How to Build an AI Sales Chatbot for WhatsApp: Step-by-Step

WhatsApp is not just a messaging app. It is the main way people communicate with businesses across the Middle East, South Asia, Latin America, and large parts of Europe and Africa. Here is why it should be a core part of your sales setup in 2026.

**Open rates are huge.** WhatsApp messages get opened up to 98% of the time, compared to about 20% for email, according to MessageBird research. Almost every message your AI agent sends actually gets read.

**Buyers expect it.** A 2024 Salesforce "State of the Connected Customer" report found that 71% of consumers expect companies to reach them on messaging channels. That number has only gone up since then.

**Speed matters.** Harvard Business Review showed that firms responding to a lead within five minutes are 100 times more likely to connect than those waiting 30 minutes. An AI agent on WhatsApp responds in seconds. Every time.

**Rich media built in.** WhatsApp supports images, videos, PDFs, catalogs, and interactive buttons. Your AI agent can do product demos right inside the chat.

If you are already running outbound campaigns on the platform, pair this guide with our [WhatsApp campaign anti-ban guide](/blog/the-anti-ban-blueprint-how-to-run-high-volume-whatsapp) to keep your number healthy as you scale.

Before building anything, you need access to the WhatsApp Business API. This is the official, enterprise-grade interface from Meta that lets businesses send and receive WhatsApp messages at scale. It is different from the free WhatsApp Business App that small shops use by hand.

Here is what you need.

**A verified Facebook Business Manager account.** This is where Meta manages your business identity and API permissions.

**A dedicated phone number.** The number you connect cannot also be used with the regular WhatsApp app or the WhatsApp Business App. Pick a number you are OK dedicating to automated sales conversations.

**An approved WhatsApp Business profile.** This includes your business name, logo, description, and category. A complete profile builds trust and gets better response rates.

**A platform that connects to the API.** Orki plugs directly into the WhatsApp Business API. It handles the technical stuff so you do not have to deal with webhooks, token refreshes, or message queues yourself. Follow the [WhatsApp Business setup](https://docs.orki.ai/docs/integrations/whatsapp/setup) guide to connect your number.

For more detail on API setup, costs, and business verification, read our [WhatsApp Business API](/blog/the-ceos-guide-to-the-whatsapp-business-api-how-to-turn-the) overview.

Once your WhatsApp number is connected, you are ready to start building.

A WhatsApp sales chatbot is only as good as the workflow behind it. Before you touch any settings, map out the sales process you want the AI agent to handle.

**Figure out the entry points.** How will prospects end up in your WhatsApp inbox? Common sources include click-to-WhatsApp ads on Facebook and Instagram, QR codes on packaging or in-store displays, website chat widgets that route to WhatsApp, outbound [WhatsApp marketing campaigns](/blog/ai-whatsapp-marketing-campaigns), and organic search where your WhatsApp number is listed.

**Map the conversation stages.** A typical sales conversation on WhatsApp goes like this. Greeting and intent detection. Understand why the prospect reached out. Product discovery. Ask about their needs and recommend relevant products or services. Objection handling. Address pricing concerns, competitor comparisons, or feature questions. Lead qualification. Figure out if the prospect fits your ideal customer profile. Conversion action. Book a demo, send a payment link, or hand off to a human closer.

**Set your success metrics.** Define what a "converted" lead means for your business. A booked meeting, a completed purchase, or a signed-up trial. Knowing this up front shapes every design decision after.

Write this workflow down in a simple document. It becomes your blueprint for everything below.

Your AI agent is only as smart as the information you give it. The knowledge base decides whether the chatbot gives accurate, helpful answers or vague filler.

**Gather your sales content.** Pull together product catalogs with descriptions, pricing tiers, and feature lists. Grab the frequently asked questions your sales team already handles. Include objection-handling scripts and competitor comparison sheets. Add shipping, return, and warranty policies. Throw in case studies or testimonials the agent can reference.

**Organize it for easy retrieval.** Break the content into clear categories. In Orki, you can upload documents, add FAQ pairs, or scrape content straight from your website. The AI agent uses retrieval-augmented generation (RAG) to pull the most relevant snippet for each customer query. Clean, well-organized data gives noticeably better responses.

**Keep it fresh.** A knowledge base is not something you set up once and forget. Schedule monthly reviews to update pricing, add new products, and remove discontinued items. Outdated information kills trust faster than no information at all.

If you sell online and want the AI agent to handle product recommendations, check our [e-commerce AI playbook](/blog/ai-customer-service-ecommerce-playbook) for strategies built around online stores.

