Before writing a single line of Rainmaker, we spent time inside the two most-hyped AI SDR tools in the market. Not as critics reviewing screenshots, but as users, logged in, trying to run an actual outbound campaign.
What we found changed what we built.
Why we started by tearing things apart
The AI SDR category moves fast and claims a lot. Every tool promises it will find your buyers, write the emails, and book the meetings. The marketing is almost identical across the space. The only way to know what is actually different is to use the products, stress-test the demos, and read what is happening underneath.
We picked the two tools that had the most attention: Gojiberry (YC S26, LinkedIn-signal AI SDR) and Explee, also known as AutoGTM (a London-based email-first agent with a large proprietary database).
We went in without a thesis. We wanted to see what they did well, where they fell short, and whether either was the tool we would actually use.
What we found inside Gojiberry
Gojiberry's strengths are real. It uses LinkedIn intent signals - job changes, funding announcements, hiring activity, competitor engagement - to identify companies that might be in-market. The onboarding is genuinely fast. The interface is clean.
But three things stopped us from trusting it with our company name on outbound.
Paywall before value. You cannot see a single lead until you enter a credit card. This is a product decision, not a pricing decision. It means you cannot evaluate whether the targeting is right for your ICP before you commit.
Autopilot by default. The default state of the product is autopilot - it finds leads, writes emails, and sends them. Your company's name goes on communications you did not review. The approval gate exists as an option, but the default is the real product decision. Most users never change defaults.
No citations. Every email Gojiberry writes contains factual claims about the prospect and their company.
None of those claims link to a source. If the agent gets a fact wrong, the email sends anyway.
We also noticed that the AI-score column for each lead was empty - no per-lead reasoning was shown. You see a number, not a rationale.
What we found inside Explee
Explee's strengths are also real. The database is large (105M companies, 536M people, per their published documentation). The natural-language ICP search is genuinely impressive. For a user who wants to find leads without an SDR doing research, the promise is real.
But two things we observed during testing gave us pause.
The demo breaks for anonymous users. Explee's hero experience - the thing designed to convert visitors into users - silently returns a 401 error for users who are not logged in. The demo appears to run. The results do not load. There is no error message. This is a significant trust signal about the reliability of the product and the attention paid to the conversion funnel.
The ICP engine returns competitors. When we ran our own company through Explee's ICP search, it returned a list that included our direct competitors as "ideal customers." It also returned a nuclear energy facility tagged as a real-estate company. Lookalike embeddings find companies that resemble your existing customers - they do not model who actually has the problem you solve.
Like Gojiberry, Explee sends on autopilot with no approval gate and no citations in the emails. The compliance page is empty.
The four product decisions we made in response
What we found did not make us pessimistic about the category. It made our product decisions obvious.
BuyerPersona, not lookalikes. We model who actually buys - their pains, their purchase triggers, the things that disqualify them. Every lead is scored against this model and the reasoning is shown for free, before you approve anything. We do not use embedding similarity to find companies that look like your customers.
SourcedClaim, or the email does not exist. Every factual sentence in a Rainmaker email links to a real, verifiable source. If the agent cannot find a source for a claim, the email is blocked before it reaches your approval queue. You cannot override this check. "No source, no send" is not a policy - it is a function.
ApprovalBatch as the default state. Rainmaker is a copilot. The agent drafts, you approve the batch, it sends. Autopilot requires an explicit opt-in. The default is approval-required because defaults are what most users ship with, and shipping unapproved outbound under your company name is a reputational risk that should require a deliberate choice.
Meeting as a first-class outcome object. A reply is not a meeting. A click is not a meeting. Rainmaker defines "qualified meeting" in writing: booked on calendar, RSVP-accepted by the prospect, not cancelled before the start time, and the start time has passed. This definition is the basis for the $25 outcome fee. We charge on the meeting because it is what you are actually buying.
What Rainmaker is, and what it deliberately isn't
We are not building the broadest AI SDR. We are building the one you can put your name on.
That means slower drafting than pure autopilot (you are in the loop). It means emails with citations (which take more to generate). It means an outcome fee that only triggers when a real meeting holds (which means we only win when you do).
If you want volume sends with no approval gate, Rainmaker is not the right tool. If you want outbound where you know why each lead was chosen, where every claim in every email is sourced, and where the pricing is aligned with whether it actually works - that is what we built.
You can paste your URL and see your first leads without signing up at the homepage demo.