Catch the patient slipping
before it becomes a crisis.
MahCare AI is the platform UK and US care teams use to spot deterioration early — through remote monitoring, virtual wards and calibrated risk — and to run the whole service around it: referrals, appointments, medications, visits and messaging in one place. Safe AI assistants do the admin and the chasing, predictions flag who needs attention now, and a qualified clinician always makes the final call.
What is MahCare AI
One place to run care — from intake to outcome.
MahCare AI brings the whole operational side of care into one place: referrals and intake, scheduling, care plans, home and remote monitoring, medications, patient messaging, billing prep and compliance. Work is prioritised by who needs attention most — not just what landed in the inbox first — and it connects to the record systems, pharmacies and devices you already use.
On top of that, it does what a plain dashboard cannot: it forecasts who is likely to deteriorate, miss an appointment or fall behind on medication — so your team can step in early — and it shows the reasoning behind every flag. Most tools in care only monitor and report. MahCare AI helps you act, with a person always in control.
Care execution layer
Turns referrals, plans, alerts, and visits into managed work — prioritised by predicted risk, not just queue order. Reduces missed handoffs, overdue work, and operational chaos.
Patient engagement layer
Runs patient, proxy, and caregiver messaging across app, SMS, email, voice, and letter, personalised to each patient's state. Improves adherence and attendance.
AI workforce layer
A governed roster of AI workers for intake, triage, drafting, outreach, and coding — every consequential output challenged and risk-checked before it is shown or acted on.
Reasoning & prediction layer
Causal models, calibrated forecasts, and what-if simulation that explain why something is likely and what the next-best action is — with honest uncertainty on every number.
Evidence & compliance layer
Verifiable provenance for every write, decision, and AI output. Audit trails, DSAR packages, access reviews, and investigation packs that an external auditor can check.
Developer & marketplace layer
APIs, SDKs, event streams, configuration packs, and partner apps. Expands distribution and product stickiness without forking the core.
What makes it different
Seven things monitoring platforms cannot do.
These are not add-ons bolted onto a dashboard. They are the reasoning substrate of the platform — the reason a recommendation is trustworthy enough to act on.
Causal reasoning
Recommendations are grounded in an estimable cause-and-effect relationship, not a correlation. When an effect cannot be established, MahCare says so instead of guessing.
Honest uncertainty
Every risk score arrives with a calibrated confidence band, validated against real outcomes. No bare probability is ever shown on its own.
Self-checking AI
Every consequential AI output is argued against by independent critics and passed through a risk gate before it is shown or executed. Weak or unsafe outputs are blocked or sent to review.
Care simulation
Roll the future forward before you act — "what happens if we escalate now versus wait" — and compare the chosen action against the next-best alternatives in real time.
Adaptive autonomy
An AI worker earns more independence as it proves itself, and loses it the moment calibration drifts or incidents rise. Autonomy is a dial that responds to trust, never a fixed setting.
Verifiable evidence
Every decision carries cryptographically anchored, externally verifiable provenance — the inputs, model and prompt versions, and reviewer decisions behind it. Proof, not assertion.
Federated learning
Opt-in, privacy-preserving learning across organisations — the models get better from collective outcomes without any raw patient data ever leaving your walls.
Eight pillars, one platform
Every pillar a first-class product surface.
These are not modules bolted together. Each pillar is engineered as a primary surface with its own deployment model, data guarantees, and commercial uplift path.
Command Centre
Unified operational command surface for work, risk, capacity, patient state, and branch performance — with a live forecast and the reasoning behind every alert.
Care Graph
Longitudinal, causally-aware graph connecting patients, episodes, tasks, messages, and outcomes — the substrate that lets the platform reason about cause and effect.
AI Workforce
Governed AI worker runtime. 23 task-specific agents with adaptive autonomy, model routing, self-checking, and human-review gates.
Evidence Ledger
Verifiable audit, evidence, policy, and review fabric across every workflow and AI decision — externally checkable, not just internally logged.
Studio
Low-code tooling for pathways, forms, roles, policies, automations, prompts, and integrations.
Marketplace
App, connector, template, and pathway distribution channel. Partners amplify distribution and extensibility.
Benchmark Network
Opt-in, privacy-preserving analytics across tenants for operational, engagement, and pathway benchmarks.
Integrations
FHIR-aligned canonical model, OpenAPI 3.1 contracts, SSO, SCIM, messaging, and payer adapters.
A governed AI workforce
Twenty-three AI workers.
Every one checks its own work.
Twenty-three named agents, led by the patient-care ones — remote monitoring, virtual wards, deterioration. Every worker records model, sources, reviewer, and outcome; every consequential output is argued against and risk-checked before it is shown. Clinically influential outputs require human approval. Medication plans are untouchable.
