For years, "chatbot" became a catch-all word. Many companies rolled out simple flows to answer frequent questions, collect basic data, or try to lighten the load on the support team. Some worked well. Many turned into long menus, stuck responses, and frustrating experiences for the customer.
The current movement is different. The promise behind AI agents is to answer questions, recommend products, qualify leads, support sales, schedule actions, and escalate service with more context. Instead of just following a fixed decision tree, AI starts acting as an active layer inside the commercial journey — it's no longer just AI customer service on WhatsApp answering FAQs.
For businesses using WhatsApp in sales and support, that changes the conversation. The question is no longer "should we have a chatbot?" The right question is "which parts of the operation can be safely handled by AI, with quality and a positive financial impact?"
A company receiving hundreds or thousands of messages a month needs to separate types of demand. There are simple questions — status updates, deadlines, documentation, availability, commercial policies, and next steps. There are commercial questions — comparing solutions, recommendations based on need, objections, contract terms, and interest qualification. And there are sensitive cases — complaints, delivery issues, technical support, negotiation, and situations that require a human.
Putting everything in the same flow is a mistake. Automating everything without criteria is too. The real value of AI is in classifying, resolving what's repetitive, handing the rep useful context, and cutting the time spent on conversations that shouldn't have to start from zero.
Meta Business Agent also puts pressure on measurement. If AI becomes part of the sales and support operation, it needs to be measured like one. How many conversations did it start? How many did it resolve? How many needed a human? How many generated opportunities? How many moved through the funnel? How many cut response time? What did each interaction cost?
This matters because usage-based pricing changes how the channel is perceived. When automation feels "free," companies tolerate waste. When it carries a clearer cost, the design of the operation has to get better.
Get MergeNews in your inbox
Every week, news about WhatsApp, AI, CRM, and sales automation to turn customer service into sales.
On WhatsApp, this gets even more sensitive because the channel blends support, relationship, and sales. A contact might start with a simple question, then ask for a recommendation, compare alternatives, negotiate a deadline, request support, or come back days later with a new intent. Without integration with a CRM, history, and journey data, AI stays limited. With context, it can actually help.
The human's role changes too. A rep shouldn't get just any conversation. They should get qualified conversations — with history, intent, product interest, main objection, and a suggested next step. That increases productivity and reduces the feeling that automation is competing with the sales team.
For businesses, the smartest move is to start small. Before trying to automate the entire operation, map the 20 most frequent questions, separate simple questions from commercial ones, and set objective handoff rules. AI should resolve what's repetitive, prepare the ground for the rep or agent, and protect the customer experience — the same principle behind where to automate on WhatsApp without seeming like a bot.
The risk lies in using AI as a trend. A poorly trained automation can answer incorrectly, irritate customers, create rework, and still raise costs. A well-designed one can reduce operational load, speed up responses, recover opportunities, and improve conversion.
This is where Merge comes in: AI on WhatsApp can't be treated as a toy. It's an operational line item. If a company isn't measuring it, it might end up paying for automation without knowing whether it's selling more, serving customers better, retaining them, or just answering more messages. The future of WhatsApp customer service won't be human versus AI. It will be a hybrid operation, with tasks clearly split: AI for scale, triage, and repetitive answers; humans for negotiation, relationship, and decisions that require commercial judgment.
Practical action for businesses
- Map the 20 most frequent support questions.
- Separate simple questions, commercial questions, and cases that require a human.
- Create handoff rules for sales or support reps.
- Measure conversations started, resolved, transferred, and converted.
- Assess whether AI is reducing cost or just increasing the volume of replies.