Humanlinker

A genuine middle case, and the same shape as the other two vendors sitting here. Strip the models out and a working product survives: a contact base with recommendations, a multichannel sequencer across mail and the professional network, a unified inbox, enrichment, warm up and two directional record synchronisation. That is a saleable tool and several competitors sell exactly it.

But what survives is precisely what this vendor's own comparison pages argue is insufficient, since it publishes head to head pages against five rivals and each contest turns on the intelligence layer rather than the sending layer.

The packaging supports the same reading from both directions: credits meter only the generative work, with the vendor stating explicitly that navigation and data access consume none, so the mechanical product runs free of the metered resource while every differentiating capability draws it down.

The vendor offers a real choice and then describes no mechanism around it. Campaigns run in either a hybrid mode combining generated content with the seller's own inputs or a fully generated mode, and both are named as distinct product capabilities, which is more honest than platforms that blur the two.

Genuine containment controls exist and are enumerated: volume and sending limit settings, a domain suppression list, campaign status management, and administrative visibility through shared inbox access and a team performance dashboard. What is absent is any account of what happens before a fully generated message reaches a real person.

No approval step, no review queue, no confidence threshold on the behavioural inference that shapes the message, no statement of what the system may not generate, and no action level audit trail were located. One capability sharpens the question: an administrator can send campaigns on behalf of a team member, so generated messages can go out under an employee's name without that employee composing them.

The disclosure that counts here is architectural and it is unusual: at the top tier the customer chooses the model. The published feature list offers deployment of dedicated large language models, selection and configuration of the chosen model including fine tuning, integration of the customer's own records and sales playbook into that model, and collaboration with a prompt engineer, with provider logos displayed against the line, one of which is identifiable as a European model developer.

A model context protocol server means an external assistant can drive the platform, so for that path the buyer knows exactly which model is working because they supply it. Elsewhere the credit documentation describes the mechanism in outline, explaining that prospect data is combined with the customer's value proposition and rules to guide the models.

Held off the top band because the providers appear as images rather than named in text, no model or version is identified for the default configuration every customer below the top tier actually uses, no inference location is stated, and nothing describes what is sent to which provider.

The evidence base is broad and unusually well attributed for a company at this stage. Roughly twenty testimonials each carry a full name, a job title and an employer, spanning a cloud provider's chief sales officer, agency heads, chief executives and individual representatives, and most contain a specific figure such as a reply rate quadrupling to above twenty eight percent, a sales cycle nearly halved, a six and a half percent reply rate after four months, or an appointment rate up fifty percent.

The customer wall carries recognisable European names including a major cloud provider, a leisure group, a consultancy, a national news title, a telecommunications equipment maker and an agency inside a global advertising group.

Independent recognition is real and checkable: a national business magazine feature, selection as a winner of a major software company's generative artificial intelligence programme, a national television segment, and inclusion in a consultancy's published radar of the country's generative artificial intelligence startups.

Held off the top band because none of the four headline aggregate claims, covering reply rate, hours saved, opportunities and closed deals, carries a method, period, baseline or sample, and because media features and programme awards are recognition rather than independent evaluation of the product.

Two real instruments ship on every tier and no regulation is named anywhere. Unsubscribe links are included from the trial upward, and a domain suppression list lets a customer exclude named organisations before a campaign runs. The enrichment partnership carries a stated compliance claim for phone number acquisition.

Against that, the product surface names no sending regulation at all: no lawful basis for unsolicited commercial mail to individuals in the vendor's own jurisdiction, no consent standard, no position on the European transparency obligations that took effect in August 2026 despite generated content being the core proposition, and no telephone rules for the call steps and call task manager the platform includes.

The gap is more conspicuous here than for most vendors in this index precisely because the company is established in Europe, displays a compliance badge in its own footer, and sells primarily into European markets where the rules it declines to name are the ones that apply to its customers.

The structure is properly built for a French company and the specific question this product raises is unaddressed on the pages read. Four separate documents are published covering the statutory legal notice, a privacy policy, a cookies policy and user terms, and the enrichment partnership carries an explicit claim of compliance with the European data protection regulation through a secure and transparent process. None of the document contents were read on this pass.

What needs checking is narrower and sharper than the usual list: this is a controller established in the jurisdiction that wrote the regulation, inferring behavioural profiles of identified individuals from data those individuals did not provide to it, which engages the profiling definition, the notification duty owed when data is obtained from other sources, and the right of access to the resulting inference.

