Lumeris Tom

Best Next Action is the engine and the agent is its voice. The engine reranks each patient's next step as claims, labs, medications and hospital events arrive, drawing on more than 200 risk and prediction models, and the conversational agent turns those priorities into calls and texts, symptom checks and summaries for the care team. Neither works without its models.

Other parts of the platform would stand on their own: claims and EHR data ingestion and normalization, population segmentation, the Care Team Hub for tasks and reporting, all long standing pieces of Lumeris's value based care toolkit, plus the enablement services sold beside Tom. That puts the AI at the center of Tom's two signature modules inside a platform that carries other value.

Tom acts on its own by design. Lumeris calls it agentic AI that reaches out to patients asynchronously, flags risks and starts follow ups without waiting for a person, and its pages describe real time monitoring, escalation, and conversation summaries handed to clinicians. After each symptom check the care team receives a structured summary with a recommended priority.

Against that sits an unqualified claim, made in two places: Best Next Action ensures nothing gets missed even when care teams are stretched thin, and Tom ensures no one falls through the cracks. Outreach, reminders and prioritization run without a person reviewing each step, so the claim covers work no one checks, and no escalation thresholds or limits on what the agent may say to a patient are given.

Best Next Action is described by its inputs rather than its makeup: 60 billion unique clinical data points, more than 200 risk and prediction models, claims and EHR data from over 119 pipelines, labs, medications and diagnostics, updated in real time. The conversational agent is pitched as agentic AI with bedside manner.

No model, model family, provider, training data set or version is named, and the health system page credits Tom with rapid innovation through evolving AI models without saying how a customer learns that a model changed. Google Cloud is named as the platform underneath, a hosting fact rather than a model disclosure.

Two outside parties are named. Google Cloud hosts the platform, and Schmitt-Thompson Clinical Content, which licenses telehealth triage guidelines to medical call centers across North America, supplies the protocols behind symptom checking.

The model layer is not named: nothing says whether the conversational agent runs on Lumeris's own models or on a licensed language model, and if licensed, whose, or under what retention and training terms. The phone and text channels imply telephony and messaging carriers that also go unnamed. Lumeris's privacy policy describes service providers for its websites, such as hosting, analytics and email delivery, rather than for Tom. With claims, EHR data and patient conversations all moving through the platform, every party past Google Cloud that touches them stays out of view.

The outcome figures belong to Lumeris's enablement partnerships rather than to Tom. Partners are reported to save more than $100 per member per month in Medicare Advantage, Medicare ACO partners average 94 percent quality scores, and the enablement model improves documentation accuracy by more than 21 percent, all without a method, baseline, period or named partner.

Scale is substantial, with more than 1.4 million lives on the platform, $90 billion in claims ingested and 60 billion clinical data points, and that measures use rather than benefit. Tom's own track record is anecdotal: a weekly Tom in Practice field report narrating outreach conversations, plus industry awards. No study measures Best Next Action's prioritization, the agent's conversations or the symptom checks against a clinical reference.

Assurances stop at secure health data and a secure architecture on Google Cloud. Tom draws on past visits, lab results and care plans to decide what to send each patient, yet the questions an AI buyer asks go unanswered: whether call transcripts and records feed model training, what retention applies to audio and text messages, and whether anything is de identified on the way to a model.

The health system page notes that Tom keeps pace with rapidly evolving AI models, which makes the training and retention questions sharper rather than settling them. Symptom checking leans on licensed Schmitt-Thompson triage content, a guard on what the agent says, and beyond that no guardrails, hallucination controls or handling of a safety event are described for the conversational agent.

No HIPAA position is stated for Tom. Business associate status and an agreement appear nowhere on the platform, health system or patient pages, and the only privacy document is the Lumeris Group Holdings website privacy policy, effective September 24, 2025. It covers information collected through the sites and mentions sharing with healthcare providers that contract with Lumeris, without describing how patient records inside Tom are handled.

Lumeris runs a Medicare Advantage plan, Essence Healthcare, and Tom ingests claims and EHR data from more than 119 pipelines, which means protected health information moves through the platform at scale. None of the terms that govern it are set out for a health system evaluating Tom.

Safeguards are asserted, not attested. The website privacy policy describes physical, electronic and procedural safeguards that follow industry best practice and a defense in depth infrastructure, while also saying Lumeris cannot guarantee security, and the platform pages call Tom's Google Cloud architecture secure and scalable.

