, Mark Zuckerberg, Andy Jassy and, more tellingly, the hard ones: Waymo’s co-CEOs, and Atlassian’s Mike Cannon-Brookes, correct only if you know Scott Farquhar had stepped down. The receipt that mattered was Activision. There is no current CEO; Bobby Kotick left after the Microsoft acquisition. Claygent did not guess. It returned “Unknown,” with a written rationale “the current leadership team identifies Rob Kostich as President, but does not list a Chief Executive Officer” , a confidence rating, and the two pages it had actually visited the company’s About page and the Activision Blizzard newsroom . For a tool whose entire job is to fill cells, refusing to fill one when the answer is “no data” is the behaviour you want and rarely get. It cost about two credits a row, and Clay’s own AI wrote the structured prompt and picked the model clay-neon without us touching either. The “without us touching either” is worth dwelling on. We typed a plain sentence; Clay’s own AI classified it as a web-research task, selected its mid-tier model clay-neon over the cheaper clay-helium and pricier clay-argon , and wrote a structured prompt that included “do not infer or guess; extract only what is explicitly stated.” That is the thing that flattens the learning curve: you describe the outcome, and the tool configures the agent. Finding the leads is the easy part We typed a plain-English target into Clay’s lead finder, Sculptor: “heads of marketing at B2B SaaS companies, 11 to 200 staff, in the US.” It built nine structured filters from that sentence, queried a database it reports as 338 million people, and returned 3,063 matches, real working marketers at real small SaaS companies, not the famous names a demo would show. We capped the import at 50 two per company, for variety and pulled them into a table. That import was free. In Clay, finding and loading leads costs nothing; the meter only starts on enrichment. The waterfall is the real product The headline feature is “waterfall enrichment,” and it earns the name. We added a Work Email column, and Clay expanded it into a visible chain of providers: it tries the first, validates the result, and only if that comes back empty does it spend a credit on the next, all the way down. On our 50 real SMB leads it found a valid work email for nine of every ten, and we watched it work: for one lead the first four finders returned nothing and the fifth, Findymail, landed the address. Downstream columns read “Run condition not met,” which is the waterfall doing exactly what it claims, stopping the moment it has a match so it does not pay every provider on every row. It ran about a credit a row, and the run dialog quoted the cost “run 49 rows, 24.5 credits” before we committed to it. That is the genuine advantage over a single-source tool like Apollo: instead of one provider’s coverage, you get the union of many, in priority order, billed only for the ones that run. One flag on that nine-of-ten, because it matters: these were marketers at real but findable SaaS companies, the friendly end of the distribution. Coverage falls on smaller, older, or non-US firms, and on people who keep a thin web footprint. Treat 90% as the top of the range, not the average. The independent test we trust, chaining four providers across 200 harder prospects, landed between 40% and 78% https://hackceleration.com/clay-review . The waterfall genuinely beats a single source; it does not repeal the limits of the underlying data. Personalization works, with an asterisk We asked Clay’s AI to write a one-line cold-email opener for each lead from their title, company, and LinkedIn headline. It auto-routed the job to a cheaper model than the research agent one credit a row instead of two and produced openers that were genuinely specific: “As VP Marketing at 1mind, your AI-native GTM focus and GTM AI Academy work stood out to me.” The asterisk is that four of five opened with the same “As title at company ” scaffold. This is real personalization at scale, but it is a draft, not a send. The tool will pull the right detail; it will not give you a voice. Then you act on it A list you cannot use is just a bill, and this is where Clay stops being a spreadsheet. From the same table you can download a CSV, push to a CRM, send to a sequencer like Smartlead or Instantly, or run an email campaign inside Clay itself. It even exposes its enrichments to ChatGPT and Claude over an MCP server, so you can ask an assistant to enrich a lead and have Clay answer. The other half is Signals: turn any enrichment into a trigger and Clay watches for the moment to reach out, a job change, a new hire, a promotion, a funding round. That is the difference between a static list and timed outbound, and it is the part most single-source tools simply do not have. The money is the catch Clay’s pricing is a metered credit system, and it is where people get surprised. Two facts set