RPM Monitor Agent
Watches remote-monitoring streams and proposes an escalation when a patient drifts outside agreed parameters.
Virtual Ward Agent
Watches a virtual-ward caseload and proposes who to escalate, visit, or step down.
Hazard Field Curator Agent
Curates a live, caseload-wide risk map of who is deteriorating or at risk now.
Medication Adherence Agent
Spots patients likely missing doses and proposes a nudge or clinician follow-up.
Care Coordination Agent
Watches care-plan execution and proposes the next tasks and chase actions.
Documentation Agent
Drafts notes, visit summaries, transfer summaries, and discharge packs.
Assessment Agent
Drafts a structured pre-visit assessment with cited drivers for a clinician to complete.
Handover Agent
Drafts a structured SBAR shift handover — state, tasks, escalations, and risks.
Referral Intake Agent
Reads inbound referrals, extracts facts, and proposes a triage band and routing.
Inbox Triage Agent
Sorts messages, tasks, and alerts by priority, surfacing urgent items first.
Coding & Revenue Agent
Proposes evidence-cited billing codes and flags upcoding and under-capture.
Safety Reviewer Agent
Adversarially reviews clinically-influential outputs and can block or escalate how they are shown.
No AI autonomously prescribes, discontinues, or silently writes to the legal clinical record. Clinically influential outputs require human review. This is a hard product safety boundary, not a toggle.
Who uses MahCare AI
Built for the teams keeping patients safe between visits.
Community and home-care operators
High coordination burden. Medication follow-up complexity. Multi-site visibility need. Compliance pressure from CQC, DTAC, DSPT.
Private clinic groups
Growth pressure. Patient communication at scale. No-show reduction. Pathway standardisation. Private billing handoff and quote management.
Ambulatory specialty groups
Scheduling friction. Prior authorisation burden. Documentation overhead. Patient follow-up leakage. Value-based care reporting needs.
Care management organisations
Longitudinal coordination. Risk stratification with calibrated uncertainty. Outreach burden. ROI sensitivity. Strong fit for Care Graph and predictive outreach.
Pharmacy-linked services
Adherence monitoring. Refill coordination. Patient communications. Task routing. Evidence and audit for dispensing workflows.
Digital-first healthtech
Operators building new models that need clinical safety governance, audit infrastructure, and multi-channel patient engagement out of the box.
Measurable journeys
Eight journeys. Each with a hard outcome metric.
MahCare is judged by hard metrics. Every journey traces to source events. Every source event is on the ledger. Every metric is exportable and independently verifiable.
Referral to first contact
Intake, eligibility, triage, scheduling, reminders, handoff, escalation.
Care-plan execution
Versioned plans, task emission, reminders, observations, reviews, closure summaries.
Medication coordination
Reconciliation, refill reminders, omission capture, pharmacy follow-up, adherence interventions.
Observation escalation
Threshold checks, alert routing, acknowledgement, action tasks, closure evidence.
Visit execution
Scheduling, mobile offline checklist, capture, sync, follow-up tasks, documents.
Patient engagement loop
Templates, channel routing, reminder sequences, reply triage, proxy handling.
Prior auth & revenue prep
Evidence assembly, checklist completion, tasking, payer communication, outcome tracking.
Compliance response
DSAR, access review, incident pack, legal hold, export approval.
Deployment-ready
One core. Two country packs. Zero forks.
United Kingdom
Built for NHS-adjacent reality
NHS Login and NHS Notify adapters; Spine PDS, GP Connect and ODS lookups. DTAC and DSPT workflow support. Clinical safety workflow support (requires customer-side Clinical Safety Officer engagement). dm+d medication terminology. GDPR and DPA 2018 operating workflows.
United States
Ambulatory & value-based care
Designed to HIPAA Security Rule principles. US Core and SMART on FHIR adapter patterns; CMS Blue Button 2.0 claims, openFDA drug data and NPPES provider lookups. Prior-authorization workflow design. NPI provider identifier support. RxNorm and NDC medication terminology.
Engineering principles
Evidence over assertion. Always.
Causal and evidenced by default
Recommendations are tied to an estimable cause, and every write, decision, and AI output is committed to the Evidence Ledger — independently reviewable, cryptographically anchored, exportable as verifiable evidence on demand.
Self-checking, human where it matters
Every consequential AI output is challenged and risk-gated before it acts. Clinically influential outputs require review. Medication plans are immutable without confirmation. Break-glass access creates visible review items. No silent mutations.
Configurable without forking
Tenants configure agent autonomy, review thresholds, model providers, retention, and market-specific policies through Studio. Enterprise control without engineering intervention.
Design partner programme open
See care that predicts, reasons, and proves itself.
We are onboarding design partners across the UK and the USA. Clinicians, operators, and founders welcome. Bring the hardest workflow you have — we want to see it run on MahCare.