Customer service mapping
Most common questions, information that qualifies a lead and the moment to hand over to a person.
Implementation of an AI agent on the official WhatsApp API. It answers, qualifies and hands the conversation to your team at the right moment, with the ad source recorded. It is not a SaaS for you to configure on your own.
The service
The campaign works, the lead messages on WhatsApp and nobody replies for two hours. By the time someone does, the person has already talked to a competitor. That is the first problem an AI WhatsApp chatbot solves: response time.
A menu bot makes customers type "1", "2" or "3". It works for little and annoys people with a specific question. An AI agent understands the question as the customer writes it, answers based on your company's information and asks qualifying questions naturally.
The AI doesn't negotiate, doesn't promise what it can't and doesn't make up answers. When the conversation needs judgment, or when the customer asks for a person, it hands over to the team with a summary of what was discussed.
The chatbot is designed together with paid traffic management: it knows which campaign the lead came from, records the qualification and feeds that data back into the media review. That shows which ad brings good conversations, not just conversations.
Scope
The chatbot is the visible part. The work is in what comes before.
Most common questions, information that qualifies a lead and the moment to hand over to a person.
Services, rules, hours and limits written so the AI answers correctly and doesn't make things up.
Connection through the official WhatsApp Business platform, without the blocking risk of QR-code connections.
CRM, calendar or spreadsheet, when the tools allow it. The lead arrives organized, with the ad source.
Clear rules for when the conversation goes to the team and alerts for whoever needs to reply.
Test conversations before going live and review of real answers in the first weeks.
Comparison
Both exist for good reasons. The choice depends on volume and how much customer service weighs on the sale.
| Criterion | Ready-made platform (DIY) | Custom implementation |
|---|---|---|
| Who configures it | The company itself | The implementer, with the company validating |
| AI in the plan | Not always in the entry plan | Defined by the project |
| Integration with ads | Limited to what the platform offers | Designed to record the lead's source |
| CRM and calendar integration | Depends on ready-made connectors | Built around the company's tools |
| Best for | Small volume and simple questions | Ad leads, qualification and multi-step processes |
The process
Five steps, from mapping to follow-up.
Conversation volume, repeated questions, lead sources and existing tools.
What the AI answers, what it asks and when it hands over to a person.
Company knowledge, official API, CRM, calendar and ad source tracking.
Simulated conversations with real questions, including the hard ones, before opening to customers.
Review of real conversations and adjustments to answers and handover rules.
Who leads
Ramon Saium leads the strategy: how the chatbot talks to the campaign, which qualification data matters and how that feedback improves the media. A qualified team handles the technical implementation under his direction.
If you've tried a bot that annoyed customers or a tool nobody could maintain, the conversation starts with what went wrong. Scope and exclusions come in writing.
Investment
The cost has three parts: implementation, the platform or maintenance subscription, and usage costs that run separately. Many proposals only show the first.
Brazilian market reference, Oct/2026 Ranges in Brazilian reais (BRL) published by other companies in Brazil. These are not ArcWords prices.
| Component | Market range (BRL) | Note | Source |
|---|---|---|---|
| Ready-made platform (SaaS) | R$ 199 – 3,900/mo | Plans published by several platforms; some charge setup separately. | Certu |
| Simple implementation | R$ 700 – 1,200 | One WhatsApp flow, no integrations. | Hora de Codar |
| Intermediate implementation | R$ 1,500 – 3,000 | Integration with spreadsheet, CRM or calendar. | Hora de Codar |
| Advanced implementation | R$ 3,500 – 8,000+ | Generative AI, multi-agent service, CRM and dashboards. | Hora de Codar |
| Monthly maintenance | R$ 200 – 1,000/mo | Adjustments, monitoring and support. | IA do Brasil |
| Agency in São Paulo | from R$ 2,000/mo | AI systems, CRM and funnels: R$ 3,000 to R$ 8,000 per month. | Agência 365 (SP) |
Usage costs paid separately, per Zap Trend: AI model (tokens) R$ 30 – 250/month, WhatsApp API R$ 0 – 100/month, server R$ 80 – 250/month and Meta template messages outside the 24-hour window at about R$ 0.04 – 0.31 each, for an estimated R$ 500 – 1,400/month total. Replies within the 24-hour service window are not charged by Meta. Ranges reviewed quarterly.
More conversations mean more AI usage and, sometimes, more messages charged by Meta.
Each connected system (CRM, calendar, spreadsheet, e-commerce) increases the scope.
A few services with simple rules or a large catalog with exceptions.
Reminders, confirmations and campaigns outside the 24-hour window are charged per message.
How many people receive the handover and at what hours.
One-time delivery or ongoing review of conversations and answers.
ArcWords sends a proposal after the diagnosis, with scope, deliverables and exclusions in writing.
Before hiring
The number that receives the leads is a company asset. Use the official API and keep the business account in the company's name. Tools that connect via QR code may stop working or get the number blocked.
Unbounded AI can confidently give a wrong answer. The knowledge base must state what the AI doesn't know and when it should hand over to a person. That is part of the project, not a detail.
Collect only what is necessary and say what it is for. Conversations with personal data should be kept in tools with access control, in line with Brazil's data protection law (LGPD).
Questions
It depends on three parts: implementation, subscription and usage costs. In Oct/2026, ready-made platforms published plans from R$ 199 to R$ 3,900 per month (Certu). Custom implementations range from about R$ 700 to more than R$ 8,000 depending on integrations (Hora de Codar). AI usage, server and API add R$ 500 to R$ 1,400 per month, according to Zap Trend. These are Brazilian market references, not ArcWords prices.
For a custom implementation, yes: it covers mapping customer service, writing the knowledge base, connecting the API and integrating with CRM or calendar. Some ready-made platforms also charge setup, others don't. Compare what each proposal delivers at this stage.
On the official API, replies within the 24-hour service window, opened when the customer writes, are free. Template messages sent outside that window are charged per message. Zap Trend cites about R$ 0.04 for utility and R$ 0.31 for marketing. That amount is paid to Meta or the provider, not to the agency.
The official WhatsApp Business API is the recommended path for companies. Unofficial QR-code connections imitate the app and risk getting the number blocked. For a channel that receives ad leads, that risk isn't worth it.
No. It answers repeated questions, collects initial information and routes. Negotiation, exceptions and decisions stay with people. Human handover is part of the project from the start.
Yes, when the tools allow integration. The qualified lead can enter the CRM with the ad source, and bookings can be made directly in the conversation. Each integration increases the implementation scope.
Next step
Tell us how many conversations come in, where they come from and where service gets stuck. The proposal comes after the diagnosis.