Unify

Agent runs are a metered unit of account, which is the test this axis applies.

Credits are consumed by data actions and by agent activity alike, at published per action rates, so inference is a direct cost line rather than a bundled feature. That is the clearest evidence that the model layer is the product rather than an addition to it.

What those agents do: prompt driven research on accounts and people, qualification of whether a signal is worth acting on, and copywriting for the outreach that follows. An agentic chat surface lets an operator direct the platform conversationally. An observation model surfaces insights to representatives. Each has its own published page.

The organising concept is the play, a workflow triggered by a signal firing, where agents research and draft and a person reviews before anything sends. That structure only exists because the agents can do the intermediate work.

What remains deterministic is substantial and is graded elsewhere: aggregation across forty plus data vendors, waterfall enrichment, sequencing and the managed sending estate. A customer running only those would still have a platform, and it would not be the one the vendor markets.

Ask what a single agent run costs in credits and what it produces.

Human review sits inside the agent design, and an independent roundup describes the platform in exactly those terms.

The vendor's own agent page describes research, qualification and personalised outreach with human review. A competitor's comparison guide, which has no reason to be generous, characterises this platform as the signal to send layer with human approval and defines that control explicitly as a person reviewing and approving a drafted message before it is sent. Being described that way by a rival is stronger corroboration than a vendor claim.

The play structure is itself a constraint. An agent acts because a defined signal fired, not continuously, so the trigger conditions bound what runs and when. A team that has built ten plays has ten defined circumstances in which automation engages.

Credit metering functions as a spending ceiling: agent activity depletes a pool, so runaway behaviour is visible and finite rather than silent.

Write access to the record system is escalated deliberately, read only at the middle tier and read write only at the top, which is the right order.

What two passes could not locate: any audit trail, and any statement of whether the review step can be disabled.

Ask whether human review can be turned off and who can turn it off.

Agents are the product and nothing published says what runs them.

Two passes across the vendor's own content library, its agent and research pages as cited by third parties, and multiple independent reviews located no model provider, no model name, no version, no statement of what account or prospect data is transmitted for inference, no training or retention position and no accuracy figure.

The credit rates disclose cost rather than mechanism. A buyer learns that a multi step research run consumes more than a simple enrichment, which is useful for budgeting and says nothing about what performs the reasoning or how reliable it is.

Qualification is the capability where that gap matters most. An agent deciding whether a fired signal is worth acting on is deciding which prospects a team pursues and which are silently dropped, and no error rate is published in either direction.

One indirect signal exists and is worth recording without overreading it: the company's seed round included a major model provider among its investors, which suggests a supplier relationship without confirming one.

The copywriting agent drafts messages that reach real people after review.

Ask which provider runs the agents and what data leaves the platform.

Several case studies, each carrying multiple specific figures, at named and checkable addresses.

One describes a hundred and fourteen qualified opportunities a month, eighty percent less time spent prospecting, emails produced ten times faster, one point one million dollars in closed business and a one week ramp. Another reports a quarter of representative time saved at two hours a day, open rates between seventy and eighty percent, two point five nine million in pipeline and two hundred and fifty thousand in revenue. A third reports reply rates two and a half times higher with all outbound moved onto the platform. A fourth reports four point two times return, ten workflows built in two weeks and three hundred thousand in pipeline. A fifth describes consolidating three named competing tools into this one with workflows four times faster.

Multiple figures per customer, each at its own published address, is a different standard of evidence from a percentage on a home page.

The named customer set reinforces it, spanning several of the most recognisable software companies of the current cycle alongside established mid market names.

Company level disclosure is unusual too: revenue stated as having grown more than tenfold from a named month, and sixteen million dollars of total pipeline production attributed.

Ask which case study most resembles your own motion and team size.

The founding premise is a compliance argument dressed as a market thesis, and the mechanics behind it are undescribed.

The founders state the position plainly: outbound sales is dominated by cold mass outreach that floods people's inboxes and converts at a tiny rate, and the company exists to reach people at the moment they are actually looking for a solution. That is a philosophical commitment to contacting fewer people with better reason, and it is the correct instinct on this axis.

