A 60-minute alignment session for business and technical stakeholders. This page summarises the full proposal — what we will build, how it connects to your existing systems, and what we need from SPH to begin.
Prepared for
SPH Media Limited
Prepared by
QUICKSALES.AI
Date
8 July 2026
Validity
90 days
Delivery options
Option A / B
Today's session
Project kick-off
Private & Confidential — for authorised SPH Media and QUICKSALES.AI participants only
The proposal in 60 seconds
QUICKSALES.AI will deliver a managed WhatsApp orchestration platform for SPH Media — across The Straits Times, Lianhe Zaobao, The Business Times, magazines, and student titles.
Subscribers message SPH on WhatsApp. AI answers routine questions in four languages. When a human is needed, the conversation hands off to a ServiceNow agent — no new agent workspace. Commerce triggers payment and delivery notifications. OneTrust governs consent. MoEngage powers Phase 2 marketing.
SPH does not buy servers or install software. Teams use ServiceNow (agents), a web Admin Portal (supervisors & marketing), and WhatsApp (subscribers). Everything runs on Singapore cloud infrastructure, managed by QUICKSALES.AI.
Two ways to read this proposal
Everyone in today's meeting gets value from this page. Technical teams can drill into the linked sections afterwards.
Business & operations
What matters to you
Subscribers get help on WhatsApp — no app download
Agents stay in ServiceNow; supervisors get a new web dashboard
Full API specs, sequence diagrams, and compliance matrix in Sections 6–14
Who uses what — three interfaces
No one learns a system they don't need. Each group keeps their familiar tools.
Subscribers
Message SPH on WhatsApp. Ask questions, manage subscriptions, receive payment and delivery alerts — in EN, ZH, MS, or TA.
Uses: WhatsApp
CS agents
Continue in ServiceNow Agent Workspace. WhatsApp appears as another channel with AI handoff summary and redacted transcript.
Uses: ServiceNow CSM
Supervisors & marketing
New QUICKSALES.AI Admin Portal — journeys, templates, analytics, consent controls, and WABA health monitoring via browser.
Uses: Admin Portal (web)
Technology stack at a glance
SPH operations teams never touch these layers — they are managed by QUICKSALES.AI. Included here so IT can validate architecture in today's session.
In plain English
SubscribersWhatsApp on any phone
AgentsServiceNow (unchanged)
SupervisorsWeb browser → Admin Portal
HostingSingapore cloud — no SPH hardware
DataMaster records stay in SPH systems
Layer
Technology
Managed by
Why?
Admin Portal
React
QUICKSALES.AI
Browser UI for supervisors — journeys, templates, analytics, and WABA health. No software to install at SPH.
API / orchestration
Node.js/FastAPI
QUICKSALES.AI
The routing hub — WhatsApp webhooks, PCI redaction, Commerce events, and handoff to ServiceNow all run here.
AI engine
GPT-4o/Claude+ NLU
QUICKSALES.AI
Primary WhatsApp bot — classifies intent, runs subscriber journeys, and drafts answers. ServiceNow VA handles live-agent handoff only.
Knowledge RAG
Pinecone/pgvector
QUICKSALES.AI
KB articles stay in ServiceNow (and Google). This is a synced AI search index — not a second knowledge base. Needed because our bot runs outside ServiceNow and must search SN + Google together.
Session cache
Redis(TTL 24–72 hrs)
QUICKSALES.AI
Short-lived routing tokens only (24–72 hrs). No message bodies or subscriber PII stored on our platform.
Config DB
PostgreSQL
QUICKSALES.AI
Stores journey rules, template metadata, and admin settings — operational config only, never customer records.
Hosting
AWS/AzureSingapore
QUICKSALES.AI
Singapore region for PDPA-aligned residency. Fully managed cloud — SPH does not purchase or host hardware.
Integrations
Configured jointly
Six connectors to SPH systems you already own — Meta (WhatsApp), ServiceNow, Commerce, OneTrust, Google KB, and MoEngage.
Common IT questions about this stack
Why not use ServiceNow's built-in knowledge search?
ServiceNow has its own RAG inside Virtual Agent (AI Search + Now Assist). Our WhatsApp bot is the primary bot on QUICKSALES.AI, not SN VA — so we need a search index next to our LLM. ServiceNow remains where articles are authored and where agents work.
Why Pinecone or pgvector — isn't that duplicating the KB?
No. ServiceNow and Google remain the source of truth. The vector DB holds synced article embeddings for fast semantic search at answer time — KB article text only, never customer chats. Pinecone and pgvector are two implementation options (managed vs. PostgreSQL extension).
What data lives on QUICKSALES.AI vs SPH systems?
Conversations, cases, and subscriber records stay in ServiceNow and Commerce. Our platform holds routing cache (Redis), config (PostgreSQL), and KB embeddings (vector DB) — plus aggregated analytics with no message text.
AI answers common questions 24/7 in four languages. Agents step in via ServiceNow. Subscribers receive payment, delivery, and subscription updates on WhatsApp.
Phase 2 — growth
Marketing & retention
Consent-based campaigns through MoEngage — win-back offers, renewal reminders, and lifecycle messages for subscribers who opt in via OneTrust.
Commercial options
Both deliver a working pilot. Option B adds full RFP coverage and Phase 2 foundation. Final pricing after discovery workshop.
Option
Scope
Timeline
Effort
Option A — Essential
8 priority journeys, core integrations, pilot-ready support & notifications
~5 months
~128 mandays
Option B — Professional
Full Phase 1 + Phase 2 foundation, all RFP requirements, 22 journeys