No notification route, access mechanism or objection path for a profiled individual was located from the pages read. Re verify at the privacy policy specifically for how the personality analysis is characterised and what rights attach to it.

One supplier is named and the rest is not. The enrichment partner is identified by name across the pricing page, the credit packs and the questions section, with email verification described as triple validated and phone enrichment sold as a separately priced option through that partner with a compliance claim attached. Naming the supplier puts this ahead of the many vendors here that describe enrichment as coming from various sources. Everything else is absent.

The recommendation base is stated at seven hundred million contacts with no source, no lawful basis for holding records on that many individuals, no accuracy or match rate, and no notification or removal route for any of them. The inferred behavioural profiles are themselves derived personal data generated by the vendor rather than collected, and nothing published addresses where the inputs come from, how long a profile persists, or whether it is retained after the prospect is no longer being pursued.

This is the broadest automated professional network action set graded in the index and the least governed. Campaign step types cover connection invitations, direct messages, sponsored messages and public comments, all sequenced and sent automatically, with sponsored message sending marked as a new capability. A browser extension imports contacts and lists from the platform in bulk. Social accounts are warmed up by the vendor, which is a practice aimed squarely at avoiding restriction.

Against that, nothing is published: no daily ceiling, no pacing rule, no compliance queue, no activity log, and no statement anywhere that automating these actions carries any risk to the buyer's account. Two vendors elsewhere in this index automate less and publish more, one naming a numeric daily limit with an auditable log and another stating the restriction risk in its own terms.

Held above the bottom band because there is no evasion marketing, no rotation of multiple social accounts sold as a scaling mechanism since the entry tier connects one, and because the mail side is clean, running through delegated authorisation with the major providers or standard protocols.

The top tier offers something the lower tiers do not, and the gap between them is the finding. Enterprise customers can have dedicated models deployed, select and fine tune the model, and have their own records and sales playbook integrated into it, which implies a degree of isolation and is the strongest stewardship position on offer. Nothing equivalent is stated for the tiers every other customer buys.

No statement covers whether customer material, prospect records or generated content train any model, whether processing crosses tenants, what the providers retain, or how long anything is held. The corpus at stake is not trivial: connected mailbox content, professional network conversations, records synchronised from the customer's system of record including notes and meeting reports at the upper tiers, and inferred behavioural profiles of named third parties. A single published sentence on the training and tenancy question would move this row.

This is a new shape for the index and it should be read as the strongest candidate yet for tightening this band. The product infers a four factor behavioural profile of a named individual from their public professional activity, without that person's knowledge, and returns guidance on how to approach them, including what to do and what to avoid.

The vendor's own homepage sells the resulting asymmetry as the benefit, promising the buyer will understand their prospects better than those prospects understand themselves, and its blog demonstrates the capability by profiling a named technology chief executive. The subject is never told a profile exists, cannot see it, and cannot correct it, and inference means there is nothing to correct it against.

Alongside that sits automated public commentary: generating and sequencing comments on a prospect's posts is a shipped, unlimited campaign step on every tier, published under the operator's real name where that prospect's whole network reads it, and campaigns can run in a fully generated mode.

Held at this band rather than lower, on the same reasoning that held two earlier vendors here: nothing manufactures an identity, there is no synthetic voice, invented persona or assumed location, messages come from the real seller, and undetectability is nowhere marketed as the feature.

Real counterweights exist and are published: unsubscribe links ship on every tier, a domain suppression list is included, out of office replies are detected and rescheduled, and contacts who have left an employer are flagged so they stop being pursued.

The connector set is narrow by count and deep by behaviour. Three major systems of record are supported with the depth gated by tier, and the integration is genuinely two directional: contacts and accounts are written back, messages sent are logged against the record, campaign contacts are added automatically with five configurable options, and deduplication runs against each system separately.

Both major mail and calendar providers connect, standard mail protocols are supported for anything else, a team messaging integration ships and another is announced, and a browser extension covers three browsers. A public interface is offered in a standard version on every tier and an advanced version higher up.

The agent era surface is the notable part for a company this size: a model context protocol server naming a specific assistant is included on every tier including the trial, has its own page, and was announced with a press release. Held off the top band because the catalogue proper is roughly ten integrations by the vendor's own count, with no marketplace, no partner directory and no third party application ecosystem.

