Current brand activation technology fails at the moment of truth — the personal conversation.
FAQ-tree bots answer pre-scripted questions but cannot hold a genuine consultative dialogue. The moment a user deviates from expected patterns, the experience breaks entirely. Average disengagement occurs within 30 seconds — and the brand loses the interaction without generating any lasting value.
Thousands of brand interactions occur at each activation with no structured record. No consumer need patterns are captured, no health data is tracked, no competitive intelligence is built. Every event restarts from zero — wasting enormous consumer research potential that was right there in each conversation.
Human promoters are constrained by headcount, training consistency, and working hours. Peak times create queues. Off-peak times leave staff idle. Messaging quality varies between individuals and degrades through the day. No system accumulates institutional knowledge across thousands of interactions.
Symptoms, chronic conditions, medication history — these are private matters. Users in a public event setting will not share them with strangers or on a screen others can see. The most valuable consultation data never gets collected because the environment makes disclosure feel fundamentally unsafe.
Every response is grounded in verified clinical data and logical reasoning — never hallucinated, never guessed. Justification is traceable back to the source data.
Contextual understanding that persists across sessions. Neo learns each user's history, conditions, and preferences — building a progressively richer profile with every interaction.
Knowing when not to recommend. Recognising the limit of what AI should handle alone — and dispatching human expertise before a situation escalates beyond safe containment.
Amplifying human capability — not replacing it. JIWA handles volume and consistency so your people can focus on the relationships and exceptions that genuinely require human judgment.
We do not build robots. We do not replace people. We build AI that behaves as humanly as possible — empathetic, contextual, and wise enough to know its own limits.
This philosophy is not a tagline. It is a design constraint that governs every feature decision in JIWA — from how Neo speaks, to how it handles emergencies, to why it never oversells or pressures a user into a purchase they should not make.
Metahuman AI Brand Ambassador
Neo is not a cartoon avatar or a generic animation. Neo is a photorealistic digital human rendered in real-time using Unreal Engine's Metahuman technology — the same production pipeline used in feature film and AAA game development.
Displayed on large-format digital signage, Neo creates a commanding presence that earns genuine attention and trust — bridging the gap between brand visual identity and a truly human interaction experience.
Every spoken word is accompanied by precisely synchronised lip movement, natural blink patterns, subtle facial microexpressions, and breathing cadence. The result reads as a living presence — not a video loop or a face with animated lips applied.
Neo modulates its delivery based on the live conversation state — warmer and slower when a user describes health concerns, more precise and confident when delivering a recommendation. Tone is not scripted; it responds dynamically to what the user is expressing.
Native Bahasa Indonesia and English — not machine-translated. Sentence construction, appropriate honorifics, and conversational pacing are calibrated for each language. Extension to regional dialects is on the product roadmap.
Every interaction follows the brand messaging guidelines exactly. No improvisation, no off-script moments, no performance degradation through the day. The ten-thousandth user receives the same quality of experience as the first.
Six steps from stranger to served — the complete user journey in a single event interaction.
Most kiosks use a shared touchscreen. Users feel watched. They share less. The most valuable data stays hidden.
JIWA routes the conversation through the user's own phone. This is not a technical convenience — it is a calculated psychological design choice with measurable impact on data quality.
When a person types on their own device, they are operating in their private sphere. The interaction feels like messaging a trusted advisor — not completing a public form in front of a crowd.
Users disclose sensitive health information because no one else can see their screen. No adjacent users, no event staff, no passersby. Psychological safety is the necessary precondition for honest disclosure — and honest disclosure is the precondition for a useful recommendation.
A user who feels safe will share that they have hypertension, that they are pregnant, or that they are currently on a specific medication. Without this context, recommendations are generic. With it, they are genuinely personalised and medically appropriate.
Event halls carry substantial ambient noise. Voice input from a personal device held close to the mouth delivers significantly cleaner audio to the STT engine than a shared kiosk microphone mounted at fixed head height in a crowded space.
A mobile browser chat interface is the most familiar digital interaction format in Indonesia today. Users need no instruction — they begin typing immediately, as they would with any messaging application they use daily.
Neo addresses the user by name from registration data and establishes conversational context — first visit or returning session. It sets a warm, unhurried tone that signals there is no rush, encouraging the user to describe their situation in their own words without feeling they are filling out a form.
The user describes their concern in natural language — no dropdown menus, no structured forms. Neo extracts clinical signals from free text using natural language processing, then reflects them back for confirmation before proceeding. Nothing is assumed; everything is verified.
Neo asks targeted follow-up questions: known drug allergies, existing chronic conditions such as hypertension or diabetes, pregnancy or breastfeeding status, and current medications. These answers directly constrain the recommendation algorithm — excluding unsafe options before any product is considered.
The user profile is matched against the 150-product Halodoc database. Contraindicated products are filtered out entirely. Dosage options are adjusted for declared age group. The algorithm converges to a single safest, most clinically appropriate match for this specific user's profile.