To start building your knowledge base, [get started with Orki](https://docs.orki.ai/docs) and follow the guided setup.

With your workflow mapped and your knowledge base loaded, it is time to create the AI agent itself.

**Pick a personality.** The tone your AI agent uses has a real impact on conversion rates. A formal, corporate tone works for B2B financial services. A friendly, casual voice does better for consumer brands. In Orki, you can adjust tone, formality, and even humor through the [agent personality configuration](https://docs.orki.ai/docs/ai-agents/personality) settings.

**Set the language.** If your audience speaks multiple languages, common in regions like the GCC or Southeast Asia, enable the languages your customers are likely to use. Modern AI agents can detect language automatically and respond in kind. That removes a big friction point for non-English speakers.

**Define guardrails.** Tell the AI agent what it should never do. Never make up pricing or discounts that do not exist. Never make claims about competitors that cannot be verified. Never share internal company information. Always tell the customer they are talking to an AI agent when asked directly.

These guardrails keep your brand safe while the AI agent runs on its own.

For a full walkthrough of agent creation across use cases, see [how to set up any AI agent](/blog/how-to-set-up-ai-agent-7-steps).

Conversation flows are the structured paths your AI agent follows when certain conditions come up. Think of them as the skeleton of the conversation. The AI fills in the rest with natural language.

**Welcome flow.** This triggers when a new contact messages for the first time. The AI agent should introduce itself, say how it can help, and ask an opening question to figure out what the person wants.

**Product recommendation flow.** Based on what the customer says they need, the AI agent asks qualifying questions like budget range, use case, and team size. Then it recommends one to three products with short explanations. Include product images or catalog links where possible.

**Objection-handling flow.** Set up trigger phrases like "too expensive," "I need to think about it," or "what about [Competitor]?" to pull specific responses from your knowledge base. The AI agent should acknowledge the concern, share a relevant data point or testimonial, and then gently move the conversation forward.

**Checkout or booking flow.** When the prospect is ready to buy, the AI agent should send a direct action. A payment link, a calendar booking URL, or a summary of next steps. Keep the number of taps between "I'm interested" and "done" as low as possible.

**Re-engagement flow.** If a conversation goes quiet for 24 hours, the AI agent can send a follow-up message. WhatsApp's 24-hour messaging window policy means you will need an approved template for messages sent outside that window. Prepare these templates ahead of time.

Not everyone who messages your WhatsApp number is a qualified buyer. Lead qualification makes sure your human sales team spends time on the prospects most likely to close.

**Define your qualification criteria.** Use a framework like BANT (Budget, Authority, Need, Timeline) or a simpler model that fits your business. Budget. Can they afford your solution? Authority. Are they the decision-maker or just researching? Need. Do they have a real problem your product solves? Timeline. Are they looking to buy now, this quarter, or "someday"?

**Ask qualification questions naturally.** The AI agent should weave these into the conversation instead of firing them off like a survey. For example, instead of "What is your budget?", the agent might say, "To recommend the best option, it helps to know your expected investment range. Are you looking at the standard plan or the enterprise tier?"

**Score and tag leads.** Set up your AI agent to assign scores or tags based on responses. In Orki, you can use customer attributes to store qualification data like budget tier, company size, and purchase intent. Then use those attributes to segment leads for follow-up campaigns or priority routing.

**Route by score.** High-scoring leads go straight to your best closers. Medium-scoring leads stay with the AI agent for nurturing with more content. Low-scoring leads go into a drip campaign for the long term.

According to Gartner, organizations that use AI for lead scoring and qualification see a 30% improvement in conversion rates compared to manual qualification. Automating this on WhatsApp combines that efficiency gain with the channel advantage.

No AI agent should run without a safety net. There will always be edge cases, complex deals, and emotionally charged situations where a real person is needed.

**Define handover triggers.** Common ones include when the prospect asks to speak with a human, when the conversation hits a negotiation phase with custom pricing, when the AI agent picks up on frustration or repeated misunderstanding, when the lead score is high enough to justify personal attention, and when the query falls outside the knowledge base and the agent cannot give a confident answer.

**Make transitions smooth.** When a handover happens, the AI agent should tell the customer a team member is joining. It should pass the full conversation history and lead qualification data to the human agent. The customer should never have to repeat themselves.

In Orki, human handover routes the conversation to a specific team or person based on rules you set, like language, product line, or lead score. This keeps things smooth for the customer and organized for your team.

