← Agent Teams

Crew 10 · HavenKrew by wGrow

HavenKrew

Community living,
confidently.

Personal support. Familiar places. A connected care circle.

A familiar journey can make the neighbourhood feel more accessible. HavenKrew connects resident needs, verified places and human-approved guidance—online and throughout the estate.

One coordinated AI crew prepares the materials. Residents, caregivers and community teams shape how they are used.

Concept previewIllustrative screens and data · Proposed capabilities for pilot evaluation · No HDB endorsement
HavenKrew / Community overviewView original (opens in a new tab) ↗
HavenKrew overview concept showing a fictional precinct, resident support plans, publication queues and the next supervised trial.
A shared workspace for the support plan, local environment, content and review queue. All figures shown are illustrative.

From digital coordination to everyday confidence

A support plan should
reach the places
people actually use.

Resident preferences, route information and caregiver guidance often live in separate places. HavenKrew proposes a shared source for preparing, approving and maintaining support across the digital and physical environment.

01

Familiar places. Personal support.

Connect a resident’s preferred language, recognisable landmarks and agreed assistance to one support plan. Prepare guidance around the person and the actual place.

02

One plan, many useful formats.

Prepare large-print cards, landmark images, spoken prompts and captioned videos from the same reviewed route. Carry the guidance from the workspace into homes and shared spaces.

03

Human judgement, throughout.

Give facilitators and reviewers a shared view of source photographs, proposed instructions and open questions. People verify the route, approve materials and decide the support needed.

04

Learn from the journey.

Record supervised observations of hesitation, wrong turns and assistance. Compare approaches and refine materials through review, with evidence to support the next pilot decision.

One everyday journey

From home to the
community hub.
One familiar step at a time.

A resident and their companion prepare for a visit to a nearby Active Ageing Centre. The team checks the actual route, chooses recognisable cues and prepares a consistent set of guidance.

Illustrative scenario. Locations and people shown in the screens are fictional.

The support travels with the person

  1. Home & lift lobbyConfirm the starting point and familiar block entrance.
  2. Covered walkwayUse a verified photo and a short, clear instruction.
  3. Garden pavilionRecognise a landmark using an image and spoken cue.
  4. Community hubArrive with the agreed companion support.

A printed card, an audio prompt and a resident guide all refer to the same approved route version.

How the system works

Understand. Prepare. Approve.
Use, observe and improve.

The proposed workflow joins resident context and verified environmental evidence. Approval applies to a specific version; changes return to review.

  1. 01

    Understand

    Confirm resident preferences, permissions, goals and the care circle. Agree the level of support with the responsible team.

  2. 02

    Verify the place

    A facilitator walks the route, checks accessibility and records recognisable landmarks with real photographs and review dates.

  3. 03

    Prepare together

    Six specialist agents assemble a support plan and draft text, images, audio and video from the verified source set.

  4. 04

    Review & release

    Named reviewers resolve concerns and approve a specific version. Physical installations need the estate authority’s approval.

  5. 05

    Use & observe

    Distribute approved print and downloadable media. Run supervised journeys, record observations and return proposed changes for review.

The publication boundary

Draft → specialist checks → authorised human review → approved release → versioned distribution. A changed route or unresolved safety concern pauses publication until the responsible team has checked it.

01 / Text

Clear instructions

Large-print route cards, daily routines, caregiver guides and proposed signage wording.

02 / Images

Recognisable places

Verified landmark photographs, visual step cards and consistent destination cues.

03 / Audio

A familiar prompt

Short spoken directions, paced for the resident and reviewed in the chosen language.

04 / Video

Prepare for the journey

Captioned familiarisation sequences and downloadable materials for agreed devices.

Our Agentic Krew team setup

Six agents.
One reviewed support plan.
People hold the authority.

Each proposed specialist has a bounded responsibility. Shared evidence links resident preferences, verified landmarks, route records and the current material version. Software permissions and review gates govern release.