The managed sending estate reinforces it, since deliverability monitoring and rotation constrain volume rather than maximising it.

What two passes could not locate: any unsubscribe mechanism named in the product, any suppression or do not contact list, any consent basis for a contact dataset exceeding a billion records aggregated from more than forty vendors, and any handling of the person level website identification that determines who gets contacted.

That last item is the sharpest. Identifying a named individual who visited a website anonymously, then contacting them about it, is the practice this axis exists to examine, and the vendor's warm outbound framing does not address how that person came to be identifiable.

Ask how unsubscribes are honoured across managed mailboxes and plays.

Two access decisions are deliberate, and the holdings are among the largest in this index.

The deliberate part: record system access is read only at the middle tier and read write only at the top, which escalates the ability to alter a customer's system of record rather than granting it by default. Single sign on through two named identity providers appears at the top tier.

The holdings are the concern. More than a billion contact records and sixty five million company records aggregated from over forty vendors. Person level website intent, meaning identification of named individuals who visited a site. Champion tracking, which follows a specific known person across a change of employer. Product usage signals. Third party intent from named providers.

Champion tracking deserves separating out. Maintaining a persistent record of a named individual and detecting when they move jobs is longitudinal monitoring of a person across their career, and it is metered at one credit each, which tells a buyer the price and nothing about the basis.

Two passes located no processing agreement, subprocessor list, retention schedule or residency statement on the surfaces examined.

Ask what lawful basis covers champion tracking and person level identification.

The aggregation model is disclosed precisely, and not one supplier is named.

What is stated: more than a billion contacts and sixty five million companies, assembled from over forty vendors, queried in waterfall so that multiple providers are tried in sequence until a match is found. That is more architectural disclosure than most data platforms offer, and it explains the commercial value plainly, since a customer gets the coverage of forty contracts while holding one.

Credit rates make the economics visible per record: roughly one credit for an email address, four for a telephone number, with stated worst cases considerably higher.

What is absent is the identity of any single supplier, which matters because in a waterfall the customer cannot know which provider furnished a given record, and therefore cannot trace its origin or its consent basis.

One disclosure runs in the buyer's favour and is worth crediting. Third party intent licences from named providers are not included natively, and independent analysis states customers must hold their own. Being clear that a headline capability requires a separate contract is the honest position.

The person level identity graph behind website de anonymisation is undescribed.

Ask which providers sit in the waterfall and in what order.

Data arrives through commercial aggregation and mail leaves through a major provider's own infrastructure.

The data position is the significant one. More than forty vendors supply records under commercial arrangements, which is the licensed route rather than the extracted one, and it stands in contrast to several records in this index whose databases have no stated origin at all.

Sending runs on managed mailboxes provisioned on a major mail provider's platform, which is that provider's intended commercial use rather than a workaround. Record system synchronisation uses published interfaces with two named platforms.

Two passes located no scraping claim, no account renting, no undetectability marketing and no browser automation of a third party property.

Two questions remain open. Sequencing is described as multichannel with professional network steps among the actions, and nothing states whether those execute automatically or as manual tasks a representative completes, which is the difference between exposure and none. And person level website identification depends on an identity graph assembled from sources not described.

Ask whether professional network steps are automated or presented as manual tasks.

The safety control is real and the disclosure around it is not.

Human review before agent drafted outreach reaches a prospect is the substantive control, corroborated by a competitor's own comparison guide, and it is credited in full on the oversight axis. Credit metering adds a second bound, since agent activity depletes a finite pool and cannot run away silently.

What is missing on the model side: no provider, no evaluation, no accuracy figure and no constraint description for agents performing research, qualification and copywriting. The qualification agent in particular decides which signals warrant pursuit, and a false negative is invisible by definition.

What two passes could not locate on the security side: any dedicated security page, any certification, any audit report, any encryption statement, any incident response process, any breach notification commitment, any named security contact, any vulnerability disclosure route and any status page on the surfaces examined.

