Private beta · care fidelity for dementia programs
GentleNext combines working GUIDE navigation workflows with a research-mode evidence layer now in design-partner validation. Turn plans, assessments, contacts, and follow-up into source-linked review queues—without replacing clinical judgment or the EHR.
The care-fidelity preview below is synthetic and proposed. Current product demos show implemented navigator and family workflows.
Navigator console
Family app
Today
Margaret has a doctor's visit Tuesday at 2:00 PM. Read the brief before you go.
From Priya, your navigator
Refill Donepezil at CVS
Before Tue, 2:00 PM
I need help now
Navigator console (left) + family PWA card (right).
Built for HIPAA-gated pilot readiness
Built for the CMS GUIDE Model
G-code-aware billing, tier-aware cadence
HIPAA Business Associate posture
BAA gates tracked before PHI pilots
Hash-chained audit log
7-year retention, operator-verifiable
PHI never leaves the tenant
Every query is scoped to your program and re-verified on every request
GUIDE creates a reachable first market. The larger problem is making multicomponent dementia care observable without turning navigators into research coordinators.
A task can be complete while the intended component, dose, adaptation, and outcome remain invisible. Supervisors reconstruct delivery manually from assessments, calls, plans, and messages.
Source: Common dementia-care program definition, TRCI 2023
A recent multi-site dementia intervention study found uneven fidelity and incomplete longitudinal enactment data. Without provenance, teams cannot tell whether delivery failed or the record did.
Source: ADS Plus fidelity study, 2026 (PMC12448453)
Dementia-care trials report different effects across caregiver burden, quality of life, utilization, and context. A program label does not reveal which components reached which dyads.
Source: Care Ecosystem and D-CARE outcomes, 2024–2025
Research-mode concept · synthetic preview
GentleNext is being narrowed around a harder problem: knowing which dementia-care components were intended, what actually happened, and what still needs a qualified human to resolve.
Evidence compiler preview
Intended care
Reassess caregiver burden, document the response, and record whether the care plan changed.
Source evidence
ZBI-22 completed
Assessment · Feb 14 · score and instrument version linked
Follow-up contact documented
Contact log · Feb 18 · navigator and source record linked
Care-plan adaptation
No source record satisfies the proposed pathway rule
Start with deterministic evidence and human review. Earn the right to learn later.
Program + clinical lead
A versioned pathway names the trigger, component, responsible role, allowed adaptation, required evidence, and outcome window.
Evidence compiler
The system proposes mappings from assessments, contacts, tasks, and care plans while retaining the source record, rule version, and uncertainty.
Qualified reviewer
Deterministic rules surface supported, missing, contradictory, and indeterminate components. A human resolves material ambiguity.
Program + research team
Reviewed fidelity data can be linked to caregiver, beneficiary, operational, and economic outcomes. Adaptive policy work remains a later, separately governed stage.
Two apps. One case.
GentleNext is one platform with two intentionally separate interfaces — because an RN building a care plan and a daughter reading it at 2am have completely different needs.
Navigator console
The navigator sees a caseload table with burden-risk flags, overdue-cadence indicators, and tier classifications. Assessments, contact logs, escalation routing, and GUIDE-compliant documentation export are all one click away. The AI doc-assist produces a structured SOAP draft the navigator edits and approves before anything is filed.
Audience: RN, LCSW, care coordinator — 4–8 hours/day at the screen, desktop-first.
Family PWA
The family caregiver opens a magic-link on her phone. She sees one sentence — what to do today, from her navigator by name. Large text, 56×56 tap targets, dark-mode default, no gamification. If something is wrong, one tap reaches a human.
Audience: family caregiver, age 55–75, short sessions, often late at night. No account creation — magic-link only.
How they connect: When a navigator logs a contact, updates a care plan, or publishes a visit-prep brief, the family sees it on their next open — not as a notification flood, but as a single updated card or task. The navigator is always named, and the family has a named navigator to reach.
GentleNext is built around the situations that happen every week in GUIDE programs.
Newly diagnosed
Sarah, her daughter, doesn't know what to do first. Priya, the navigator, enrolls the family, completes the ZBI, and publishes a single next step to the family app. Sarah sees it that evening.
Enrollment + ZBI: ~12 minutes. Sarah's first calm next step: same day.
Overdue reassessment
The ZBI is due. GentleNext surfaces it in Priya's caseload before it becomes a compliance gap. Priya schedules the call, documents the score, and the GUIDE reporting record updates automatically.
Surfaced 14 days before the cadence breach. Reporting record: instant.
Urgent concern
She opens the family app and taps 'I need help now.' The concern routes to the on-call navigator. AI does not answer, intercept, or close the contact — a human responds.
Routed to the on-call navigator in under 60 seconds. Callback timing is an operational program target — not a measured product SLA.
Visit prep
Priya drafts a one-page brief: current medications, cognitive scores, and the three questions to ask the geriatrician. Sarah can read it or print it. No login required.
