Qualified

The removal test splits this product rather than settling it, and that split is what places it one band below the agent natives. Two capabilities cease to exist without models rather than degrading: the generated video agent that holds a face to face conversation with a visitor, whose avatar supplier the vendor names in its own subprocessor list, and the personalised email the agent composes and sends. Strip the models and neither becomes a lesser version of itself, it is absent.

What remains, however, is not nothing. The company shipped a conversational marketing platform from 2018, and visitor identification, segmentation, conversation routing, meeting scheduling and bidirectional record synchronisation all predate the agent and survive its removal as a complete and previously saleable product.

This is the Mindtickle shape rather than the incumbent shape, a genuine model dependency in a signature capability sitting on a conventional platform, which is why it sits above the C that every established platform in this index has taken and below the A reserved for companies with nothing underneath.

Marketing volume did not decide it: every product line here has been renamed around the agent and the vendor has renamed the category itself, and that was set aside in favour of what survives removal. Ask which capabilities stop working entirely without the model layer, and what the platform still does for a customer who switches the agent off.

An agent that converses, qualifies, books and sends without a person, and no published account of what constrains it. The vendor describes the agent operating around the clock, answering questions, qualifying leads, scheduling meetings and sending follow up email so that nothing waits for a human representative to come online.

Two mechanisms are named and both are workflow rather than oversight: an agent configuration studio included in every tier, and routing that hands a qualified conversation to the right representative. Passing a good lead to a person is a business process, not a control on a bad run. What a control would look like here is absent from every published surface.

Nothing states what the agent may commit to in a conversation with a buyer, whether generated email can be held for approval before it sends, what volume ceiling applies, what escalates automatically, or what halts a conversation or a nurture sequence already in flight. The gap weighs more for a conversational agent than for a sending agent, because this one answers product questions in real time to a prospect who may act on the answer. Ask what the agent is permitted to state about capability, availability and price, whether generated email can be gated for review, and what stops it mid conversation.

The model supply chain is disclosed in a legal instrument rather than on a marketing page, which is rare enough in this category to carry the grade. The subprocessor list, effective June 2026 and incorporated by the data processing addendum, names five generative suppliers individually with the location of each: two separate routes to one large model provider including a hyperscaler hosted deployment of it, a second hyperscaler managed service that names the model family it serves, a third cloud provider service, and a distinct supplier for the generated video avatar.

A buyer can therefore establish which companies process their visitors' conversations, evaluate each on its own terms, and receive notice when the list changes. No comparable record in this index publishes that. Missing is everything above the supply chain: no model family or version is stated for any capability, no model card exists, no evaluation, accuracy or error rate is published for qualification, conversation or generated email, and nothing states which supplier serves which capability or what happens on failover between them.

The qualification and routing logic appears to be operator configured rather than learned, but that is inferred from the product description rather than stated. Ask which supplier serves which capability, what evaluation covers agent output before it reaches a buyer, and what the fallback path is when a provider is unavailable.

Named customers with linked case studies, an independent analyst evaluation, and not one figure carrying a denominator. The vendor publishes five named customer outcomes, each linked to its own write up: a billing software company at six times revenue, a company data provider at three times meetings and twice the qualified leads, a payables company at four times engagement, a rights management company at four times meetings booked, and a healthcare software company at seven and a half million dollars of pipeline.

Only the last is an absolute figure. The rest are multiples with no baseline, no period and no method, and the vendor states that more than five hundred companies have hired the agent without publishing anything measured across that population.

The independent element is real and was checked rather than assumed: the vendor cites a leading position in a named analyst evaluation of conversation automation for business buyers from the fourth quarter of 2025, and the reprint is hosted on the analyst's own domain rather than on the vendor's blog, which is the host domain test this index applies to any comparative study a vendor cites.

Review aggregation is substantial at roughly fifteen hundred ratings averaging four point nine out of five. Ask for conversation volume, qualification accuracy and meeting conversion measured across a stated customer population and period, and for the baseline underneath any multiple.

The population contacted is bounded by architecture, and nothing is published about how that boundary is maintained. The agent engages people who arrived on the customer's own website or who already sit in the customer's lead database, and the vendor states its focus is inbound rather than outbound prospecting, so this product does not source cold contacts and send to them.

