Iceberg above and below the waterline
Iceberg AI Solutions · Malaysia

Most organisations only see 10% of what AI can do.

We build the 90% underneath: the automation infrastructure that enterprise and government rely on.

Trusted by
Government agencies · Enterprise · Public sector
Scroll
Waterline
Voice
Chatbot
Outbound
Data
Workflow
Persona
Vision
Governance
Depth · 5m · Surface layer

AI Receptionist

Never miss an inbound call. Bilingual, calendar-aware, handling intake and triage the moment your line rings. From day one.

Sub-3s response24/7EN + BM
Depth · 40m · Mid layer

AI Chatbot

A single agent across WhatsApp, website, Instagram and Facebook. Trained on your actual product, policies and tone. Consistent on every channel.

OmnichannelTrained on your dataSub-3s reply
Depth · 120m · Deeper layer

AI Outbound Caller

Automated prospecting, qualification and follow-up at scale. Speaks to ten thousand contacts a day without losing tone, judgement, or cadence.

Qualified leadsCRM-nativeScales linearly
Depth · 240m · Base layer

Data & Intelligence

Pipelines, monitoring, executive dashboards. The substrate beneath the visible work. Built for ministerial review and board-level scrutiny.

PipelinesDashboardsCustom ETL
Depth · 400m · Abyss layer

Enterprise Workflow Automation

Multi-agent pipelines that qualify, route, process, and escalate cases end-to-end. Human involvement triggered only at moments of genuine value.

Multi-agent orchestrationCRM + ERP nativeFull audit trail
Depth · 600m · Deep layer

Interactive AI Personas & Avatars

Custom AI personas with bespoke voice, personality, and domain knowledge. Deployed on kiosks, hologram displays, and exhibition installations.

Custom voice + personaKiosk · Hologram · WebBM + EN + dialects
Depth · 800m · Hadal layer

Multimodal Voice + Vision

Process text, voice, and image input within a single conversation thread. Facility check-in by document photo. Claim submission by receipt image. Built on ILMU's native multimodal architecture.

Text · Voice · ImageILMU multimodalSingle thread
Depth · 1000m · Foundational layer

Sovereign Governance Layer

AIGE seven-principle alignment built into every deployment. Per-decision audit logs. Bias monitoring. Human-in-the-loop checkpoints. Quarterly governance review for your compliance officer.

AIGE-alignedPer-decision audit logHuman-in-the-loop
01 · The Malaysian AI imperative

Malaysia has already moved. The question is how deep you go.

i.
2.4M
Malaysian businesses already using AI, with adoption jumping 35% in a single year. The curve is no longer "whether". It is "how fast." AWS Malaysia AI Adoption Study, 2025
ii.
90%
Of Malaysian businesses using AI remain stuck at basic efficiency tasks. Only 10% have reached transformative deployment. The 90% underneath is where we build. AWS Malaysia AI Adoption Study, 2025
iii.
75%
Of Malaysian CEOs believe AI will determine their competitive advantage within the next 24 months. IBM Malaysia, 2025
iv.
RM 1B
Generated by Malaysia’s 140 registered AI solution providers to date. The market is here. Most vendors are not yet building at depth. MDEC, 2024
02 · Our AI provider

Built on ILMU. Malaysia’s own LLM.

ILMU by YTL AI Labs

Every AI system we build (voice agents, chatbots, automation pipelines, avatars) runs on ILMU, the large language model developed by YTL AI Labs. ILMU is Malaysia’s first homegrown multimodal LLM, recognised by the Prime Minister and the Ministry of Digital as a national AI capability.

This means your data stays in Malaysia, your AI speaks Malaysian, and your deployment is compliant with national AI governance standards by default.

