Designing AI Chatbots That Actually Drive Sales

Many enterprises in Southeast Asia have deployed an AI chatbot, but only a handful can clearly answer: “How much revenue or cost efficiency does it really generate?” In most cases, bots end up as glorified FAQ widgets—good for ticking an innovation box, weak on business impact.

This article looks at a strategic approach inspired by Janice Tjen—who works at the intersection of product, CX, and data—to move chatbots from a cost item into a genuine growth engine. The focus: how to connect conversation design, data, and messaging channels like WhatsApp Business API and Omnichannel so that business metrics actually move.

From Cost Center to Growth Engine: Resetting the Chatbot Mindset

The first mistake in many enterprise chatbot projects is treating them purely as a way to reduce contact center volume. When that’s the main lens, chatbot design over-indexes on complaint handling and under-invests in:

• identifying and warming up leads,
• helping customers complete transactions,
• nudging up-sell and cross-sell,
• accelerating post-sales processes.

Janice’s approach starts from a simple question: “At which point in the customer journey would a 1% uplift in conversion have the biggest impact on revenue?”

From there, the chatbot is designed not as an isolated channel but as an orchestrator of conversations across SMS, WhatsApp Official API, email, and other omnichannel touchpoints—with clear commercial objectives.

Map the Journey Before You Write a Single Line of Dialogue

Before choosing AI models or vendors, the priority is to map the journey with brutal clarity. Across consumer-facing industries and fintech, a typical pattern looks like this:

1. Acquisition: prospects come in via ads, referrals, or offline touchpoints.
2. Activation: form fill, phone verification via SMS OTP, onboarding steps.
3. First Transaction: purchase, top-up, booking, or application submission.
4. Repeat & Retention: repeat use, subscription renewal, feature adoption.

At each stage, the chatbot should play a measurable role. For example:

• Addressing objections on product pages to increase add-to-cart and sign-up rates.
• Reducing friction around OTP, KYC, and document upload flows.
• Driving repeat purchases with relevant product or feature recommendations on WhatsApp.

This is where the choice of messaging channel matters. Integrating an AI chatbot with WhatsApp Business API via SMSMasking.id (https://smsmasking.id/id/whatsapp/waba) enables real-time conversations in the channel your customers already use daily.

A Four-Layer Strategy: How to Architect Chatbots for Outcomes

Instead of bolting AI onto existing scripts, Janice’s approach breaks chatbot strategy into four mutually reinforcing layers:

1. Business Intents, Not Just Conversation Intents

Most chatbot projects stop at narrow conversation intents like “check balance”, “track order”, or “reset password”. What they miss are broader business intents like:

• Lead intent: the user is curious but not ready to buy.
• Purchase intent: the user has a clear product in mind and needs help to complete the transaction.
• Risk intent: the user shows signs of churn risk or is about to escalate publicly.

Once business intents are defined, the bot can route conversations to the most valuable next action: self-service, contextual offers, or fast-track escalation to a human agent.

2. Content & Knowledge: Where Data Meets Brand Voice

An effective AI customer service chatbot is not only technically accurate; it also speaks in a voice that reflects the brand and aligns with regulatory constraints. This layer covers:

• Base knowledge: FAQs, policy explanations, product details.
• Conversation playbooks: structured flows for objection handling, risk education, upsell, and more.
• Guardrails: what the bot cannot say or do, particularly in regulated sectors like finance and healthcare.

Using an Omnichannel platform from SMSMasking.id, the same knowledge base can be reused across channels: web chat, WhatsApp, and even SMS for less digitally mature user segments.

3. Channel Orchestration: From Web Chat to WhatsApp and SMS

One of Janice’s recurring points is that customers do not think in terms of channels. They just want their issue resolved with minimal effort. Channel orchestration is critical:

• A visitor starts a chat on your website while browsing a product.
• Without losing context, the conversation continues seamlessly on WhatsApp after the visitor leaves the site.
• If data coverage is poor or WhatsApp remains unopened, a short reminder can be pushed via SMS Masking to ensure critical messages are still delivered.

With Omnichannel from SMSMasking.id (https://smsmasking.id/id/omnichannel), enterprises can manage all these interactions in a single workspace. That means:

• conversation history stays intact across channels,
• sales and support teams don’t have to switch tools,
• the chatbot can recognize the same customer regardless of entry point.