The custody is significant. This platform holds a contact dataset exceeding a billion records, person level website identification, longitudinal tracking of named individuals across employers, and operating access to customer record systems with write permission at the top tier.

Ask what independent security assessment exists and where it is published.

Real identity, drafted content, and a recipient who does not know how they were found.

Identity is genuine. Mail leaves from mailboxes provisioned for the customer under the customer's own domain and sender name. Two passes located no persona, no rented account, no synthetic sender, no cloned voice and no automated social action manufacturing the appearance of interest.

Human review before sending means a person read the message, which is a meaningful difference from the fully automated records elsewhere in this index. The recipient receives something a colleague approved.

Three things are not disclosed to that recipient and should be recorded. The copy was drafted by an agent and is presented as personally written. The contact may have been triggered by person level website identification, meaning they browsed anonymously and were named. And if they are a tracked champion, their move to a new employer was detected and priced at one credit.

None of those is deception about who is writing. All three concern how the recipient came to be selected, which is invisible to them.

The warm outbound premise means most recipients did engage in some way first.

Ask what a recipient would be told about why they were contacted.

Deep on data and identity, thinner on programmatic access than the positioning suggests.

The data ecosystem is the strength and it is genuine: more than forty vendors aggregated behind a single waterfall, named third party intent providers, and product usage signals. A customer buying this is buying access to a supply chain rather than a single source.

Record system synchronisation covers two named platforms, escalating from read only at the middle tier to read write at the top. Single sign on runs through two named identity providers at the top tier.

What two passes could not locate on the surfaces examined: any public interface documentation, any authentication scheme, any webhooks and any protocol server. That last absence is worth noting because ten vendors in this sweep ship a protocol server, several of them very much smaller, and a platform describing itself as a system of action for revenue teams would be an obvious candidate.

The vendor runs an unusually large published content library including direct comparisons against eight named competitors, which functions as an ecosystem asset in its own right.

Ask whether a documented interface and webhooks exist.

The sending infrastructure is described in operational detail and the platform infrastructure is not described at all.

On sending, the detail is specific: mailboxes provisioned and operated by the vendor on a major mail provider's platform, with dedicated high delivery addresses, domain and mailbox health monitoring, warm up, rotation across the pool and deliverability analytics. Mailboxes are priced individually per month, so a customer builds the estate deliberately and knows what each one costs.

That is more infrastructure disclosure than most records in this index provide about anything.

What two passes could not locate on the surfaces examined: any hosting provider or region for the platform itself, any data centre, any residency commitment, any tenancy or isolation model, any encryption statement, any backup or continuity position, any uptime commitment and any status page.

The residency question is live given the holdings. A contact dataset exceeding a billion records, person level website identification and longitudinal tracking of named individuals all imply processing of European personal data somewhere, and nothing states where.

Single sign on at the top tier runs through two named identity providers.

Ask where platform data is processed and whether a regional option exists.

Enterprise controls appear in the top tier feature list and no assurance material was located.

What is visible: single sign on through two named identity providers, reserved for the highest tier, alongside read write record system access and a dedicated consultant. Those are the controls an enterprise buyer expects and their presence indicates the vendor has met that requirement at least once.

What two passes could not locate on the surfaces examined: any certification, any audit report, any examination, any penetration test, any trust portal, any dedicated security page, any encryption or access control statement, any subprocessor list, any vulnerability disclosure route, any named security contact and any status page.

The customer base makes an absence unlikely rather than proven. Several named customers are software companies of a size and profile that would run a full supplier assurance review before granting write access to their record system, so material almost certainly exists that these surfaces did not surface.

A buyer should ask directly rather than infer from this record either way.

The holdings that would be assessed are substantial: over a billion contact records, person level identification and operating access to customer systems.

Ask for the security documentation and certification scope.

This vendor repriced by roughly an order of magnitude inside two months, and a buyer reading older comparisons will be badly misled.

