Cadence by Ringover

Cadence itself is sequencing and task orchestration, not a model product: templates, steps, timing, assignment to a person or team, and CRM write back. Every AI capability in the Ringover estate is a separately priced SKU sitting outside it, and the rate card is the evidence, in the same class as Apollo's entry tier shipping with no AI at all.

Empower conversation intelligence is 39 USD per user per month, the AIRO voice agent is 0.39 USD per minute, and agent assist, AI coaching, transcription, call summaries and topic detection are all listed as add ons. A customer can buy and run Cadence with none of them. That is the removal test answered by the vendor's own pricing page rather than by an analyst.

The core design is supervised by form factor. Cadence turns a sequence into a queue of dated tasks assigned to a named person, and the calls in it are placed by that person, so the product's main output is a work list rather than an autonomous actor. That is the same architectural oversight that earned Autobound credit on this axis: a system whose job is to tell a human what to do next cannot act on its own initiative.

Two things hold it at C. Email and SMS steps do fire automatically once configured, so part of the sequence proceeds without a person in the loop, and the neighbouring AIRO voice agent handles inbound and outbound calls autonomously at a per minute rate. Nothing published describes an approval gate, a guardrail that withholds a step, an escalation rule or an audit trail of automated actions.

The AI modules are described functionally rather than technically. Empower is said to use natural language processing models to analyse customer interactions, transcribe calls, generate summaries, detect topics and read sentiment, and the voice agent handles conversations, but no model provider, family or version is named, nothing states where inference runs, and nothing addresses accuracy or error rates on transcription despite transcripts feeding coaching and performance review. For a vendor whose entire residency proposition rests on data never leaving France, the absence of any statement about where model inference happens is the conspicuous gap.

Scale is stated and credible, around 14,000 companies on the platform, and third party recognition is real if generic: G2 Leader and High Performer Europe, Capterra Shortlist, GetApp Leaders, Software Advice. A testimonials library exists. What is missing is anything specific to Cadence.

The launch material argues from general industry statistics about follow up attempts rather than from results the module produced, and no case study, named reference or quantified outcome attributable to Cadence was located. The distinction matters because Cadence is an add on: the customer evidence supporting Ringover the phone system does not transfer to the sequencer sold on top of it.

Recorded as observed rather than concluded: no published outreach compliance posture for Cadence was located, no regulation named, no consent framework, no do not call handling and no suppression mechanism. Two contextual facts sit either side of that gap and neither resolves it.

In Ringover's favour, it is a licensed telecommunications operator running its own network with national numbering in at least ten countries, which necessarily carries regulatory obligations a pure software sequencer does not have. Against it, the estate ships the two features this index has repeatedly flagged, voicemail drop and local number presence, the latter matching an outbound number to the recipient's area.

Worth recording as a framing contrast: Ringover markets voicemail drop as less intrusive prospecting, where Buzz markets the same class of technology on the basis that the recipient's phone never rings. Same mechanism, opposite justification, and neither vendor states a consent position.

A dedicated GDPR page sits alongside the privacy policy, which is more than a link in a footer, and the company is a French controller under CNIL supervision, so the regulator with jurisdiction is unambiguous. Two details signal actual work rather than boilerplate. A separate, scoped privacy policy exists specifically for Google Contacts, which means the limited use obligations attached to that API have been addressed on their own terms rather than folded into general text.

And the cookie banner offers a genuine refuse all alongside granular categories, states that the choice is retained for six months, and gives a standing route to change it, which is the shape a compliant consent interface actually takes rather than the dark pattern most sites ship.

Not A because the policy body was not read this pass, so no controller or processor designation, lawful basis enumeration or DPA offering has been verified, and nothing addresses the third parties whose call recordings and contact records flow through the platform.

Cadence ships no contact database, buys no third party data and holds no inventory, so provenance is determinate by construction: prospect records arrive from the customer's own CRM, whether Salesforce, HubSpot, Pipedrive, Zoho, Bullhorn or another connected system, and the vendor never becomes a data broker.