Draft: ~90 seconds (AI). Navigator review + publish: ~5 minutes. Sarah sees it: instant.
Not a feature tour. Three moments the software was built around.
Every contact closes with a structured SOAP draft the navigator edits and approves. Clinical-advice language is blocked at generation time. Every AI call is logged with prompt and model version so a reviewer can retrace the decision two years later — without digging through email.
Caregivers open the app at 2am. They see one calm sentence, one primary action. Large type, 56×56 tap targets, no streaks, no dark patterns, no red. Installable as a PWA with an offline fallback page when the network is unavailable.
Eligibility checks at intake. Tier-aware contact cadence monitoring. Monthly DCMP billing that runs idempotently and exports a billing CSV (G-code and amount per enrollment). Re-running the export is safe — unchanged rows are skipped.
One closed loop
The same action that closes documentation updates the caregiver surface — without a chatbot and without a notification flood.
Step 1
Navigator closes a call or visit in the console.
Step 2
SOAP or visit-prep draft is edited and explicitly approved — never auto-filed.
Step 3
The caregiver opens a magic link and gets one calm next step in the navigator's voice.
GUIDE domains covered
From ZBI/CDR/FAST assessments through respite caps, HRSN, and helpline routing — plus navigator training gates and CE records for GUIDE readiness.
Security & compliance
PHI-bearing production pilots remain gated on signed vendor BAAs, but the product already enforces tenant scope, audit trails, server-side redaction, and human review.
BAA gates
Microsoft coverage is documented; Clerk, DigitalOcean, DO Inference, and Sentry stay launch blockers until signed BAAs are archived.
Tenancy
Every PHI query is scoped to your tenant and re-verified on every request, so a navigator at Program A is kept out of Program B's data.
Keys
Each program's encryption key is wrapped separately and unwrapped only when needed, held in memory for at most 5 minutes, with rotation that can't leave data half-encrypted.
Audit
Every access and change is written to a tamper-evident audit log — any attempt to alter history breaks the chain, and an auditor can verify it directly.
Sessions
Family magic-link sessions rotate on auth change; HttpOnly + SameSite=Lax + Secure in prod.
No third-party analytics on family surface
No GA, no Mixpanel, no session replay. First-party telemetry only.
We are seeking dementia-care programs with a costly, recurring question about whether intended care was actually delivered. A paid, scoped design-partner pilot should prove that value before either side commits to a broader platform rollout.
Validate first
An 8–12 week engagement focused on one pathway, one evidence problem, and one accountable program owner.
Scoped with your team
Priced to case volume, data preparation, and research support.
Extend after evidence
Expand the working navigation platform only after the pilot shows measurable value, usable evidence, and an accountable buyer.
Value-based proposal
No permanent public PMPM until willingness to pay is validated.
Paid pilots only. PHI remains blocked until required agreements and controls are complete.
GentleNext is purpose-built for structured dementia care programs, not general caregiving apps.
CMS lists 320 active GUIDE participants in a model that runs eight years, through mid-2032 — a durable program, not a closing window. Every aligned patient earns the program a tiered monthly Dementia Care Management Payment, plus up to $2,500/yr in respite. Assessment workflows (CDR, FAST, ZBI, HRSN), DCMP PBPM tracking, five-tier classification, respite cap enforcement, and the GUIDE reporting export run end-to-end today. Billing lines reserve PBA and health-equity adjustment fields; quality-measure PBA math lands with CMS PY2 pay-for-performance (Year 1 remains pay-for-reporting).
Health systems running memory-care outside GUIDE get the same navigator workflows — assessment cadence, caregiver-burden tracking, care plans, family communication — without the CMS G-code reporting and PBPM modules. Same console, narrower compliance surface.
Practices with a care-coordination function use GentleNext between visits: post-appointment follow-through, ZBI-tracked caregiver burden, family-side resource handoff, escalation routing to the practice's own on-call. Reduces re-asking the same questions visit-over-visit.
MA plans operating dementia-specific supplemental benefits or in-house care-navigation teams use the same workflows their plan-side care managers already follow, with HIPAA-aware tenant scope per plan and a calm family surface their members will actually open. Not employer-side caregiver concierge; plan-operated programs only.
About
GentleNext is a product of Mahumtech, an engineering studio that ships HIPAA-grade software for regulated healthcare programs. The platform is maintained by a small senior team — clinical advisors, a care navigator, and the engineers who wrote the audit layer — not an outsourced contract shop.
We are pilot-stage and say so out loud. That means we over-invest in launch support, we write detailed compliance briefs for your CISO, and we answer every email within a business day. If you're considering us, expect to talk to the team that wrote the code — not a BDR.
Design-partner conversation
In 30 minutes, we will test whether GentleNext can reduce the cost of answering it. We can show the implemented navigator and family workflows, then map a proposed evidence pilot. No claim that the fidelity engine is already live; no production PHI is required.
Loading request form…