Prospi AI

The model does real work in four places: a research agent reads each prospect's website and public activity to write a unique opening line, the onboarding flow reads the customer's own site to build an ideal customer profile and draft sequences, inbound replies are sorted automatically into interested, not interested, meeting ready and out of office, and responses are suggested for one tap sending.

Apply the removal test and a substantial product still stands, because the load bearing assets are not models. Two contact databases totalling more than five hundred million records, address verification, mailbox provisioning, warmup, inbox rotation, sequencing and a unified inbox all survive intact. What goes is the copy assistance and the sorting. Sold as an artificial intelligence platform, structured as a database and a sending engine with assistance layered on.

The automation is extensive and runs unattended by design: campaigns send on schedule across rotated mailboxes, a model writes each opening line with no review step described, inbound mail is categorised automatically into four states, and replies are suggested for dispatch in one tap from a phone. Against that, no approval queue, review gate before a generated line goes out, customer set sending ceiling, role definition, permission set or audit record appears on either page read.

One genuine control is named rather than gestured at, and it is included on every plan rather than gated: do not contact list management, which is a suppression mechanism rather than a posture. That single feature is the reason this row is not lower.

Five distinct model capabilities are marketed and no provider, model family, version or hosting arrangement is named for any of them. Two statements go further than most vendors manage, and both concern training data rather than architecture. The research agent's inputs are identified explicitly, being the prospect's own website and their public activity, so a buyer knows what the opener was written from.

And the reply suggestion feature is stated to be trained on more than five hundred agencies, which names a corpus even though it names no model. Identifying what a system learned from is more useful to a buyer than a provider name, but neither statement is accompanied by anything about the model itself.

Every quantified figure on this site sits inside a fabricated interface, and the fabrication is provable from the page source rather than inferred. The lead database screen displays six prospects with names, employers, addresses, revenue figures, headcounts and buying signals such as a recent funding round, three open sales roles and a named competitor in use, presented as the output of a verified database, and the photographs attached to all six are served from a public random face generator.

The campaign statistics, the warmup scores, the inbox placement panel and the reply thread are constructed the same way. Around them, eight customer logos support a claim of more than three hundred business customers and every file is named logo one through logo eight with alternative text reading client one through client eight, so not one is identified.

Exactly one testimonial appears, carrying a generated avatar and attributed to a co founder whose employer's name places them in the inbox provisioning trade rather than among ordinary buyers. A case studies page exists in the footer and was not opened. The vendor also embeds a widget on its own homepage directing readers to an independent review platform where the record is contested by unverified individual accounts, one of which alleges the platform sent to contacts who had opted out.

Two things sit above the category floor. Do not contact list management ships as an included feature on every plan rather than as an upgrade, and authentication records across all three types are configured automatically for every mailbox, which is a genuine hygiene control even though it is sold as deliverability.

Against that, no statute, regime or regulator is named anywhere on the homepage or the pricing page, no consent standard is stated for a database of more than five hundred million contacts sold expressly for cold outreach, and no unsubscribe or complaint mechanic is described.

One marketing line sharpens the whole row: the product offers to pull verified personal email addresses of company owners, and a personal address carries a materially different consent analysis from a work address in every regime that governs this activity. Terms of service and privacy policy exist at their own routes and neither was opened, both flagged.

A privacy policy and terms of service exist at their own routes and neither was opened, so this row rests on the surrounding surface and is flagged for re verification. What the surface establishes is the size of the question rather than any answer: two databases totalling more than five hundred million contact records, fifty or more data points held per record, intent driven buying signals attached to individuals, and personal email addresses marketed as part of the offer.

No lawful basis, notification route, subject request route or processing addendum appears on any page read. One documented inconsistency belongs here: the store listing for the mobile application declares that no data is collected and none is shared with third parties, which cannot be reconciled with an application whose stated purpose is surfacing the customer's inbound replies and contact records on a phone.

Two databases are sold, three hundred and twenty five million general business records and one hundred and eighty seven million ecommerce records, and nothing anywhere identifies where a single record came from. No source, supplier, partner, licence, collection method, refresh cadence, jurisdiction, notification route or subject request path is published on any page read.

The specific offer sharpens it beyond an ordinary absence: the product markets verified personal email addresses of company owners, and personal addresses are not compiled from business directories. Two further defects sit in the vendor's own numbers. Email accuracy is stated to one decimal place at ninety eight point two percent with no sample, method, date or measurement basis behind it.

And the core verification claim escalates between pages, appearing as double verified on the pricing page where a credit is defined and as triple verified on the homepage where the database is sold, so the buyer cannot tell what they are being charged for. Intent driven buying signals are attached to records with no explanation of how the intent was observed.

Inbox rotation is a named feature and the vendor also sells the mailboxes being rotated through, at three dollars fifty each inside the application, which is the pattern this axis exists to watch. Three things hold it at the middle rather than lower. The mailboxes are provisioned on a mainstream business mail platform rather than rented, shared or disposable. Authentication records are configured properly rather than circumvented, which is the opposite of an evasion posture.

And warmup is published as a gradual ramp rather than marketed as a way past sending limits. Nothing rents an identity, ships a scraping extension or touches a professional network. The exposure that goes unaddressed is upstream rather than in the sending: hundreds of millions of records including personal addresses were assembled from sources that are never identified, and whatever terms governed that collection are discussed nowhere.

The cross client question is answered here in the affirmative, volunteered as a selling point rather than extracted from a policy. The pricing page lists suggested replies to interested leads and states in the same line that the feature is trained on more than five hundred agencies.