**Set availability windows.** If your team only works business hours, set up the AI agent to handle off-hours conversations on its own and queue handover requests for the next available slot. The customer should get a clear expectation, like "A member of our team will follow up with you at 9 AM GST."

Skipping testing is the most common mistake. A WhatsApp sales chatbot that gives wrong prices or fails to hand over to a human will burn trust with real prospects.

**Internal testing.** Have your sales team message the AI agent and try to break it. Ask edge-case questions. Misspell product names. Switch languages mid-conversation. Request things outside the scope. Write down every failure and fix it in the knowledge base or conversation flows.

**Pilot launch.** Start with a small slice of traffic. Maybe leads from a single ad campaign or a specific region. Watch conversations in real time for the first 48 hours. Look for accuracy of product information, tone that feels right, fast response times, correct handover behavior, and accurate lead qualification.

**Fix before you scale.** Use pilot data to make adjustments. Common fixes include adding FAQ entries for questions the AI agent could not answer, tweaking the personality to better match your brand, and adjusting qualification thresholds based on real conversion data.

**Full launch.** Once you trust the agent's performance, open it to all traffic sources. Connect it across channels. You can [connect WhatsApp, Instagram, and web chat](https://docs.orki.ai/docs/integrations/overview) from a single Orki workspace. Start routing all inbound sales inquiries through the AI agent.

Once your WhatsApp sales chatbot is live, track these metrics to see how it performs.

**Response time.** How fast does the AI agent reply? Aim for under 5 seconds on the first message.

**Qualification rate.** What percentage of inbound conversations turn into a qualified lead? Compare this to your old numbers.

**Conversion rate.** Of the qualified leads, how many reach your target action? A booked demo, a completed purchase, a started trial.

**Handover rate.** What percentage of conversations need a human? A declining handover rate over time means the AI agent is getting better.

**Customer satisfaction.** Run post-conversation surveys or look at sentiment within the chats. A Zendesk 2025 CX Trends report found that 67% of consumers are willing to interact with AI agents for buying, as long as the experience is smooth. Your satisfaction scores tell you if you are hitting that bar.

**Revenue attributed.** Track actual revenue from WhatsApp conversations. This is the number that justifies the investment.

Review these weekly for the first month, then move to monthly. Build dashboards your sales leadership can check without digging through raw data.

There are three cost pieces. The WhatsApp Business API fees from Meta, which are charged per conversation and vary by country and type. The AI platform fee, where Orki offers tiered plans based on conversation volume. And the time you spend building your knowledge base and flows. Most small-to-midsize businesses can get a fully working WhatsApp sales chatbot running for a few hundred dollars per month. You can [try Orki free](https://app.orki.ai) to explore the platform before committing.

No. Platforms like Orki are built for non-technical teams. You configure your AI agent, upload your knowledge base, and connect your WhatsApp number through a visual interface. No code, no API calls, no server management. See [how to set up any AI agent](/blog/how-to-set-up-ai-agent-7-steps) for a walkthrough.

Yes. Modern AI agents can detect the language a customer is using and respond in that language. This is especially useful in multilingual markets. In Orki, you can turn on specific languages and even adjust the agent's personality for each one to match cultural communication norms.

It should be honest about it and offer to connect the customer with a real person. In Orki, you set up fallback behavior so the agent never makes up an answer. It either gives verified information or hands the conversation to someone who can help.

Yes, as long as your knowledge base has accurate, current pricing and your guardrails stop the agent from inventing discounts or terms. The AI agent should always pull from the pricing data you loaded. It should never generate numbers on its own. For complex or negotiated pricing, set it up to hand over to a human rep.

Meta has strict quality and spam rules for WhatsApp Business API users. The key points are to only message contacts who have opted in, keep a high quality rating by giving relevant responses, and avoid sending bulk promotional messages without approved templates. Our [WhatsApp campaign anti-ban guide](/blog/the-anti-ban-blueprint-how-to-run-high-volume-whatsapp) covers this in detail.

Yes. Orki lets you build one AI agent and run it on multiple channels. Your knowledge base, personality settings, and qualification logic stay the same whether the customer reaches you on WhatsApp, Instagram DMs, or your website's live chat. Learn more about connecting channels in the [integrations overview](https://docs.orki.ai/docs/integrations/overview).

Most teams can have a basic WhatsApp sales chatbot running within a few days. The timeline mostly depends on how fast you can pull together and organize your knowledge base content. The technical setup, connecting your WhatsApp number, creating the AI agent, and configuring conversation flows, usually takes a few hours in Orki. Plan an extra week for testing and tweaking before opening the agent to full traffic.