4. Data & Experimentation: Turning the Bot into a Conversation Lab

This is where the chatbot becomes a growth engine instead of a static automation layer. Every dialogue is an experiment:

• At the top of the funnel, does “Can I help you choose a plan?” outperform “How can I help you today?” in terms of lead capture?
• Does a WhatsApp reminder plus an SMS backup improve payment completion compared with WhatsApp alone?
• What is the repeat purchase rate for users who receive AI-driven recommendations vs those who don’t?

These metrics feed into your analytics stack, while the execution layer is handled by a messaging partner such as SMSMasking.id—covering WhatsApp Business API, SMS Masking, Voice OTP, and Omnichannel.

High-Impact Use Cases: Where Chatbots Should Start

From a commercial standpoint, your AI chatbot should be pointed at high-impact scenarios first—not “answer everything”. Below are three priority use cases Janice often recommends for the first phase.

1. Sales Assistant on WhatsApp

Instead of waiting for customers to find a contact form, invite them into a WhatsApp conversation: from ads, QR codes in-store, or short links on social media.

Using WhatsApp Official API by SMSMasking.id, enterprises can:

• deploy a chatbot that answers product and pricing questions 24/7,
• send dynamic catalogues and in-chat carts,
• route high-value conversations to human sales when appropriate.

Key metrics to track include:

• Chat-to-lead rate: percentage of chats that yield lead data (name, contact, product interest).
• Lead-to-order rate: how many leads convert to actual orders.
• Average order value for chatbot-assisted vs non-assisted transactions.

2. Reducing Onboarding & Verification Friction

In fintech, insurtech, education, and B2B SaaS, a large number of prospects drop off during onboarding: forms are too long, verification is confusing, or network conditions are poor.

Janice’s recommended playbook:

• Use a WhatsApp chatbot to turn long forms into a sequence of short, conversational questions.
• Automatically trigger SMS OTP or Voice OTP when a user reaches critical verification steps.
• Provide real-time guidance for sensitive fields (income, tax ID, business details) to build trust and reduce hesitation.

By leveraging local-direct SMS infrastructure, such as SMSMasking.id Local Direct SMS, you can ensure OTPs and notifications arrive quickly and reliably—without forcing users to juggle too many apps.

3. Proactive Retention and Soft Upsell

Instead of waiting for angry reviews on social media, an AI chatbot can be used to check in and intervene early:

• After a purchase or activation, the bot sends a short WhatsApp message: “How was your experience?” with quick-reply options.
• Positive responses trigger tailored recommendations or loyalty offers.
• Negative responses are fast-tracked to human agents, along with full conversation context.

Omnichannel capabilities become crucial here. Customers who originally bought via marketplace or offline can be gradually migrated into owned channels (WhatsApp or SMS) for more controlled, personalized post-purchase journeys.

Setting the Right KPIs: Balancing Conversion and Efficiency

Many chatbot initiatives report success only in terms of “reduced ticket volume”. Janice argues that a more balanced KPI framework is needed—one that combines conversion, efficiency, experience, and AI quality.

1. Conversion Metrics

• Lead capture rate from chatbot interactions.
• Assisted conversion rate: conversion where the chatbot played a role vs the baseline.
• Average order value when AI recommendations are shown vs not shown.

2. Efficiency Metrics

• Deflection rate: share of inquiries fully resolved by the bot.
• Average handling time for agents—does it fall because the bot pre-collects context?
• Tickets per customer: does better education via chatbot reduce repeated contacts?

3. Customer Experience Metrics

• Bot-specific CSAT after each conversation.
• Drop-off rate in mid-conversation—at which nodes do users abandon?
• NPS for journeys that involve the bot vs those that don’t.

4. AI Quality Metrics

• Frequency of “I don’t understand” responses.
• Escalation rates for topics that should be handled by the bot.
• Cycle time to improve bot knowledge based on feedback.

When paired with messaging infrastructure such as WhatsApp Business API and Omnichannel, these metrics can be analysed across channels instead of in silos.

A Practical Architecture for Lean Product Teams

Many teams delay chatbot projects because they fear technical complexity. Janice recommends a pragmatic starting point: build a minimum lovable product (MLP)—limited in scope but clearly useful for both users and the business.