That is a materially better starting position than the outbound agents graded in this cohort, and it comes from the architecture rather than from a policy page, which is the strongest form of evidence this index recognises. Everything after that is unstated.

Nothing published describes the consent basis on which nurture email reaches a lead who has not replied, how suppression and unsubscribe requests are captured and honoured across chat, email and voice, whether an objection recorded in the connected record platform stops the agent, or which obligations the vendor considers to transfer to the customer. An acceptable use policy exists on the vendor's own domain and was not retrieved on this pass, so part of this may be answered there. Ask how unsubscribe and suppression propagate between the agent and the connected record platform, and what consent basis supports agent sent nurture email to a lead who has not engaged.

A full instrument set published and reachable, with two specifics holding it below the top band. Available without a sales conversation are a data processing addendum, a subprocessor list carrying an effective date, a privacy policy, a dedicated regulation page, and an enumeration of exactly what is collected about a website visitor, stated as the network address at minimum and in many cases the name and business email address, alongside a statement that payment card numbers, government identifiers and other highly sensitive personal data are not stored.

Cross border transfer is handled correctly and unusually openly: the vendor states that it was previously certified under the transfer framework that a European court invalidated in 2020, explains the decision that invalidated it, and states that transfers now run on standard contractual clauses covering the vendor and its subprocessors. This index treats continued reliance on that invalidated framework in 2026 as a live warning sign, and this vendor explicitly disclaims it.

Individual visitor deletion is offered with a stated timeline of one quarter to hard delete from production systems. Two things hold the grade. The visitor identification cookie is published with a ten year expiry, which is long against any purpose a sales conversation serves.

And the trust page states that customers may change their retention settings in the application while the pricing page lists custom cookie and data retention policies as a capability of the second tier, so the control the privacy surface describes as available is packaged as an upgrade. Ask which retention settings are available at the tier quoted, and what purpose a ten year visitor identifier serves.

Two enrichment suppliers named, a third layer unnamed, and a stated method that implies more than the list contains. The subprocessor list names two firmographic and enrichment providers, both separately indexed here, so a buyer can trace that part of the corpus to known vendors and evaluate them directly.

The architecture also reduces how much has to be sourced at all, since the vendor states that the visitor's email address is the key used to retrieve that person's record from the customer's own connected platform, meaning much of the person level data originates with the customer rather than being bought.

Against that, the pricing page describes account segmentation with waterfall enrichment, and a waterfall by definition cascades across a series of providers, which does not reconcile with two enrichment suppliers appearing on the list. The intent capability gated at the second tier is described only as third party research intent signals, with no provider named on any surface.

No licensing terms, no statement of the basis on which either named supplier collected its records, and no indemnification position were located. Ask which providers make up the enrichment waterfall and in what order, who supplies the intent signals, and whether provenance is indemnified.

Own sending infrastructure and sanctioned connectors, which is exactly where the convention places the middle band. Outbound email runs through a delivery provider named in the subprocessor list rather than through any account borrowed from a platform, and the integration surface is a set of official connectors into record, marketing automation, sales engagement, messaging and account intelligence systems, together with a first party listing on the record platform's own application marketplace, which carries that platform's review process as a condition of listing.

Nothing here scrapes a professional network, drives a browser extension against one, or automates messaging on a platform whose terms prohibit it, so the exposures that separate this category fastest do not arise at all. Held below the top band because no conformance position is stated in the vendor's own words: nothing published describes how the vendor keeps each connector within the terms of the platform it connects to, what happens to a deployment if a platform changes those terms, or what rate and volume limits it observes against each.

The acquisition changes the shape of the largest dependency rather than removing it, since the platform this product is built against is now also its owner. Ask for the stated conformance position on each connected platform, and what the migration path is if a connector is withdrawn or restricted.

A framework of principles, and one word doing a great deal of work in the commitment underneath it. The vendor publishes an artificial intelligence trust framework covering transparency, fairness, accountability, privacy and governance, and every element is written as an intention rather than a control: it says the vendor will explain how the systems work, will ensure fairness, will review and approve projects.

No evaluation record, no red teaming artifact, no independently audited management standard for artificial intelligence, and no statement of whether visitor conversations, transcripts or generated output train or tune any model. The one artificial intelligence policy named among seventeen listed policies governs employee use of generative tools in the workplace, which is a workforce policy rather than governance of the agent that speaks to a customer's prospects.