01

Resident needs

Receives
Confirmed resident preferences, familiar places, communication needs and caregiver observations shared with appropriate permission.
Produces
A structured summary of preferences, support needs and open questions for the support-planning handoff.
Human control
The resident and authorised care team confirm what is accurate, relevant and appropriate to use; the agent does not diagnose or determine capacity.
02

Environment & route

Receives
Staff-verified place records, source photographs, route checks, access conditions and landmark review dates.
Produces
A candidate landmark-based journey with evidence links and flags for missing, changed or unverified environmental information.
Human control
A facilitator verifies the route on site. The relevant estate authority approves any proposed physical installation before use.
03

Support planning

Receives
Confirmed resident needs, verified route information, agreed goals and the available companion or facilitator support.
Produces
A draft support plan with step-by-step cues, assistance conditions and questions requiring professional or caregiver judgement.
Human control
The authorised care team agrees suitability, support level and trial conditions. A plan is revised or withheld when suitability is uncertain.
04

Multimedia creation

Receives
The reviewed support brief, verified landmark imagery, approved terminology, language preferences and output-format requirements.
Produces
Draft large-print text, image cards, spoken cues and captioned familiarisation video, each linked to its source and version.
Human control
Reviewers check language, visual fidelity, accessibility and rights. Navigation landmarks must remain recognisable; drafts require approval before publication.
05

Safety & accessibility

Receives
Draft support plans and media, source references, programme rules, accessibility criteria and pending verification checks.
Produces
A review checklist highlighting unsupported instructions, unclear cues, missing evidence and issues that should block publication.
Human control
Named reviewers resolve concerns and authorise a specific version. Software permission rules enforce the gate; an agent cannot waive a failed check.
06

Outcomes & learning

Receives
Facilitator-recorded trial observations, assistance records, material versions and the pilot’s agreed evaluation measures.
Produces
An evidence-linked summary of observed use, recurring difficulties and proposed changes for the next review cycle.
Human control
The programme lead interprets findings, checks limitations and approves changes. Observations do not establish clinical effectiveness or justify unsupervised use.
Coordination does not confer approval.

Agents draft, check and flag gaps. Care teams agree resident support, facilitators verify places, authorised reviewers approve content, and estate authorities approve physical installations. Model output alone cannot authorise release.

Inside HavenKrew

Twelve views.
One connected support cycle.

Explore the original screen designs, from resident needs to offline materials and field evaluation. These are concept designs, not screenshots of a deployed service. All resident details, places and dashboard results are illustrative.

01 / 12 · Concept design

Overview dashboard

Coordinate resident support, verified places, content queues and supervised trials in one programme view. Identify work needing attention before a material reaches a resident.

HavenKrew overview concept showing a fictional precinct, resident support plans, publication queues and the next supervised trial.
Overview dashboard · Illustrative dataView full size ↗ (opens in a new tab)

02 / 12 · Concept design

Resident support profile

Record familiar places, preferred language, communication preferences and agreed support goals. A shared profile gives the crew and care circle a consistent starting point for review.

HavenKrew resident profile concept showing fictional preferences, support goals, care-circle access and a recent observation.
Resident support profile · Illustrative dataView full size ↗ (opens in a new tab)

03 / 12 · Concept design

Places & landmarks

Maintain photographs, approved wording, verification status and review dates for local places. Staff check the real environment so published cues stay recognisable as routes and landmarks change.

HavenKrew places and landmarks concept showing verified photographs of a block entrance, garden pavilion, lift lobby and community hub.
Places & landmarks · Illustrative dataView full size ↗ (opens in a new tab)

04 / 12 · Concept design

Route studio

Arrange a journey around familiar landmarks and review each decision point. Keep the route, resident cues and assistance conditions together before an authorised person approves the guidance.

HavenKrew route studio concept showing a fictional home-to-community-hub journey, landmark steps, an illustrative map and publication checks.
Route studio · Illustrative dataView full size ↗ (opens in a new tab)

05 / 12 · Concept design

AI Agentic Crew

Follow six bounded specialist roles from resident needs and local evidence to a draft support pack. The shared evidence panel makes pending verification and the human publication gate visible.