the frame. First, the good data is the expensive data: a mobile-phone lookup is a fourteen-provider waterfall at about 11 credits a row, roughly ten times the cost of an email, so phoning the same 50 leads would run about 570 credits. Second, the Free tier is a demo: 2,000 credits a month, and phone enrichment is blocked entirely. When we tried to run it, Clay refused with a flat “your subscription does not allow this integration to be added.” Email and AI enrichment work on Free; phones and the heavier providers want a paid plan. The paid plans are not cheap. Clay restructured its pricing https://www.clay.com/pricing in March 2026 into Launch about $167 to $185 a month and Growth about $446 to $495 a month , plus custom Enterprise. Credits meter from roughly $0.05 each. Used well, on a tight ICP with the right waterfalls, that is a fair price for the coverage. Used carelessly, with broad lists and the expensive enrichments switched on, it is how a $349 plan becomes a $349-plus-overage plan. The real cost per lead Clay never quotes you a price per lead, so here is the math from our runs, at Clay’s own credit prices data credits start around five cents each . A verified work email ran about a credit, call it a nickel. A Claygent research column, like our CEO lookup, ran about two credits, a dime. A mobile number is the outlier at roughly eleven credits, north of fifty cents, and it is paid-tier only. Now put it on a real list. Enriching a thousand leads with a work email and one AI research column is around three thousand credits, on the order of $150 of credit value, before you have written a word of outreach, and that sits on top of the monthly plan. Add phone numbers and you can double or triple it. That is the number nobody puts in front of you, and it is the whole reason the meter, not the sticker price, is the thing to watch. The flip side is the good news: an email at a nickel, found by the union of six providers, is genuinely cheap per unit. Clay does not get expensive because any one lookup is dear. It gets expensive when you run a lot of them, or reach for the dear ones, without watching the total. Where it bites The learning curve is the real tax. Clay’s surface area is enormous: tables, waterfalls, signals job changes, new hires, funding, hiring , functions, a sandbox mode, and an MCP server that pipes its data into ChatGPT and Claude. The power is the point, but it is a lot, and we fumbled the sandbox toggle ourselves on the first try, which is the single most common complaint in reviews. Plan a week of real use before it pays you back. Two more cautions. Clay’s marketing implies it can triple your match rate; the independent test we trust put the real uplift closer to a doubling, which is still good but not the headline. And, like every tool in this category, it will not fix a bad list or a weak offer. It is a faster way to be right or wrong; the strategy is still on you. Clay, or something cheaper? Clay is not the only way to find an email, and most readers should not start here. The real question is whether the waterfall and Claygent earn the premium over a flat-rate tool. If you need a few hundred leads a month for a tight, well-known ICP, a single-source tool is cheaper and far simpler. Apollo bundles a contact database and a sequencer for tens of dollars a seat, not hundreds; Instantly and Smartlead handle sending and basic enrichment in the same range. None of them match Clay’s coverage, but for an obvious ICP, one provider often gets you most of the emails for a fraction of the price and none of the learning curve. You move to Clay when you hit the ceiling of that: when single-source coverage leaves too many rows blank, when you need research or scoring per lead that no database ships with, or when you want signals-triggered outbound at volume. That is the job the waterfall and Claygent are built for, and nothing flat-rate competes. The order is the whole point: buy the cheap tool first, and graduate to Clay when you have a reason, not a hunch. The verdict Clay is the most capable prospecting and enrichment tool we have driven, full stop. The AI agent is accurate and refuses to hallucinate, the email waterfall genuinely beats single-source coverage, and the whole thing is wired to act on what it finds. It earns the verdict. The scope on that verdict is tight, and it is the whole point. Clay is worth it for someone running real outbound volume who will learn the tool and watch the meter. For that person it is a clear yes. For the solo builder who needs a few dozen leads a month, it is overkill and overpriced, and a $30 Apollo or Instantly plan does the job with a fraction of the learning curve. Buy Clay for the volume, not the novelty, and keep one eye on the credits. Sources Every outside quote in this review was re-fetched from its source before we used it. | Source | Link | |---|---| | Clay, Waterfall enrichment | |