Neo presents exactly one product — with the specific reason it was selected for this user, the recommended dosage, what to watch for during use, and when to stop. A medical disclaimer is displayed at both session start and conclusion, establishing the informational scope of Neo's role.
If Neo detects symptom patterns consistent with a medical emergency — chest pain, breathing difficulty, anaphylaxis signs — it immediately dispatches a WhatsApp notification via Unreal Engine to on-site medical staff, including the user's instance location. Neo keeps the user calm while help is en route.
Each product carries a structured list of clinical indications — the specific symptoms and conditions it is appropriate for, ranked by evidence strength. Neo maps user-reported symptoms against this taxonomy using semantic matching rather than keyword search.
Products that are clinically unsafe for specific patient profiles — hypertensive users, pregnant users, those with hepatic or renal conditions — are automatically excluded before the recommendation pool is generated. Exclusion happens silently; the user is never exposed to an unsafe option to decline.
Complete compositional data enables cross-checking for ingredient-level allergies and for polypharmacy interactions with user-disclosed medications — a depth of clinical screening that human promoters rarely have the training or time to perform at the point of sale.
Known side effect profiles and drug interaction data are disclosed in the recommendation output, giving users the information they need to make an informed decision — and reducing brand liability from uninformed or inappropriate product selections driven by enthusiastic but undertrained staff.
Dosing guidance is applied dynamically based on declared age group — paediatric, adult, and elderly thresholds are applied where product data specifies different dosing requirements. Recommendations are never one-size-fits-all; they are calibrated to the individual profile.
Decision research shows that presenting multiple options increases cognitive load and reduces the probability of any purchase being made — by up to 40%. A single clear recommendation with clear rationale drives action. It also eliminates the risk of a user selecting an inappropriate option from a list they may not fully understand.
When AI recognises a crisis, human help is dispatched automatically — before the user needs to ask for it.
Neo detects trigger pattern through conversation NLP layer
Unreal Engine fires WhatsApp Business API notification
On-site medical staff receive alert with instance location ID
Staff navigate to the correct signage unit within the venue
Neo calmly informs user that help is on the way and keeps them engaged
No other retail chatbot or kiosk system has an active clinical safety layer. For a brand in health and wellness, deploying JIWA demonstrates that user wellbeing is taken seriously at an operational level — not just in advertising copy. This is a defensible, meaningful differentiator that competitor systems cannot quickly replicate.
Unlike every other event chatbot, JIWA remembers each user — across sessions, across days, across reconnections.
Complete conversation history is stored in the user's browser session on their own device. This persists after the browser is closed, enabling seamless resumption the following day — or several days later — without re-registration or re-entering health context that was already provided.
Both storage layers synchronise on every interaction. If the network drops mid-session, the device holds the last known state. On reconnect, the instance worker reconciles and continues the conversation from the exact point of interruption — with no context lost.
A mirror of every session is maintained in the dedicated instance worker process. This enables continuity from any device the user reconnects from, and provides a server-authoritative record that synchronises continuously with the client's analytics database.
Neo knows the user's name, prior symptoms, and previous recommendation on Day 2 of the event. The interaction feels like a follow-up with a personal advisor — not a reset to zero.
A dropped connection at any point does not lose the conversation. Reconnect and continue from exactly where the session paused — no frustration, no repeated disclosure of already-provided information.
Session persistence means the relationship extends beyond the event. Follow-up consultations are possible — turning a single event touchpoint into an ongoing brand engagement channel.
Persistent sessions enable longitudinal data — repeat consultations, changing symptom patterns, repurchase signals — providing research depth unavailable from single-session interactions.
Each instance is a fully independent process with its own memory allocation, conversation state, and storage access. A heavy query, network lag, or unusual input in one session has zero effect on any concurrent session. There is no shared application state that could create contention or data leakage between users.
Adding capacity for a larger event crowd is a configuration change, not an engineering project. The system supports N concurrent users where N is bounded only by the client's network infrastructure — not by any architectural ceiling in JIWA. Scaling up between event days requires no system downtime or code changes.
If one instance encounters an error condition, it fails independently. There are no cascading failures, no system-wide outages triggered by a single edge case input. Each instance can restart cleanly without any other user experiencing an interruption. The system degrades gracefully rather than catastrophically.
Session data is sandboxed at the process level. There is no mechanism — intentional or accidental — by which one user's health information can be accessed by, associated with, or commingled with another user's session record. Isolation is an architectural property, not a policy statement that relies on correct runtime behaviour.
Product recommendation frequency plotted against time, with segmentation by hour, day, and week. Identifies organic demand spikes that correlate with foot traffic patterns — enabling brands to align promotional content and staff deployment with real consumer interest curves rather than schedule assumptions.
Session data aggregated by product category — cold and flu, vitamins and supplements, digestive health, topical treatments, and others. Reveals whether consumer need distribution matches the brand's marketing investment allocation across categories — often showing significant misalignment with assumed priorities.