HavenKrew AI Agentic Crew concept showing six specialist roles, shared evidence, draft media and a publication gate blocked until route verification.
AI Agentic Crew · Illustrative dataView full size ↗ (opens in a new tab)

06 / 12 · Concept design

Multimedia content studio

Prepare large-print text, landmark image cards, spoken prompts and captioned route videos from the same reviewed brief. Keep navigation imagery tied to the actual place and send every pack for approval.

HavenKrew multimedia studio concept showing text, image, audio and video formats, source landmarks, narration, captions and review controls.
Multimedia content studio · Illustrative dataView full size ↗ (opens in a new tab)

07 / 12 · Concept design

Resident guide

Present one familiar landmark and one instruction at a time, with replayable audio and clear help options. The resident experience follows the agreed support plan and companion arrangements.

HavenKrew resident guide concept showing a large garden-pavilion image, a single instruction, audio replay, pause and family contact controls.
Resident guide · Illustrative dataView full size ↗ (opens in a new tab)

08 / 12 · Concept design

Care circle

Bring the day’s plan, shared observations and follow-up requests into one coordinated view. Family, facilitators and programme staff work from information appropriate to their agreed responsibilities.

HavenKrew care-circle concept showing a fictional daily plan, authorised family and facilitator roles, shared observations and follow-up requests.
Care circle · Illustrative dataView full size ↗ (opens in a new tab)

09 / 12 · Concept design

Offline publishing

Package reviewed guidance for print, family tablets, facilitator kits and community displays. Track material versions and distribution so changed routes trigger a clear replacement process.

HavenKrew offline publishing concept showing a versioned pack of printed cards, spoken guidance, route video and a distribution register.
Offline publishing · Illustrative dataView full size ↗ (opens in a new tab)

10 / 12 · Concept design

Review & approval

Review wording, accessibility, source links, route checks and physical placement before release. An unresolved site check pauses publication; revisions return to an authorised reviewer.

HavenKrew review and approval concept showing source linkage, required checks, version history and publication paused for an incomplete site check.
Review & approval · Illustrative dataView full size ↗ (opens in a new tab)

11 / 12 · Concept design

Field trials

Record supervised journey observations such as hesitation, wrong turns, backtracking and assistance. Facilitators can pause or end a session and capture context for later review.

HavenKrew field-trial concept showing an accompanied journey, landmark check-in, facilitator observation controls and session safeguards.
Field trials · Illustrative dataView full size ↗ (opens in a new tab)

12 / 12 · Concept design

Outcomes & learning

Compare observed journeys and different support approaches using agreed measures. Surface places needing attention and prepare an evaluation pack for human interpretation rather than treating a dashboard as proof of effectiveness.

HavenKrew outcomes concept showing route-completion measures, assistance records, interventions under review and a human interpretation step.
Outcomes & learning · Illustrative dataView full size ↗ (opens in a new tab)

For Singapore public-sector and community partners

Start with a place.
Agree the evidence.
Make the next decision clear.

A focused pilot lets agency, estate, care and technology teams assess the concept against a real operating context. Begin with one agreed setting, a small set of journeys and supervised use.

Discuss a HavenKrew pilot ↗
01 / Define

Scope with the people involved.

Agree the use case, site access, participant permissions, responsible teams, support needs and baseline measures before implementation.

02 / Demonstrate

Build a reviewable pilot.

Prepare verified route records, the online workflow and an approved offline pack. Review functionality, accessibility and security before supervised use.

03 / Evaluate

Decide with evidence.

Compare observed journeys, record limitations and maintenance effort, then agree refinements, handover or a separate expansion scope.

Proposed evaluation criteria to agree with the buyer
Evaluation areaEvidence to review
Place accuracyWalked routes, recognisable source photographs, landmark owners and review dates.
Resident suitabilityLanguage, legibility, audio clarity and agreed assistance, reviewed with residents and caregivers.
Publication controlNamed approvers, unresolved-issue blocking and approval tied to the released version.
Offline continuityReadable print samples, downloadable media and a register of where each version is used.
Observed usabilitySupervised route completion, hesitation, wrong turns, assistance and participant feedback.
Operational handoverExport formats, maintenance ownership, training, support and replacement of outdated materials.