The load bearing sentence sits elsewhere: the vendor states it does not access or use customer content for any purpose other than providing, maintaining and improving its services. Improving is the carve out, and it is precisely the one a buyer would want closed. That sentence appears on the trust page while the agreements it would need to live in are linked but were not opened on this pass, so its contractual status is unestablished.

Ask whether that commitment appears in the agreement itself, whether improving the services includes training or tuning models on customer content or visitor conversations, and what retention applies to transcripts.

The label travels with the product everywhere except to the person it concerns. On every vendor surface the agent is called an artificial intelligence agent, the tiers are sold as agent capabilities, and the email is described as generated, so a buyer is never misled about what they are purchasing.

That is a real difference from the records in this cohort that sell an agent under a human first name as a hire and leave the machine authorship to be inferred, and it is why this sits above the bottom of the scale. It stops at the buyer. Nothing published states what a website visitor is told when a conversation opens, whether the agent identifies itself as automated when asked directly, or whether generated email carries any indication of its authorship.

Two features raise the question rather than settle it. The agent is given a human first name and referred to with a human pronoun in the vendor's own copy. And the highest profile capability is a face to face video conversation with a generated avatar, supplied by a provider the vendor names in its subprocessor list as a generative avatar service, which puts a synthetic human face in front of a visitor and makes the disclosure question sharper than text chat ever does. Ask what the agent says when a conversation opens and when asked directly whether it is a person, and whether generated email is marked as generated.

Depth in one ecosystem that few records here match, and a boundary the vendor states less plainly than its own material implies. The product is built against a single record platform with real time bidirectional synchronisation, and the access model is documented rather than asserted: the vendor states that the objects and fields visible to the integration user determine exactly what can be read and written, which puts the customer in control of the surface area rather than the vendor.

Around it sit more than twenty named connectors across marketing automation, sales engagement, conversation intelligence, account intelligence and messaging, public developer interface documentation, applications on both mobile platforms, a published implementation process, and a training and certification programme. Two things hold it below the top band.

The depth is ecosystem specific, and independent reviewers consistently report that a buyer on a different record platform gets a materially thinner deployment, while the vendor's own description of its ideal customer names the one platform. And parts of the integration surface are commercially gated, with enterprise grade interfaces and sandbox support listed at the second tier rather than included. Ask what a deployment looks like on a different record platform, and which interfaces and sandbox access are included at the tier quoted.

A published negative, which serves a buyer better than the silence this axis usually meets. The vendor names its cloud provider, names the single region it runs in as the eastern United States in Northern Virginia, and states plainly that it has no data storage option in the European Union at this time. That sentence is the finding.

Most vendors in this index leave a European buyer to infer the answer from a privacy policy; this one answers directly and the answer is no. Around it sits real architectural detail: encryption in transit on current transport security with maintained cipher configurations, encryption at rest at block level on a named standard, logical segregation between customers, and continuity and recovery plans tested annually with a named executive owner and a required postmortem after any use.

Held below the top band on what remains missing rather than on the residency answer itself. No recovery time or recovery point objective is published, no tenancy model beyond logical segregation is described, no single tenant or regional election exists at any price, and no failover architecture or second region is stated for a service whose customers depend on it responding to live website traffic. Ask for the recovery objectives, and whether any European processing option exists for a buyer whose own regulator requires one.

A detailed control narrative published in the open, and not one artifact a buyer can verify before a sales conversation. What is published is substantial and better than most records at this grade: an annual service organization control type two audit, annual black box penetration testing by an independent agency against an isolated clone of the application, an inventory of seventeen named policies reviewed annually and examined in each external audit, background checks enumerated to seven specific components, security training at onboarding and annually with an additional technical workshop for engineers, least privilege access with regular reviews, intrusion detection and security event monitoring, and a public vulnerability disclosure route carrying a stated five business day acknowledgement and a seven business day target for critical issues, alongside a bug bounty programme.

Absent is the verifiable layer. No audit period, audit date, auditor name or report is visible, the report itself is available only by contacting a representative, the penetration test summary and the remaining policies sit behind the trust centre and a confidentiality agreement, two compliance badges are displayed without a link to any register, and no information security management certification is claimed at all, where the closest peer in this index holds one and names its auditor and scope unprompted.

