mPulse

The origin is messaging infrastructure and much of the platform still is. Four billion touchpoints a year is a communications operation, and three acquisitions since 2023 added member portals, experience analytics and cross channel print, none of which needs a model.

Where the intelligence genuinely sits is in two places: predictive analytics deciding which members to contact and when, and conversational handling of what a member sends back. Both are real and both are consequential, since targeting determines who receives anything at all.

Graded C on the same basis as the other platform records in this index whose artificial intelligence arrived on top of an existing business. A buyer should establish what proportion of their programme is model driven targeting and what proportion is scheduled campaign delivery, because the second is a mailing operation with a conversational layer.

Substantial autonomy over contact, with a described human path. Campaigns run automatically at scale, the conversational layer answers member messages and returns resources without a person, and published material shows programmes handing off mid conversation to a live agent when the member needs one.

The autonomy that matters is upstream. The system decides which members are contacted, through which channel and at what moment, and a member the model does not select receives nothing. Held at B because the automated actions are communications rather than clinical decisions, and because escalation to a person is a described feature rather than an inference.

The capabilities are named and the mechanisms are not. Conversational artificial intelligence, natural language processing and predictive analytics appear throughout published material without a model, an architecture, or a measure of how well any of it works.

No accuracy figure for intent recognition, no containment or escalation rate for the conversational layer, and no description of what the predictive models use as features. For a system whose main effect is deciding who gets contacted, the feature set is the substantive disclosure and it is absent.

Nothing identifies any party in the chain: no model or model family, no hosting arrangement and no sub processor list was located in two passes, and no retention schedule, encryption detail or secondary use position was found. The concentration is considerable and its origin is what makes it a distinct question rather than a matter of scale.

Member records span more than four hundred and fifty organisations with a reported four billion touchpoints a year, and one hundred and fifty million member communications arrived with one of three businesses folded in since 2023. Each acquisition merges another population's communication history into one platform, and those records were originally collected under agreements made with different customers, in different years, under different representations to the members concerned.

Nothing published describes how those datasets are separated or governed after integration, whether the combined set is used to develop models serving every customer, or whether a plan that contracted with one of the acquired businesses agreed to anything resembling the present arrangement. That last question is the one an acquired customer should ask, because consent and contracting travel less well through an acquisition than data does. Ask whether the acquired estates are joined, what governs the combined set, whether communications train models across customers, and retention on message and call content.

Better evidenced than most engagement vendors, and the outcomes measured are commercial rather than clinical.

Independent corroboration exists: recognition as a Strong Performer in a major analyst evaluation of healthcare experience platforms in early 2026, assessed on current offering, strategy and customer feedback. Scale is stated concretely at more than 450 organisations, over 50 of the 60 largest health plans and 80 percent of top rated Medicare Advantage plans.

Case studies carry real numbers rather than percentages without denominators, including a programme reporting 2.7 times lower disenrolment among contacted members, roughly 186 members retained, more than 1.1 million dollars of value over three years and a 776 percent return by year three.

Held at B because those endpoints are retention, return on investment and survey performance. Nothing published measures whether members were healthier, only whether they stayed and how they rated the plan.

Graded on an honest basis. No retention schedule, encryption detail or secondary use position was located in this pass.

The data concentration is considerable and worth naming: member records across more than 450 organisations, four billion touchpoints a year, and 150 million member communications acquired with one of the three businesses folded in since 2023. Each acquisition merges another population's communication history into one platform, and nothing published describes how those datasets are separated or governed after integration.

Graded on an honest basis. No compliance statement or agreement posture was located in this pass.

One good practice does appear in published material and deserves noting: at least one described programme required double opt in before messaging members. Consent discipline is the operative control in outbound engagement and a vendor showing it in a case study is showing more than most.

Recorded honestly and provisionally: the dedicated trust and security search this index requires was not run in this pass, and no attestation was encountered incidentally.

A platform serving more than 50 of the 60 largest health plans has certainly passed payer security review repeatedly, and none of it was retrieved.

No device pathway applies and none is claimed.

