How many people do you need for WhatsApp support without drowning

You need enough people to cover the hours of real conversation work that land inside your support hours, divided by the productive chat hours each person has. There's no universal ratio. A store getting thirty messages a day might be fine with one part-timer, while a brand that drops new products every week may need shifts even at the same daily average, because everything arrives in a two-hour window.

The most common mistake is hiring based on total monthly messages. That number hides what actually matters: how many conversations (one thread with one customer about one issue, start to finish) arrive per hour at peak, and how many of them are the same question you answered yesterday. Measure those two things for two or three weeks before you post a job ad.

The practical calculation has four steps: count conversations per day, measure the minutes of real work each one takes, find your peak hours, and subtract whatever already gets resolved without a human. Handle time isn't how long the chat stays open; it's the minutes someone actually spends reading, looking things up in Shopify or a spreadsheet, and writing the reply.

Here's a concrete example. An apparel store gets about 120 conversations a day. Looking through them, 70 are sizing, shipping and order-status questions that take around 3 minutes each; 35 are product or styling questions at around 8 minutes; and 15 are real problems (exchanges, double charges, late orders) at around 15 minutes. That's roughly 13 hours of chat work a day. If one person gets about 6 productive chat hours out of a shift, the store needs a bit more than two people just to keep up, before accounting for peaks.

Peaks change the math. If half those conversations arrive between 7 and 11 p.m., the issue isn't headcount, it's who's on in that window. That's why team size gets decided shift by shift, not from the daily average.

An experienced agent can usually handle two or three active conversations at a time; beyond that, quality drops. For live chat, one best-practice guide suggests a maximum of three concurrent chats per skilled agent as a starting point, and contact center practitioners broadly land on two chats as the guideline for most agents, with a third for stronger performers.

WhatsApp is more asynchronous than website chat. Customers reply when they can, sometimes half an hour later. That lets an agent keep more threads open, but not more threads active. The useful rule is to separate the two: someone can have twenty conversations waiting on the customer, but should only be typing in two or three at once. Push an agent to eight active chats and the classic mistakes show up: sending one customer another customer's order details, forgetting a promised follow-up, or firing off short replies that read as indifferent.

AI should take the repetitive questions that have a known answer and leave people the work that needs judgment. In the apparel example, the 70 sizing, shipping and order-status conversations are exactly what an AI agent connected to your catalog and orders can answer at any hour. Take that off the team's plate and those 13 daily hours drop to roughly half. The question shifts from "who do we hire?" to "how do we point the people we already have at the hard stuff?"

For that to work, the AI needs the same information you'd give a new hire: return policy, shipping times by region, size guide, brand voice. That's what an AI agent knowledge base is for. If an answer isn't in there, the AI shouldn't make one up. It also needs a clear exit: escalation rules that hand the conversation to a person when the customer is upset, a charge is disputed, or they ask for a human.

There's a middle step that suits a lot of teams: copilot mode, where the AI drafts the reply and a person reviews and sends it. It doesn't reduce how many conversations reach the team, but it cuts handle time on each one, and it builds trust before you let the AI answer simple topics on its own. If you're weighing that call, we cover when AI should suggest and when it should reply on its own.

The first mistake is staffing for the average day. A team built for a normal Tuesday falls apart on launch day, Black Friday, or the day a WhatsApp broadcast goes to your whole list, and the customer who messaged at 9 p.m. hears back the next morning, after they've already bought somewhere else. If your campaigns create spikes, plan coverage for those windows separately.

The second is measuring only speed. First response time matters, but an instant "hi, one moment" followed by two hours of silence doesn't help anyone. Also track how long conversations take to resolve and how many get reopened because the answer didn't work.

The third is running sales and support through one queue with no priorities. A complaint about a late order and a shopper ready to buy compete for the same person, and the sale usually loses. Splitting queues, or at least tagging and prioritizing, changes how many people you need on each side. The fourth is never revisiting the numbers: what you need today won't be what you need in three months, and the only way to know is to look at conversations every week.

With heysteff, WhatsApp, Instagram, Messenger and Gmail conversations land in one inbox, and Steff, the AI agent, answers what it can using data from Shopify and the knowledge base. When a rule or keyword says the case needs a person, the conversation is escalated with full context, so nobody asks the customer to repeat themselves.

Weekly conversation analytics show what people are writing about, how many conversations were handed to humans, and why. That turns hiring into a data call: if handoffs grow on a topic the AI could handle with better information, you fill in the knowledge base; if they grow on cases that genuinely need judgment, it's time to add someone. You can compare plans on the pricing page.

How many WhatsApp messages can one person answer per day? It depends on the mix: simple questions take a few minutes, order problems can take fifteen or more. Work it out from your own handle times and roughly 6 productive chat hours per shift, not the full 8 hours on the contract.

Can multiple agents use the same WhatsApp number? Yes. With a shared inbox connected to the WhatsApp Business Platform, several people work the same number, each conversation gets assigned to someone, and there's a record of who said what. That prevents double replies and chats nobody picks up.

Will AI replace my support team? Not entirely, and it shouldn't. AI handles repetitive questions with known answers well; people are still needed for complaints, exceptions, complex sales and upset customers. What changes is how big a team you need for the same volume and what kind of work that team does.

• Copilot mode: AI suggests, your team decides
• Knowledge base for your AI agent
• How to set human escalation rules
• AI customer support