Outdoo AI

One of the cleanest removal tests in this index, because the central artefact of the product is a model and cannot be anything else. What the customer buys is a synthetic buyer who talks back, adapts, objects and pushes, rendered as a video avatar with a stated personality, and up to three of them in one scenario.

Remove the model layer and there is no roleplay, no tutor, no scoring rubric applied to either practice or live calls, no objection detection, no coaching recommendation and no skill benchmark. What survives is a call recorder and a video library, which is the vendor's own earlier product under its earlier name rather than the product being sold today. The commercial design agrees: the free tier meters usage in minutes of model conversation, which is the vendor pricing the thing it actually sells.

A real administrative control set, published plainly, and one large question left open. Every action in the platform is stated to be fully audit logged, access is role based, federated sign on is supported through four named protocols including two open standards, workspaces partition by team, geography and vertical, and personal information scrubbing is offered for regulated environments.

Coaching is designed with a manager in the loop rather than delivered straight to the representative, and the scoring rubrics are built on the customer's own methodology rather than imposed. Off the top band because of the question this category makes unavoidable and this vendor does not answer: the product generates machine scores of named employees on discovery, qualification, negotiation and closing, benchmarks them against colleagues, monitors readiness over time and links the result to win rates and deal velocity.

Nothing published states whether a representative can see, contest or appeal a machine assessment of their own performance, whether human review is required before a score informs a personnel decision, or whether any fairness testing has been done on scores that vary by accent, language or region.

The architecture is described in detail and the model estate is absent entirely. Named components run to tutors, roleplay agents, scorecards, rubrics, objection detection, video avatars and a revenue intelligence layer, each with its own page. No base model, provider, family, version or hosting arrangement appears anywhere on the site or in the pricing questions.

The gap is sharper here than at most vendors because the same page that declines to name any model does discuss sub processors directly, committing that neither the vendor nor its sub processors use customer data for training, so the vendor is willing to make a promise about parties it will not identify. The video avatar layer likewise implies a speech and likeness synthesis dependency that is named nowhere.

Among the strongest evidence bases in this index and short of the top band only on method. Three testimonials carry a full name, a role and a named recognisable employer, each attached to a specific quantified result, covering better objection handling at a listed insurer, reduced call administration at a major infrastructure company, and a single coaching framework across three regions at an events platform.

Behind them sit case study pages for a dozen named customers including two marked as members of a major United States index and one as a member of a British index, spanning insurance, technology, healthcare, real estate and professional services. Four named business and technology outlets are linked directly to their coverage. The vendor is listed on four independent review directories and cites a category ranking for satisfaction.

Off the top band because not one of the quantified figures carries a sample size, measurement period, denominator or method, and the two headline aggregate claims covering ramp speed and win rate improvement are unattributed to any customer at all.

Structurally a weak fit for this axis, graded on what is knowable, with the context carried here. This product sends no marketing mail, runs no sequencer and dials no prospect. Its outbound surface is internal, in that it delivers practice scenarios and coaching to the customer's own employees.

The one place it touches a real recipient is the conversation intelligence half, which ingests recorded customer calls from a dialer or a competing conversation platform, and that is a consent question rather than a solicitation question. No statute, consent framework or suppression mechanism is published, and none of the usual apparatus of this axis would apply to it if it were.

The document set is better assembled than most and none of it was read, so this row is held at the middle band and flagged rather than graded on inference. What is verified is the structure, and it is genuinely above norm: a privacy policy, terms and a separate data processing addendum are each published at their own public route and linked in the site footer, which matters because the addendum is the document most vendors in this index make available only on request.

A trust portal exists on its own subdomain. The corporate entity is named with both a registered office and a mailing address. Four privacy regimes are named including the United States health privacy statute, which is consistent with named healthcare customers. Reading the addendum is the single highest value verification available on this vendor and would most likely move this row.

The baseline is clean and one unusual question sits unanswered. The vendor sells no contact database and buys no lead data; the material it processes is the customer's own recorded calls, its own methodology and its own uploaded content, which removes the licensing problem that dominates this axis elsewhere.

What replaces it is specific to a simulation product: the platform renders lifelike video avatars and publishes persona cards carrying a photograph, a full name, a job title and a named employer, all of them fictional. Nothing states whether those faces and voices are wholly synthetic, licensed from a stock library, or derived from real people who consented to the use of their likeness, and no synthesis provider is named. A product whose core output is a convincing human face has a provenance question about that face, and it is not addressed anywhere.

Official routes throughout and one competitive dependency worth naming. The vendor holds a listing in a major sales platform's official marketplace, ships applications through both official mobile stores, and connects to learning systems through published open standards rather than bespoke scraping.

Named connections cover two sales databases, two conversation intelligence platforms, a forecasting tool, two learning management systems, a messaging platform and a sales engagement dialer, and the vendor states that integrations are included in the license with no additional connection fees.

Off the top band because two exposures are unaddressed: recorded calls arrive from conferencing and dialer platforms with no stated conformance position on their recording or bot participation rules, and the vendor publishes a dedicated migration path off a named competing conversation intelligence platform, which moves recordings and transcripts held under that competitor's subscription into this one with no position stated on the terms governing that extraction.

The training question is answered in the strongest wording this index has recorded outside a contract. The vendor states flatly that customer and user data is never used for training or model improvement, and then closes the gap almost every other vendor leaves open by extending that commitment to its sub processors by name of category, so the promise covers the parties downstream rather than only the vendor itself.