Material generated inside customers' accounts, which in a reply suggestion feature means their actual correspondence with their actual prospects, has trained a model served to every other customer on the platform. Most vendors in this index avoid the question entirely or answer half of it; this one advertises the answer as a benefit.

What is absent is everything that would make it acceptable: no statement of what was used, whether any customer consented, whether an opt out exists, whether the material was anonymised or to what standard, or whether the agencies whose replies trained the model were told. Set against a platform that also holds mailbox credentials for unlimited connected accounts, the full inbound and outbound record of every campaign, and call activity through in application calling.

Every opening line reaching a prospect is written by a model from material scraped off that prospect's own website and public activity, and arrives under a named human sender from a mailbox bought for the purpose, with nothing marking it as machine written and no position taken on the European marking obligation.

The recipient is also the subject of research they did not agree to: the buying signals the product surfaces include a recent funding round, current hiring activity and which competing tools a company runs. Graded at the middle rather than lower because nothing in the marketing celebrates the recipient being unable to tell, inbox rotation is presented as throughput rather than as evasion, and the vendor's stated measure of success is positive replies rather than volume, which at least aligns its incentive with relevance.

Real breadth for a vendor this size: unlimited use of two named automation platforms, a messaging platform integration, reply and positive reply webhooks published as a feature rather than mentioned, all email providers claimed, in application calling, native applications on both mobile platforms, domain purchase inside the product and built in calendar scheduling. What is missing is the half a buyer would build against.

No documented interface, developer documentation, reference, authentication description, marketplace, partner directory listing or protocol server appears anywhere, and webhooks published without a documented interface behind them is an unusual combination. The sharpest omission is structural: not one system of record is named as an integration, at a platform that sits upstream of one by design and sells to teams that already run one.

Provider, country, region, data centre and residency option are absent from both pages read, and there is no security page or trust centre where such a statement would sit. Two published facts locate the company rather than the data. The mobile store listing gives a Wyoming street address of a kind used at scale by registered agent services, and a Serbian telephone number, while the founder's public profiles place the operation in that country.

For a platform holding European contact records, selling unlimited outreach into European recipients, and offering a database that includes personal addresses, no transfer position is stated anywhere.

This is an aggravated absence rather than an ordinary one, and the aggravation comes from what the platform holds rather than from anything it claims. It holds mailbox credentials for an unlimited number of connected accounts, it provisions and administers the mailboxes it sells, it holds the complete inbound and outbound correspondence of every campaign run through it, it holds call activity from in application calling, and it holds two contact databases totalling more than five hundred million records.

Against all of that, no certification, audit report, attestation, penetration test, control set, encryption statement, authentication requirement, trust centre or vulnerability disclosure route appears on any page read. The only security adjacent element on the entire site is a footer line reading all systems operational, which reports availability.

A buyer can compute their exact bill before speaking to anyone, which is what this axis asks. The unit is published at ten cents per verified contact with a base of one hundred and ninety seven dollars a month for two thousand, a volume slider running to five, ten and fifteen thousand with the discount at each step stated as twenty, thirty and forty percent, and an annual discount stated at twenty five percent.

The charging rule is defined rather than implied: a credit is consumed only when an address verifies and a failed lookup is never billed. What is included is enumerated across four sections and covers sending, warmup, personalisation, campaigns, testing, sequences, inbox and accounts, all stated as unlimited, with no per seat fee anywhere. The one meaningful add on carries a price, mailboxes at three dollars fifty each.

Payment methods are named, invoicing for larger buyers is stated, plan changes are described with their timing, and a free start is offered before any credit is spent. One inconsistency is flagged rather than deducted: a second pricing page exists showing a talk to sales step against some tiers, which the primary page does not have.

Contacts, campaign history, reply threads and the inbox record have no stated export function, format, post termination right, deletion timeline or retention period on either page read. Plan changes are addressed, with upgrades and downgrades allowed at any time effective from the next cycle, but billing flexibility is not portability.

The sharper question is specific to this vendor's model and is answered nowhere: the mailboxes and domains bought inside the application at three dollars fifty an account are the customer's sending identity and the accumulated sending reputation attached to it, and whether those domains and accounts transfer to the customer on departure or lapse with the subscription is stated on no page. A customer who leaves may be leaving behind the asset that took twenty eight days of warmup to build. Terms of service unread and flagged.

This is the strongest technical disclosure on the site and the reason the row sits here. A twenty eight day warmup ramp is published as a day by day volume curve rising from three sends to fifty, per mailbox warmup scores are shown against it with a readiness state for each, all three authentication record types are configured automatically and monitored, deliverability monitoring runs automatically, inbox rotation spreads volume across accounts, and the mailboxes come from a mainstream business mail platform rather than a disposable pool.

Publishing the ramp as numbers rather than asserting that warmup happens is what separates this from the category. Holding it off the top band, the headline figure is an absolute stated twice as a maintained rate, ninety six percent or better inbox placement, when no sender controls the receiving infrastructure and placement is not a thing a vendor can maintain. Complaint thresholds, bounce thresholds, per day ceilings on live campaign sending as distinct from warmup, and the response when placement degrades are all unstated.

Six audience pages exist covering business companies, agencies, sales teams, consultants, software companies and enterprise, and a whitelabel offering with unlimited workspaces and no per seat fee describes the agency buyer precisely rather than generically. A stated three hundred or more business customers supplies a scale figure.

Against that, not one customer is named in any sector, no region, country, language or industry coverage appears anywhere, no seat or headcount band is given, and there is no statement of who the product is wrong for at a vendor whose whole proposition is unlimited volume. The enterprise page sits awkwardly beside a platform with no certification, no residency answer, no federated sign on and no named security control.