The structure verified in July runs four plans: a free forever tier at no cost carrying a hundred credits per seat monthly for up to three seats; a base tier at twenty dollars per seat monthly with eight hundred credits; a professional tier at sixty dollars per seat with two thousand four hundred credits, read only record system synchronisation and a fourteen day trial; and a custom business tier billed annually with a workspace credit pool, read write synchronisation, managed mailboxes at twenty five dollars each monthly, single sign on and automated plays.

What that replaced is the finding. Through the first half of the year the entry point was an annual commitment of one thousand seven hundred and forty dollars monthly, roughly twenty thousand eight hundred and eighty dollars a year, covering one user and fifty thousand annual credits, with no free plan and no disclosed trial. Independent sources verified that figure in May and again in June, and an earlier structure at seven hundred dollars monthly is documented before it.

So the published floor moved from a twenty thousand dollar annual commitment to zero, and per seat pricing appeared where none existed.

Credit consumption is published per action type, which is the disclosure that makes a credit model usable.

Deductions: the top tier remains custom and annual, overage rates are reportedly undisclosed, and third party intent licences are not included.

Ask what credits cost once the included pool is exhausted.

Synchronisation carries the work outward and the sending estate does not travel.

Record system synchronisation with two named platforms is the route, and it escalates by tier: read only at the professional level, read write at the top. A customer on the upper tier has activity, contacts and outcomes flowing into a system they own as the work happens, which is the strongest form of portability and it operates continuously rather than at exit.

What two passes could not locate on the surfaces examined: any export mechanism described in product terms, any format, any public interface, any deletion commitment, any retention position after cancellation and any account closure process.

One asset cannot transfer by construction and it is the one the vendor is best known for. Managed mailboxes are provisioned and operated by the vendor, so the domain reputation, warm up history and inbox placement built over months belong to infrastructure the customer does not own. A departing customer starts that again elsewhere.

The play configurations, signal definitions and agent prompts a team develops are equally undescribed on exit.

Annual commitment on the top tier bounds when leaving is possible.

Ask whether managed mailbox domains transfer on cancellation.

An independent reviewer calls the deliverability infrastructure genuinely hard to replicate, and the component list supports that.

Mailboxes are provisioned and operated by the vendor on a major mail provider's platform rather than assembled by the customer. Around them run deliverability analytics, domain and mailbox health monitoring, warm up, rotation across the pool, and dedicated high delivery addresses. Each mailbox is priced individually per month, so the estate is built deliberately and its cost is legible.

Bundling the whole layer is the decision that matters. The common failure in this category is a customer assembling domains from one supplier, mailboxes from another, warm up from a third and authentication by hand, then discovering months later that placement collapsed. Operating all of it as one managed service removes that failure mode, and it is the same architectural choice made by the strongest sending records in this index.

The warm outbound premise reinforces it, since contacting fewer people at better moments generates fewer complaints than volume sending.

What two passes could not locate: any published warm up volume, any per mailbox daily ceiling and any inbox placement testing against named providers.

Ask what daily volume each managed mailbox supports and how warm up ramps.

The buyer is defined by motion, and an independent reviewer states the boundary more usefully than the vendor does.

The vendor's definition is teams running warm outbound off buying signals, and the named customer base bears it out, skewing heavily toward high growth software companies including several of the most recognisable of the current cycle.

The sharper framing comes from outside. An independent review observes that the signals are inbound shaped: the platform is strongest when an account is already engaging, through a website visit, an intent spike or a champion who moved, and that if a target list consists of enterprise accounts that have never touched the website, there is nothing for a workflow to trigger on. It adds that the intelligence layer is enrichment deep rather than account deep, telling a representative that a signal fired without building the context a complex deal requires.

That is a precise statement of who this does not suit, and it is more useful to a buyer than any positioning page.

Separate surfaces address revenue operations and individual representatives, so the product speaks to different roles.

The repricing suggests the segment is being deliberately widened downward toward smaller teams.

Ask whether the platform fits an outbound motion with no inbound traffic.