Measures are proposed evaluation criteria, not achieved outcomes. No improvement percentage or clinical effectiveness is claimed.

Data and deployment by agreement

Define hosting location, access roles, encryption, retention and deletion, audit records, model-provider processing and permitted data flows. Agency security and data-governance review must inform the implementation scope.

A maintainable service

Agree who owns route checks, translation review, expired materials and offline replacement. Scope service levels, training, exports, source-code rights, operating costs and exit arrangements in the proposal.

Evidence before expansion

Produce a review pack of approved materials, test findings, field observations, limitations and outstanding risks. Expansion follows the buyer’s acceptance and procurement process.

Extending the AdCareAgents approach

Personalised support,
in the community.

AdCareAgents provides the starting approach: resident context, personalised plans, coordinated agents and review. HavenKrew proposes the community layer—verified places, landmark-based routes and guidance that can leave the screen.

Explore AdCareAgents (opens in a new tab)
Resident context

Carry forward only the preferences and agreed support information needed for the journey.

Environment knowledge

Add checked routes, real photographs, recognisable landmarks and ownership for updates.

Reviewed guidance

Prepare consistent online and offline materials, then use supervised feedback to propose revisions.

Integration is proposed. APIs, data mappings, access rights and deployment compatibility require assessment; no completed connector is claimed.

Singapore context

A practical response to
dementia-friendly community living.

HDB’s published work on wayfinding highlights recognisable block identities, clear cues and accessible signage. HavenKrew explores the operational layer around that environment: personalised preparation, consistent materials and structured feedback.

Context references describe public priorities. They do not establish endorsement, a contract award or acceptance of this concept.

Practical questions

A clear concept.
A considered path to a pilot.

What is available to evaluate today?

This page presents the HavenKrew product concept and twelve coordinated screen designs. The sample precinct, residents, routes and dashboard figures are illustrative. A functional demonstration, integration scope, security evidence and field validation would be agreed in a pilot proposal. The designs do not establish a deployed estate service or validated resident outcomes.

Who would use HavenKrew?

The proposed workspace brings together community programme teams, facilitators, caregivers and authorised reviewers. Residents use the approved guidance with the agreed support. An agency or estate operator can evaluate shared materials and supervised journey observations across a defined pilot area.

Does a resident need a smartphone?

The concept includes printed route cards, landmark cards and downloaded audio or video for agreed devices. Online coordination prepares and maintains those materials. Device choices, offline availability, language coverage and accessibility need to be tested with the intended users.

Can AI invent or alter a landmark?

Navigation materials must preserve the identity and appearance of verified real places. The proposed workflow uses approved photographs and route records; an authorised person checks instructions against the location before release. Illustrative images in these concept screens are not instructions for a real journey.

How does this connect to AdCareAgents?

HavenKrew extends the AdCareAgents approach to personalised support, coordinated agent work and human review into community living. The proposed integration connects selected resident context, support plans and review workflows. Data fields, permissions, APIs and deployment compatibility must be assessed before implementation.

Is this an HDB-endorsed or clinically validated system?

No. HavenKrew is an independent wGrow product concept. HDB has not endorsed it. No clinical effectiveness, procurement eligibility, security certification or completed government deployment is claimed. Supervised trials and the relevant agency’s assessment would establish suitability for a specific use.

How would an agency scope procurement?

Begin with an agreed problem, pilot site, participant group and acceptance criteria. The proposal should define deliverables, milestones, data handling, hosting, model providers, security review, accessibility testing, support, intellectual property and exit arrangements. Commercial terms and any procurement route are agreed separately with the buyer.

Bring your estate, care and technology teams together

Make familiar places
part of a connected support plan.

Tell us the setting, the people you support and the journeys that matter. We’ll shape a scoped demonstration, review responsibilities and an evaluation plan with your team.

Discuss HavenKrew