Select any product to see: total recommendation count, the most frequent symptom clusters that triggered it, the demographic profile of users who received it, and trend direction versus prior period. This is SKU-level demand intelligence that is impossible to generate from point-of-sale data alone.
All data from the preceding seven days exported as a formatted Excel workbook ready for management review or input into existing BI pipelines. Includes raw session data, aggregated summaries, and product performance rankings in standardised column structures compatible with common analytics platforms.
Analytics data is only valuable if someone interprets it correctly and acts on it quickly.
Most brand managers do not have time to parse raw dashboards between events. The AI Summary converts accumulated session data into a professionally structured business intelligence report in seconds — without a data analyst, without manual work, without delay.
The report is formatted for executive consumption — concise enough to share in a message to a brand director, or attach to a morning debrief deck the same day the event closes.
Output is branded with the client's logo and Neo Pexels attribution. Exportable as PDF. Structured for direct board-level sharing without further editing.
"We do not touch your data. It flows directly to you."
User registration data — names, phone numbers, session transcripts, and recommendation histories — is written directly to the client's own database infrastructure. In the P&G deployment, data flows exclusively to P&G's own Google Drive environment. Neo Pexels has no read or write access to this data after transmission is complete.
Registration cannot be completed without the user actively ticking an explicit consent acknowledgement. The consent statement clearly specifies what data is collected (name, phone number), the purpose (personalised consultation and future promotional communications), and who will hold it (the brand). No ambiguous language, no pre-ticked defaults, no buried terms.
Individual user sessions are isolated at the process level. There is no mechanism — technical or accidental — by which one user's health information can become visible to or associated with another user's record. Isolation is enforced architecturally, not managed by access control policy alone — the distinction matters.
What is collected, the format it is stored in, and the destination it is transmitted to are disclosed to both the client and the end user. There are no silent background data collections, no analytics beaconing to third-party services, and no advertising network integrations in the JIWA technology stack.
A single JIWA deployment handles the consultation volume of multiple human promoters simultaneously — without overtime, without performance variance, and without fatigue degradation across a multi-day event. Remaining human staff can be redeployed to genuinely high-value interactions requiring personal judgment, relationship development, or complex problem resolution.
Every QR registration creates a consented marketing lead — name, verified phone number, health interest profile, and preferred product category. This structured, high-quality CRM dataset is built automatically during the event, at no additional collection cost, and is immediately available for post-event follow-up campaigns targeting verified consumer health interests.
Aggregated session data from thousands of interactions reveals the actual distribution of consumer health concerns in the target market — mapped by demographic segment, by geography, and by time of day. This is primary research data that would cost significantly more through traditional market research methodologies, updated continuously throughout the event at no incremental research cost.
| Capability | JIWA by Neo Pexels | Standard Chatbot | Human Promoter |
|---|---|---|---|
| Photorealistic presence | Unreal Engine Metahuman | Text or basic 2D avatar | Real person (variable) |
| 24 / 7 availability | Always on, no degradation | Always on | Shift-limited |
| Privacy architecture | Personal device controller | Shared public screen | Face-to-face — none |
| Clinical data layer | Halodoc-sourced, structured | None or manual FAQ tree | Training-dependent |
| Emergency escalation | Automated WhatsApp dispatch | Not available | Manual, inconsistent |
| Session memory | Multi-day persistence | Single session only | None |
| Real-time analytics | Live dashboard + AI report | Basic logs only | Not available |
| Data ownership | 100% client-controlled | Provider retains data | Internal only |
| Concurrent scalability | Unlimited isolated instances | Limited by server tier | Linear with headcount |
| Cost at event scale | Fixed deployment cost | Low — limited capability | Grows with headcount |
Product consultation at event activations and in-store deployments. Consumer health need capture at the point of engagement. SKU-level demand intelligence generated passively from consultation data. Lead generation with health-interest profiling for targeted campaign follow-up. Competitive differentiation for brands in commoditised OTC and supplement categories where product similarity makes experiential quality the primary decision driver.
OTC medication guidance in clinic waiting areas and pharmacy retail environments. Patient education on supplement protocols and preventive care product lines. Condition management advisory for chronic disease product ranges. Medical congress deployments for HCP product awareness. Prescription pathway navigation where clinical constraint compliance is architecturally enforced rather than relying on staff knowledge.
Concierge augmentation at hotels and resorts — recommendations, venue navigation, service requests in multiple languages simultaneously. Menu advisory at high-volume establishments with complete dietary restriction handling. Guest loyalty programme engagement and contextual upselling. Multilingual reception at international venues serving demographically diverse guest profiles from across Southeast Asia and beyond.
Branch engagement augmentation for product advisory — savings instruments, insurance products, and investment vehicles explained in accessible language. Digital onboarding companion for new account holders navigating complex product ranges. Financial literacy education deployments. Customer service first-contact handling with seamless escalation to human relationship managers for high-value or complex cases.
Ready to deploy JIWA at your next event?