Where the platform runs and where customer and prospect data are stored is not stated on the public pages. No region, data centre, cloud provider or residency commitment was located, and no residency option is offered at any tier. The company is French and says so prominently, which a European buyer may reasonably read as a residency signal, but corporate domicile is not a hosting commitment and the two should not be conflated.

The one adjacent capability is model rather than data isolation: enterprise customers can have dedicated models deployed, with one provider identifiable as European, which addresses where inference happens for that tier without addressing where anything is stored for any tier. For a vendor holding behavioural profiles of identified individuals across European markets, the storage question is the one a buyer's own regulator will ask first.

Two controls are named and both sit behind the highest tier, which is the pattern this index has flagged repeatedly. Federated sign on and multi factor authentication appear only on the enterprise plan, so a customer on the entry tier connects a mailbox, a social account and a system of record without either.

Beyond that no certification of any kind was located, with no service organisation report, no international standards certification, no trust centre, no penetration test statement, no encryption description, no status page and no vulnerability disclosure route found on the pages read.

Set against the access the platform holds, which spans delegated mailbox authorisation, a connected professional network session, write permissions on the customer's records and custody of inferred profiles of third parties, that is thin. A complete legal estate exists with four published documents, which keeps this above the bottom band and suggests the company is capable of publishing what it has.

The consumption economics are published with a precision very few vendors manage, which matters because credits rather than the seat price decide the bill. The entry tier is 69 euros per user per month, shown with the annual total excluding tax, a monthly against annual toggle carrying a stated twenty percent annual discount, a slider for one to five users, and a choice of three currencies.

Included allowances are stated per tier at 200 credits a year, 200 email enrichments a year, 800 contact recommendations a month and a named number of connected mail and social accounts. The credit unit is then defined by action: one credit produces one full outbound sequence, three credits produce one prepared meeting, and the vendor states plainly that navigation and data access consume nothing.

Top up packs carry published prices at 30 euros for 500 email enrichments and 47 euros for 100 phone numbers, with the enrichment partner named, and the rules governing them are set out, including that email credits remain usable for twelve months, that phone credits persist while the subscription does, and that credits pool across a shared workspace. The trial publishes its exact quotas rather than describing itself as limited. A four column comparison runs to well over a hundred rows.

Recorded as observed: the middle tier is described on the same page as suited to teams of more than five people and as starting from fifty users, so a team between those two numbers is told two different things and has no published price either way.

Export of contact lists is enumerated as a feature on every tier including the trial, which is a genuine capability rather than a promise, and cancellation terms are stated precisely, taking effect at the end of the current monthly or annual period.

Credits are addressed honestly in both directions: email credits remain usable for twelve months while the subscription is active, phone credits persist on the same condition, and both therefore lapse when it ends, which the vendor says rather than leaves to discovery. What is missing is the entire post termination half.

No retention period, no deletion timeline, no deletion confirmation, and no statement of what happens to campaign history, unified inbox conversations, enrichment already purchased, or the inferred behavioural profiles once an account closes. That last one matters more here than at most vendors, because those profiles concern people who were never customers of anyone involved.

A real named control set ships across every tier rather than being reserved for the top one. Warm up runs on both the mail and the social side, described as a pre heating system. Mailbox rotation spreads sending across two to four connected accounts depending on subscription. Volume and sending limit settings are exposed to the customer as configuration rather than being fixed by the vendor. A domain suppression list prevents contact with named organisations.

Unsubscribe links are included on every tier including the trial. Out of office replies are detected and the contact automatically rescheduled rather than burning a touch, and contacts who have left their employer are identified so sending to a dead address stops. Sending runs through the customer's own connected mailboxes, so the reputation built belongs to the buyer.

Held off the top band because no standard is named anywhere: no sender authentication protocols, no blocklist monitoring, no complaint or bounce threshold, no placement testing and no published ramp schedule for the warm up that is offered.

Two things here are rarer than they look. The vendor publishes actual headcount bands on dedicated pages, addressing small and medium businesses at fifty to two hundred and forty nine employees and large enterprises at five thousand or more, where most vendors in this index publish no number at all.

And each tier states its system of record prerequisite explicitly, with the entry tier supporting one platform, the middle tier two and the top tier three, which is a hard qualifying condition a buyer can test themselves against before a call. Four buyer roles each carry their own page covering sales leadership, individual representatives, business owners and freelancers, and the tier descriptions name the team shape each suits.

Held off the top band because the middle tier is described on one page as suiting teams of more than five people and as starting from fifty users, which leaves the entire range between those numbers unaddressed, and because no ceiling is published anywhere.