No audit framework is named anywhere: no SOC 2 report, HITRUST certification or ISO 27001 certificate, no trust portal and no statement about outside penetration testing. That leaves a platform holding claims and EHR data for more than a million lives, and placing calls and texts to patients for health systems, with nothing independent for a security team to read in advance. Whether Essence Healthcare's controls as a Medicare Advantage plan extend to Tom is not addressed.

Symptom checking brings Tom closest to clinical judgment, and it runs on a named protocol. Schmitt-Thompson Clinical Content's telehealth triage guidelines, long used by nurse call centers, drive the symptom checks, and after each one Tom gives the care team a structured summary and a recommended priority so clinicians decide the next action. Lumeris describes the feature as complementing clinical workflows while keeping clinical oversight.

No regulatory position is stated for Tom, and nothing says what a patient hears during a symptom check before the care team acts, such as a direction to seek urgent care. Best Next Action's ranking of outreach and screenings orders a population for follow up rather than judging an acute condition.

Clinical content governance is the visible piece. Symptom checking follows licensed Schmitt-Thompson triage guidelines, Lumeris's Tomorrow Labs says its teams co create Tom with physicians and patients, and a DocTakes series on clinical informatics and content standards sits behind a cookie consent wall.

What is missing is any evaluation of the models: no testing of Best Next Action's prioritization for accuracy or fairness, no results by age, race, payer type or language for an engine that decides which patients get attention first, and no description of how conversation quality is reviewed. The stated aim of supporting equitable, team based care has nothing measured behind it.

Review exists where Tom hands work to people: escalations and conversation summaries reach clinicians, and each symptom check ends in a summary and priority for the care team to act on. Those leave a trail a clinician can check.

Nothing stands behind the outputs Tom produces on its own, the outreach messages, education and prioritization. Best Next Action and the agent come without any measured miss or mistake rate, no warranty or indemnity is offered, and the site terms disclaim accuracy for website content only. Lumeris's willingness to share risk with partners on cost and quality is a commitment about financial results, not about whether Tom's outputs are right.

Data flows in at scale: Best Next Action ingests clinical data from more than 119 EHR pipelines alongside payer claims, labs and medication data, and Lumeris says Tom fits existing EHR, analytics and communication systems and embeds real time insights in the record.

No EHR is named, nor any interface standard, marketplace listing or partner site where Tom writes back into the chart, so the depth of the connection at a given health system cannot be checked from outside. Conversation summaries and escalations reach clinicians through the Care Team Hub, which may sit beside the EHR rather than inside it.

Tom runs as a hosted platform built on Google Cloud, which Lumeris names as the base of its secure, scalable architecture. Region, data residency and tenant isolation are not stated, and neither is whether call audio and text messages stay inside Google Cloud or pass to other services along the way.

Deployment otherwise means plugging Tom into a health system's EHR, analytics and communication systems, normalizing its claims and clinical data, and configuring agents to match its workflows and staffing, with escalations routed to the Care Team Hub. Lumeris pitches national reach for multi state networks. Every setting named is in the United States, which implies US hosting without committing to it, and no private or customer hosted option is offered.

No price is published for Tom. It is sold as Primary Care as a Service to health systems, physician groups and payers through a demo request, and often packaged with Lumeris's value based care enablement: payer contracting, funds flow and governance design, and coaching, with Lumeris saying it shares risk and stays accountable for results.

That alignment suggests part of the economics rides on shared savings, yet nothing states whether Tom itself is licensed per member, per patient engaged or inside a risk arrangement, and setup fees are not mentioned. A buyer can tell the commercial shape, platform plus enablement with incentives tied to performance, but not the figures.

Primary care across payer types is the target: Medicare Advantage, ACO REACH, MSSP, Medicaid, commercial and fee for service, for health systems, physician organizations and payers, with a rural health transformation offering as well. The patient side covers chronic condition support with symptom checks, vitals tracking and education, transitions of care and follow up after discharge, medication adherence, screening reminders and annual wellness visits.

Lumeris's 15 years in Medicare Advantage and multi payer ACOs give weight to the value based settings, where its partner quality and savings figures sit. The newer fee for service and commercial use, and the agent's work with particular populations or languages, have nothing published behind them.