This is the pattern this index has consistently held to the middle band, a real security page enumerating real controls with nothing independently checkable beforehand, and the enablement platform graded earlier in this project sits here on thinner evidence than this. Ask for the current report with its audit period and auditor, the penetration test summary, and whether any management system certification is held or planned.

Three tiers, a complete feature ladder, and no number anywhere on any vendor surface. The pricing page names the tiers, lists what each one adds, and states that a plan will be built for the buyer, with every button on the page leading to a demonstration request. There is no self serve path, no free tier, no published trial and no figure of any kind.

The gates are load bearing rather than cosmetic, which removes any argument for a higher grade: single sign on, custom cookie and data retention policies, enterprise interfaces, sandbox support, third party intent signals, multiple brands, multiple production instances and high volume ceilings are all tier conditional, so the capabilities that decide both the bill and the compliance posture are the ones with no price attached.

The convention this index applies is whether a buyer can budget their own purchase, and here they cannot begin. Third party figures exist and were not adopted: reported entry points run from roughly forty two thousand to sixty eight thousand dollars a year, the sources disagree by more than half, most are published by vendors selling against this one, and at least one describes its own range as unverified and directional. Household name status does not soften this.

Two enterprise vendors built earlier in this project, an enablement platform and an account intelligence platform, publish tier structures without figures in the same way and hold the same grade. Ask for the rate at the tier carrying the retention controls and single sign on, what each gated capability costs as an addition to the first tier, and whether any term shorter than a year is available.

Part of the corpus leaves by architecture and the rest is unaddressed. The product writes continuously into the customer's own record platform and the vendor states the customer controls what synchronises in both directions, so visitor activity, lead status and conversation history already sit in a system the customer owns and would keep.

That is portability by construction for the operational record, and it is worth more than an export promise because it does not depend on the vendor's cooperation at termination. Everything specific to the platform is unaddressed. Nothing describes what a departing customer receives for full conversation transcripts, voice and video recordings, agent configuration, routing rules, segment definitions or accumulated intent history, in what format, or across what period.

No post termination retention or deletion timeline for a whole instance appears, and the one deletion timeline published concerns individual visitor requests, which is a different obligation to a different party. Configuration is the underrated part of the exposure, because the agent studio, routing rules and segment definitions represent the tuning work of an implementation the vendor puts at around a month, and none of it is portable in any described form. Ask what exports at termination and in what format, what becomes of recordings and transcripts, and how long an instance is retained before deletion.

The sending path is disclosed and the discipline around it is not. Naming the email delivery provider in the subprocessor list is more than most records here manage, and it tells a buyer that mail leaves through the vendor's infrastructure rather than through the customer's own connected mailbox, which is the single most consequential fact about who carries the reputational risk. From there nothing is published.

No statement describes which domain messages are sent from, who configures sender authentication records and whether the customer's own domain is used, whether a new sending identity is warmed before volume increases, what bounce or complaint rate triggers intervention, whether one customer's sending reputation is isolated from another's on shared infrastructure, or whether any daily ceiling applies to agent generated email.

The exposure is structurally smaller than for the outbound products in this cohort, because recipients engaged first and complaint rates should follow that, but agent composed nurture email at scale to leads who have not replied is still sending that a reputation depends on. Ask which domain sends, who owns the authentication records, whether sending reputation is isolated per customer, and what bounce threshold pauses the agent.

An unusually clear statement of fit, including a statement of what the product does not do, undercut by a contradiction on that same point. The vendor states its market as mid market and larger business to business companies, particularly in technology and software, and states directly that it focuses on inbound pipeline generation rather than outbound. A vendor naming what it is not is rare in this index and saves a buyer an evaluation cycle.

Scale boundaries are expressed structurally through the tier ladder, which gates multiple websites and brands, multiple production instances, high volume websites and high volume contact databases, so a buyer can locate themselves on it before talking to anyone. A multi language agent is published as a capability of the entry tier, which matters to an international buyer. Two things hold it below the top band.

The footer carries a solution page for outbound alongside the stated focus on inbound rather than outbound, and the two are reconciled nowhere on the site. And the single most decisive fit condition, the volume of inbound traffic below which the agent has too little to work with, is never stated, though the entire value of the product depends on it. Ask what monthly traffic volume the entry tier assumes to be worth deploying, and whether the outbound solution is a distinct product or a repositioning of the same agent.