The same structural argument earns Bindago its grade here, and Ringover's version is cleaner still, because the source is the buyer's own system of record rather than an extraction from a platform whose terms are contested. Ringover's own content is explicit that a buyer should pair the tool with a separate data source, naming Sales Navigator, Hunter and ZoomInfo, which is an honest statement of what the product does not do. Not A because none of this is set out as a stated provenance position anywhere; it has to be inferred from the shape of the product.

The dominant channels run on infrastructure the vendor owns and is licensed to operate. Ringover manages its own telephony network and holds national numbering across at least ten countries, so voice and SMS are not borrowed capacity, and the CRM and ATS connectors are published marketplace applications rather than unofficial integrations, with the Cadence app documented on Pipedrive's own marketplace including its sync behaviour, which is third party verifiable in the way this index requires.

That combination, own sending infrastructure plus official connectors, is the B condition. It is not A for one specific reason that should be re verified: LinkedIn appears as a step type inside sequences, and nothing located makes clear whether that step automates an action against LinkedIn or simply queues a task telling a rep to send a message manually. Those two implementations sit in completely different exposure bands, and the vendor does not say which it ships.

Recorded as observed rather than concluded: nothing located states whether customer data trains models, how tenants are isolated, how long transcripts and recordings are retained, or whether prompt content leaves the platform. The gap is more consequential here than for a text only tool, because the AI estate processes recorded human voice on both sides of a call, transcribes it, scores sentiment, detects topics and feeds the output into coaching and performance review, which makes employees as well as prospects the subjects of the analysis.

Ringover also declines, explicitly and on security grounds, to describe its technical procedures in exhaustive detail, which is a defensible stance for intrusion resistance and an unhelpful one here. Re verify against the data processing agreement.

Three synthetic or identity adjacent surfaces ship in the estate and none carries a stated disclosure position: nearest number selection, which matches the outbound caller ID to the recipient's area and manufactures local presence; voicemail drop, which places a pre recorded message; and the AIRO voice agent, which conducts inbound and outbound calls at 0.39 USD per minute. The last is the one that matters, and it produces the sharpest Article 50 observation the index has recorded so far.

Article 50 is European law, its guidelines were finalised in July 2026 and obligations applied on 2 August 2026, the marking duty rests principally on the provider of the system rather than the deploying customer, and Ringover is a French company selling an outbound synthetic voice agent to European buyers calling European recipients. No vendor in this index is more squarely inside the regulation's scope, and none has said less about it.

Over 100 integrations with the important ones named rather than counted, spanning CRM (Salesforce, HubSpot, Pipedrive, Zoho, Sugar, Odoo), service desks (Zendesk, Freshdesk), collaboration (Slack) and an unusually deep staffing and recruiting set (Bullhorn, Avionte, JobAdder, Vincere, Loxo, Mercury) that reflects where the company actually sells. A public developer portal at developer.ringover.com and documented webhooks give a programmatic surface beyond the connector list.

The evidence quality is what earns the grade: the Cadence application is listed on Pipedrive's own marketplace with its person and organisation sync described there, which is exactly the third party verifiable standard Autotouch met and which beats a large vendor's unaudited connector count. Not A because the depth is an order of magnitude below the platforms holding A on this axis and no MCP or agent facing surface was located.

The strongest residency answer graded in this index, and it is unqualified. Ringover states that all data centres storing the data necessary to provide its services are hosted and located in France, and draws the conclusion explicitly, that no data is transferred outside the European Union or the European Economic Area.

There is no regional tier to buy, no exception for processing, no third party trust profile standing in for a vendor statement, and no US fallback of the kind that sits behind most European hosting claims. The company also runs its own network rather than reselling capacity, which is what makes the claim structurally deliverable.

For a European buyer weighing sovereignty exposure under the US cloud Act, this is a decisive difference from every American vendor in the index, and it is the reason a French telephony operator outscores Adobe and Salesforce on an axis where both sit at B.

Specific and technical where most vendors are reassuring. Named protocols rather than adjectives: calls encrypted with DTLS-SRTP, API access over HTTPS and TLS only, password hashing salted at SHA-256 strength or better. A described security programme covering staff awareness training, permanently accessible internal guidance, physical security measures for offices and staff systems, and regular testing.