Core Components to Prepare

1. Messaging Platform
   Work with a provider that natively supports multiple channels (SMS, WhatsApp Official, omnichannel). For instance, SMSMasking.id offers:
   • WhatsApp Business API with transactional notifications, product catalog, and approved message templates.
   • SMS Masking and Voice OTP for verification and reminders.
   • An Omnichannel dashboard that consolidates conversations.
2. AI & NLP Engine
   This can be:
   • a cloud-based conversational AI platform, or
   • an open-source framework, if your tech team prefers more control.
3. Orchestration & Integration Layer
   Responsible for connecting the chatbot to:
   • CRM/CDP for customer profiles,
   • transaction systems (e-commerce, booking, core apps),
   • analytics and reporting.
4. Cross-Functional Core Team
   Ideally small but empowered, consisting of:
   • a product owner who owns the journey and KPIs,
   • a CX/CS lead who understands customer language,
   • a data/engineering lead for integrations and experimentation.

Janice Tjen’s Lens: Empathy + Data in Every Exchange

One of the strongest themes in Janice’s work is the deliberate balance between empathy and data. A purely data-driven chatbot risks sounding robotic and transactional. A purely empathetic one, without data, may feel nice but fail to move business metrics.

In practice, this means:

• Every standard line—greeting, clarification, closing—is tested on two dimensions: does it help the user feel understood, and does it guide them toward a clear next step?
• Responses are concise yet informative, offering obvious next options instead of open-ended dead ends.
• Conversation logs are analysed for emotional signals: keywords and patterns that reflect confusion, frustration, or delight; these insights shape future scripts and AI fine-tuning.

By framing the chatbot as an extension of your service and sales teams—not a replacement—you can build trust that translates into loyalty and higher lifetime value.

Risks to Avoid: Where Chatbots Can Backfire

Rushed or poorly framed chatbot projects can damage brand perception. Key pitfalls to watch for:

1. Dead Ends and No Way Out

Users stuck in loops of “sorry, I didn’t understand” with no option to escalate will quickly abandon your bot and possibly your brand. To prevent this:

• Ensure every flow has a clear path to a human agent, especially for high-risk topics (finance, health, serious complaints).
• Use Omnichannel tooling so agents can seamlessly take over within the same channel and see full context.

2. Data Misuse and Regulatory Blind Spots

In regulated sectors, a bot that carelessly asks for or discloses sensitive data can create compliance exposure. You should:

• separate sensitive data from general chat logs with strong access controls,
• design the bot to comply with local regulations (e.g., OJK/BI guidance in Indonesia),
• work with messaging partners like SMSMasking.id that adhere to local security and regulatory standards.

3. Over-Automation: Trying to Replace Everyone

Enterprises that try to “automate everything” with bots often see CSAT and NPS drop. The healthier framing is:

• Let the bot handle repetitive, low-complexity queries.
• Use the bot to gather initial context and triage complex issues.
• Free human agents to handle nuanced conversations that require judgment and relationship-building.

A 90-Day Roadmap: From Idea to Business Impact

For enterprises that want to move quickly but thoughtfully, a 90-day plan is realistic.

First 30 Days: Foundation & Use Case Prioritization

• Map your customer journey and biggest friction points.
• Select 1–2 high-impact use cases (e.g., WhatsApp sales assistant plus SMS-based OTP reminders).
• Onboard a messaging partner like SMSMasking.id for WhatsApp Business API, SMS Masking, and Omnichannel.
• Draft initial conversation playbooks and escalation rules.

Second 30 Days: Build, Light Integrations, Soft Launch

• Build the first version of the chatbot with clearly scoped capabilities.
• Integrate it with WhatsApp Official API and, if needed, SMS for backup notifications.
• Run internal and small-scale customer pilots.
• Start collecting baseline data: volume, intent distribution, early CSAT.

Final 30 Days: Data-Driven Optimisation & Gradual Expansion

• Review conversations to fix confusing responses and refine flows.
• Introduce a second use case (e.g., proactive retention or soft upsell).
• Connect conversation data with your CRM for better segmentation.
• Set up a shared KPI dashboard for product, sales, and support teams.

Conclusion: AI Chatbots as Bridges, Not Replacements

Ultimately, the success of your AI chatbot strategy is not determined by the sophistication of the underlying model, but by how well it bridges customer needs with business goals.

Janice Tjen’s lens positions the chatbot as:

• a connector between data, channels, and people,
• a continuous experimentation engine for conversion uplift,
• a driver of efficiency that doesn’t sacrifice empathy.

With robust messaging infrastructure—spanning WhatsApp Business API, local-direct SMS Masking, and Omnichannel—Southeast Asian enterprises can start small, prove value within weeks, and then scale chatbots into a core pillar of their digital growth playbook.