Three other regimes do apply and matter more here. Telecommunications rules govern automated outbound contact, which is why consent discipline appears in the company's own case studies. Federal marketing and communication rules constrain what a Medicare Advantage plan may say to members and when, particularly around enrolment periods, and one published programme describes messaging members approaching Medicare eligibility, which is exactly the territory those rules police. And the Star Ratings system defines the measures the whole platform optimises toward, including the patient experience survey.

A buyer should treat the compliance surface here as marketing and telecommunications regulation rather than clinical regulation.

Nothing published on evaluation, monitoring or subgroup performance, and one published case study describes a targeting logic precise enough to examine on its own terms.

In it, predictive analytics identify two groups: members most likely to rate the plan poorly on the federal experience survey, and members most likely to respond positively to outreach. Outreach is then concentrated on them. Both criteria are rational for a plan whose revenue depends on that survey, and neither is clinical need. The first targets people by the influence their opinion will have on a score; the second selects the members easiest to move.

The consequence follows directly. A member who is unwell, unlikely to respond and unlikely to complete a survey ranks below a member who is well, responsive and likely to be counted. This index has now recorded the same structure targeting by cost saving potential, by claim value and by shift fill difficulty; this is the version that targets by measurement influence.

Nothing published addresses whether contact rates, channel choice or response differ by language, age, disability or rurality, in a population that is disproportionately old and poor.

Two passes located no accuracy figure for intent recognition, no containment or escalation rate for the conversational layer, no description of what the predictive models use as features, no evaluation methodology and no warranty, indemnity or remediation commitment. Conversational artificial intelligence, language processing and predictive analytics appear throughout published material without a model, an architecture or a measure of how well any of it works.

The missing disclosure that matters most is the feature set, because of what the system decides. Its main effect is deciding who gets contacted, so the features are the eligibility criteria in everything but name, and whether socioeconomic proxies, geography or prior non response contribute determines whether outreach reaches the members who need it or the members easiest to reach.

Those diverge in a predictable direction, and prior non response is the sharpest example: a member who did not answer before is both harder to reach and more likely to be disengaged from care, so a model that learns from response rates will deprioritise exactly the people an outreach programme exists for. Nothing published states whether that is guarded against.

The escalation rate matters equally, since a conversational layer that fails to hand off a member describing a serious symptom produces a harm nobody records. Ask for the feature set, the escalation criteria and rate, and outreach reach by member subgroup.

Stronger than the payer side norm because it reaches into provider scheduling. Published material names real time integration with the dominant record vendor's scheduling module, which is what allows an engagement campaign to book rather than merely remind, and portal capability arrived with one of the acquired businesses.

Held at B because no interface standard, certification or breadth of record system coverage is described, and because the primary data direction is payer side: eligibility, claims and member files rather than clinical records.

Not described. A hosted platform delivering across text, voice, web, portal and physical mail, with no hosting model, region or retention position located.

The integration of three acquired businesses since 2023 raises a specific question a buyer should ask: whether their programme runs on one platform or on several joined by a common interface, because the answer determines where their member data actually sits.

No rates published, and the mechanism is described by third party analysis as custom enterprise pricing on subscription and value based models, with annual contract values for comparable platforms placed between 50,000 dollars and over a million.

Value based pricing is the interesting part and the thing to pin down. If a fee moves with measured improvement, the measure defines the incentive, and here the measures are Star Ratings components, retention and survey performance. A plan should establish exactly which metric the fee is tied to, over what baseline, and who computes it, because a contract paying for movement in a patient experience score rewards changing the score by whatever means work. Also establish whether the conversational and predictive components are priced separately from message volume, since four billion annual touchpoints implies volume based economics underneath.

Broad across payer types and buyer categories. Published material names Medicare Advantage, Medicaid managed care and commercial plans alongside health systems, accountable care organisations, value based providers, management services organisations and pharmacies, with more than 450 organisations served.

Depth is greatest in Medicare Advantage, where the measures the platform targets carry the most revenue and where the company reports reaching 80 percent of top rated plans. Clinically it is measure driven rather than specialty driven, so coverage follows quality programmes: adherence, preventive screening, chronic condition management. Held at B because everything is communication and targeting rather than care, and coverage is United States only.