Hosting certifications are enumerated with their versions, HDS medical data hosting, ISO 9001:2015, ISO 14001:2013 environmental, ISO 27001:2013 and ISO 50001:2011, and HDS in particular is a demanding French certification built on top of ISO 27001.

Two things keep it off A and the first is precise: those certifications are attributed to the hosts, not to Ringover's own information security management system, which is the same distinction that holds Boomerang down where AWS attestations were offered in place of a vendor certification.

Second, the company states plainly that for security reasons it will not describe its technical and physical procedures in exhaustive detail, and there is no self service trust centre, no SOC 2 and no published penetration test.

Unusually complete on the numbers and unusually constrained on the path, and both halves matter. Published on the vendor's own pricing page: a USD and EUR currency toggle, a monthly and annual toggle with annual marked as a 25 percent saving, Talk at 15 USD per user per month against a 24 USD monthly list price, Business at 47 against 57, and Advanced quote only, all stated excluding taxes.

The add on rate card is published alongside, which is the piece ActiveCampaign lacked: Cadence 59 USD per user per month, Empower AI 39, omnichannel messaging 29 per licence, team activity analytics 6, CRM autofill 5 applied to every licence on the account, and the AIRO voice agent at 0.39 USD per minute.

The finding that survives all of it: Cadence at 59 costs more than the Business plan it attaches to at 47, and a working outbound seat of Business plus Empower plus Cadence lands near 145 USD per user per month, roughly triple the plan sticker. Two constraints are published in small type under both priced tiers and are easy to miss: a three user minimum and a twelve month commitment, so the real entry is about 45 USD a month for Talk and 141 for Business, on a year long lock.

Not A because the purchase path narrows sharply above the floor: only Talk carries a Buy now button, while Business and Advanced both route to Get a demo, and several capabilities a high volume outbound team treats as core, call campaigns, voicemail drop, local presence and single sign on, sit in the quote based Advanced tier. Worth re verifying on the next pass: third party analyses from July 2026 record a fourth Smart tier at 21 USD which is not visible on the current page and may have been retired.

One real mechanism and no stated commitments. The mechanism is a public developer API and documented webhooks, which give a customer a programmatic route to extract call logs, activity and contact records on their own schedule rather than depending on a support request, and the CRM write back means sequence activity is mirrored into a system the buyer already owns as it happens. Neither is framed as a portability guarantee.

Nothing located sets out post termination data rights, a deletion timeline or confirmation, retention periods for call recordings and transcripts, or what happens to recorded audio when a subscription ends, which is the specific question a European buyer holding voice recordings of identifiable people will ask first.

Split by channel and thin on both. On voice, owning the network is the equivalent of owning sending infrastructure and is a genuine structural advantage, but nothing published addresses the disciplines that actually govern outbound calling reputation: no STIR/SHAKEN attestation for the US numbers it sells, no answer rate or spam labelling management, no abandoned call governance, no dialling ratio guidance for the power and parallel diallers.

Aloware published exactly that material and holds the A on this axis, which is the comparison to draw. On email, Cadence sequences messages and nothing was located on warmup, domain health, inbox placement testing, bounce handling or volume thresholds. The absence is more notable on a vendor of this scale and technical seriousness than it would be on a small tool.

Coherent and explicitly structured, which is the opposite of the Aritic failure. The vendor segments twice and publishes both cuts: by company size, with a separate product for very small businesses, a small and mid tier and an enterprise tier, and by team, with distinct propositions for staffing and recruiting, sales, and customer service.

Around 14,000 companies, sales presence and national numbering across France, the UK, Ireland, Spain, Italy, Belgium, Switzerland, Sweden, the US and Canada. The recruiting concentration is real rather than incidental, visible in an ATS integration set deeper than the CRM one. Not A because Cadence itself was launched and is still positioned for small and mid sized businesses rather than across that whole range, and because coverage outside Europe and North America is not evidenced.