LLM & AI provider · ILMU by YTL AI Labs · Hosted in Malaysia
#1
on MalayMMLU benchmark. Outperforms all frontier models in Bahasa Melayu
GPT-4o
performance matched on global benchmarks, hosted entirely within Malaysia
5
languages fluent: English, Bahasa Melayu, Manglish, Kelantanese, Bahasa Istana
🏛️
Data never leaves Malaysia
ILMU is hosted on Malaysian infrastructure. Your data, your citizen records, processed and stored on Malaysian soil.
⚖️
AIGE-aligned by default
Seven-principle National AI Governance and Ethics framework built into every deployment, not retrofitted.
🗣️
Bilingual + dialect native
No translation approximation. ILMU speaks Malaysian (Manglish, Kelantanese, Bahasa Istana) natively, not through patched translation layers.
📋
Procurement compliant
For GLCs, government agencies, and regulated industries, sovereign AI is a procurement requirement, not a preference.
03 · Engagements

Two deployments. Different shapes of scale.

Numbers measured, not modelled. Every engagement handled directly by the founder.

Case 01
AI ReceptionistFederal governmentBilingual · 24/7

A federal agency’s citizen line, answered in Bahasa Malaysia and English. At 3am on a Sunday.

Built a bilingual AI receptionist handling inbound citizen enquiries for a federal agency, with calendar integration for officer appointments and live handover for sensitive cases. Launched in under 14 days; handed over with documentation and a 90-day observation window.

100%
After-hours enquiries answered within three seconds.
Case 02
AI ChatbotMultimodalVoice · Location · Vision

A chatbot that actually understands your customers: voice, location, and images.

Deployed across WhatsApp, website, Instagram and Facebook as a single coherent agent. Trained on the client’s product, policies, and tone, and fluent in every input their customers actually use.

  • Voice messages. Customers send voice notes. The bot transcribes, understands intent, and replies in voice or text, matching the language the customer used.
  • Location tracking. Captures and responds to live location for field service, delivery, and proximity-based workflows.
  • Image recognition. Customers send photos (damaged product, site condition, document, ID) and the bot identifies the issue and routes or actions automatically.
3
Modalities: voice, location, vision in a single thread.
03b · The operational reality

The issue is not effort. It is infrastructure.

The volume, timing, and channel diversity of citizen and customer communication today has outpaced what any team can consistently manage without automated infrastructure underneath.

01

The Speed Gap

A citizen or customer who enquires is in a decision window that closes in minutes. The average organisation responds in over 42 hours. By then, the moment has passed.

02

After-Hours Demand

WhatsApp messages arrive at 11PM. Citizen enquiries come on weekends. Organisations that only respond during staffed hours are unavailable for the majority of hours their stakeholders are active.

03

Leads Never Followed Up

Nearly a third of all leads that express interest receive no follow-up. The marketing spend to generate them has been paid. The return is zero.

04

Service That Doesn't Scale

The same question asked by 200 customers gets 200 different answers from different staff, at different times, with different quality. Volume grows. Consistency does not.

05

Fragmented Channels

WhatsApp, website forms, Instagram DMs, Facebook Messenger, phone, email. Each managed by a different person on a different platform. No unified view of the citizen or customer.

06

Data Without Decision

Operations generate enormous quantities of data: call logs, enquiry patterns, conversion ratios, service gaps. Most of it is never structured, never reviewed, and never converted into a decision.

03c · Industries served

Deployed across eight sectors. Real operations.

01

Government & Public Sector

AI deployments for citizen-facing agencies and government bodies. Voice agent handling of inbound enquiries, officer appointment scheduling, and 24/7 availability across all service channels.

Voice agent · Citizen services · 24/7
02

Federal Foundations & NGOs

Process automation, beneficiary engagement workflows, and structured data acquisition for federal foundations and non-profit organisations supporting national-scale social programmes.

Automation · Beneficiary engagement · Data
03

Healthcare

AI Receptionist and Chatbot for appointment booking, patient enquiry handling, and post-visit follow-up. Available 24/7 across WhatsApp and web with full clinical-language calibration.

24/7 · WhatsApp · Web
04

Financial Services

AI Outbound Caller for life insurance and renewal campaigns. Handles hundreds of simultaneous calls without additional headcount. Bilingual qualification with full audit trail for compliance.

Outbound · Bilingual · Audit
05

Hospitality & Property

Bahasa Melayu and Bahasa Indonesia localisation for international AI voice partners serving hotel clients. Property developer deployments for sales qualification and post-handover service.