Personal information scrubbing is offered for regulated environments and private cloud deployment is available, both of which are stewardship mechanics rather than assurances. A processing addendum is published openly. Off the top band because the commitment sits in a commercial questions section rather than in the subscription terms or the addendum, no supplemental terms govern the model features as a condition of enablement, no retention arrangement with any model provider is described, and the sub processors covered by the promise are never identified.

This vendor splits cleanly in two on this axis and only one half is answered. On the practice side the disclosure question genuinely does not arise: a trained employee knowingly rehearsing against a labelled simulator, on a platform they logged into for that purpose, is the narrow circumstance the European guidance preserves as obvious from the context, and this is one of the few products in the index where that exception applies honestly rather than being claimed.

The other half is unaddressed. The conversation intelligence layer records and analyses real customer calls, and nothing published states whether the recorder announces itself, joins as a visible participant, follows the customer's consent policy, or handles the jurisdictions requiring all party consent. No position on article 50 of the European artificial intelligence regulation appears anywhere, and machine generated call summaries and follow up suggestions carry no stated authorship disclosure.

Broad, standards based, and explicitly not charged for, which is the part worth crediting most. The vendor states that integrations are included in the license with no additional fee for connecting a sales database, a dialer or a conversation intelligence tool, which removes the connection tax several vendors in this index apply.

Learning system connections run on three published open standards for course packaging, activity tracking and tool interoperability, which is the right answer for an enablement product and is rarer than it should be. A protocol server for external agents and interface access are offered, alongside a listing in a major sales platform marketplace, native applications in both official mobile stores, a help centre and a customer academy.

Off the top band on three counts: the two integration totals published on the same site disagree with each other, programmatic access is gated to different tiers across the two product lines with no stated reason, and no public developer documentation was located anywhere.

One real deployment option is published and the location question is untouched. Private cloud deployment and private cloud storage are offered as named capabilities for regulated environments, which is a genuine architectural choice rather than a marketing phrase and answers the tenancy half of this axis.

Nothing anywhere answers the other half: no hosting provider, country, region, data centre or residency option is stated on any page read, and no European or other regional processing commitment appears. That gap matters more than usual here because the platform stores recorded customer conversations and employee performance records for named enterprises operating across at least three continents, on the vendor's own published customer evidence, and because a health privacy statute is claimed. The trust portal was not read and is flagged as the likely home of an answer.

A complete and specific posture, let down only by where the evidence sits. The audit is named with its type rather than asserted bare, and the vendor states that third party audits recur rather than describing a single historic certification.

Federated sign on is supported through four named protocols including two open standards, access is role based, personal information scrubbing is available for regulated environments, private cloud deployment is offered, and every action in the platform is stated to be fully audit logged. A trust portal exists on its own subdomain and is linked from the footer.

Off the top band because detailed security documentation is available on request for an information security review rather than served self service, which is the distinction that separates the two bands elsewhere in this index, and because no audit period, audit date or auditor is published anywhere.

Note also that the compliance list names an international standards body without identifying which of its standards is held, which is the same category error recorded at other vendors and should not be read as a certification.

Everything a buyer needs except a number. Two product lines each publish a three step ladder with a complete feature matrix, so a buyer can see exactly what moves between tiers, and the free tier is real rather than decorative: it carries a stated allowance of sixty minutes of model conversation credits, unlimited team members, support, and a substantial feature set including video avatars, multi persona scenarios, transcription, scorecards and coaching feedback, with no card and no sales call required.

Two exclusions are disclosed clearly and in the buyer's favour, in that integrations are stated to be included in the license with no connection fees, and implementation is named as separately scoped with a written scope and quote provided before signing rather than discovered later. Annual contracts are stated as standard and a structured proof of concept is offered.

Against all that, not one figure appears for any paid tier on either ladder, the usage based tier names its unit and publishes no rate against it, implementation carries no indicative range, and the enterprise step is quote only.

One genuine portability strength and no published exit position. The strength is standards based and easy to underrate: learning content, completion records and activity data move through three published open standards for course packaging, experience tracking and tool interoperability, which means the certification and training history a customer builds here is expressed in formats other systems can read rather than in a proprietary store.

Private cloud storage on the top tier further reduces the practical cost of leaving. Against that, no export function, file format, bulk retrieval route, post termination window, deletion timeline or retention period was located on any page read. What a departing customer would want back is substantial: recorded customer calls, transcripts, scoring histories against named employees, custom rubrics built on their own methodology, and every roleplay scenario they authored. The terms and the published processing addendum were not read and are flagged as the documents that would settle this.

Structurally a weak fit for this axis, graded on what is knowable, with the context carried here. The product operates no sending infrastructure of any kind: it does not run campaigns, does not connect customer mailboxes for outbound, and does not dial. Its messaging surface is internal notification to the customer's own staff about practice assignments, coaching sessions and scores.

No warmup, authentication, pacing, bounce handling or complaint threshold is published because none of that apparatus applies to what this vendor does. Nothing here indicates poor practice, and nothing in the sending discipline this axis measures is available to assess.

One of the more completely drawn market positions in this index. Five audience pages address revenue leaders, sales teams, enablement teams, learning and development teams and support teams, which correctly recognises that the buyer and the user are different people in this category.

Six industry pages cover software, insurance, banking, credit unions, contact centres and commercial real estate, and each is a real vertical position rather than a keyword page, with a dedicated claims handling training scenario built for the insurance case. Seven use case pages and four named competitor comparisons place the product precisely. The named customer list corroborates every one of those verticals independently.

Regional coverage has one real data point, in that a named customer describes running the same coaching framework across North America, Latin America and Europe. Off the top band because no customer count, seat band, language coverage or explicit region list is published, and there is no statement of who this is wrong for.