BM · BI · Sales qualification
06

Fitness & Wellness

AI Outbound Caller for member re-engagement at subscription end, automated feedback collection, and renewal outreach. No manual follow-up required from trainers or front desk staff.

Re-engagement · Renewal · Feedback
07

Oil & Gas

AI voice agents for field crew scheduling, safety briefing acknowledgement, and permit-to-work workflows. Automated incident reporting and compliance documentation across offshore and onshore operations.

Field ops · Compliance · Safety
08

Legal & Law Firms

Client intake automation, matter triage, and appointment scheduling across WhatsApp and web. AI-assisted document review routing and deadline reminder workflows for litigation and conveyancing practices.

Intake · Triage · Compliance
04 · How we engage

Four stages. No decks until the system works.

We don't pitch architecture diagrams. We build the thing, show it to you running against your own data, then contract.

Stage I

Discovery & audit

Two weeks inside your operation. We map every inbound channel, every handoff, every drop. Only then do we propose a system.

Stage II

System design

Architecture, data model, integration plan. Shared openly; reviewed with your engineering and compliance leads before build.

Stage III

Deployment

Production rollout in under 14 days. KPIs contracted in writing. Observed for 90 days with weekly reporting to the accountable executive.

Stage IV

Ongoing optimisation

Monthly model reviews, new surface area, and calibration against quarterly business targets. We stay embedded.

04b · How we compare

Most vendors are building on the surface.

Capability
Typical Malaysian AI Vendor
Iceberg AI Solutions
LLM Backbone
Foreign model (GPT, Gemini, Claude) via API
ILMU sovereign LLM via YTL AI Labs collaboration
Data Hosting
US, Singapore, or EU data centres
YTL AI Cloud, Malaysia. Data does not leave the country.
Bahasa Melayu Performance
Approximate; trained on translated data
Best-in-class. ILMU is #1 on MalayMMLU, fluent in Manglish, Kelantanese, Bahasa Istana
AIGE Compliance
Not addressed
AIGE seven-principle alignment built into every deployment by design
Engagement Model
Project handover, then on your own
Managed end-to-end with monthly KPI review for life of contract
Deployment Time
3 to 6 months typical
Under 14 days from discovery to live operation
Voice + Chat + Workflow
Single channel typically
Unified multi-agent architecture across voice, chat, automation, and data
05 · Why us

Depth over display.

Six principles we hold to. Every engagement is evaluated against them, including by us.

P · 01

Depth over display

The architecture matters more than the interface. We build what the customer will never see, because it's the part that makes the visible part work.

P · 02

Measured outcomes

KPIs written into the contract. Hours saved, leads qualified, response time. All stated up front and observed for 90 days after launch.

P · 03

Long-term thinking

We don't optimise for the first quarter's dashboard. Systems are built to be extended by whoever inherits them, including your future engineers.

P · 04

Local context, global standard

Bilingual by default. Malaysian regulatory, operational, and cultural grounding. Engineered to standards you'd recognise from London or Singapore.

P · 05

Transparency as practice

Weekly status, visible dashboards, open access to the codebase and the model cards. You see the work while it is being done.

P · 06

Permanence by design

No vendor lock. Systems handed over with documentation, runbooks, and the keys. Your organisation owns what we built, not us.

06 · The team

A focused founding team. No intermediaries.

Every client engagement is handled directly by the people who built the system. Enterprise business strategy, full-stack technical architecture, and hands-on AI engineering. In one team.

CEO · Business
Firdhaus Othman
Co-Founder & CEO
Enterprise engagement, commercial structuring, and government relations. Leads all client partnerships and business development.
COO · Operations & Sales
Avvienash Yogaprash
Co-Founder & COO
Leads operations, marketing, and sales. Specialises in voice agent engineering and go-to-market execution, and manages the end-to-end delivery pipeline from client onboarding through to live deployment.
CTO · AI Engineering
Navien Bala
Co-Founder & CTO
Full-stack architect and lead developer across all backend and frontend systems. Responsible for product engineering, AI integration, and the technical architecture that runs every client deployment.
07 · Next step

The first step is a 30-minute discovery session. At no cost, no obligation. If there is a case for AI infrastructure, we will present it. If there is not, we will tell you that directly.

Tweaks