WhatsApp Automation Developer: 11 Interview Questions, Scored

Eleven questions to ask a WhatsApp automation developer, with good and bad answers on the official API, templates, duplicates and proof, plus a scoring table.

By Carlos Familia, CRM & AI automation expert. Updated October 16, 2026. 7 min.

The complaint at a Spanish-language wellness clinic was easy to state: some leads got the same WhatsApp follow-up two or three times. The fix took longer to find. Four workflows each changed the lead's stage and only then, a second later, set the tag meant to block repeats. The stage change woke the next workflow before that tag existed.

I reordered the steps, put a check in front of every WhatsApp send and pulled 31 contacts out of the loop. Nothing about it was exotic. It is the kind of flaw a single interview question brings out, and it is question 5 below.

How do you hire a good WhatsApp automation developer? Ask the eleven questions below, score each answer from 0 to 2, and pay for a small test flow before the real project. The setup they should describe by default is Meta's official WhatsApp Business Platform: your number inside your own Meta business account, connected directly or through a provider you hold the account with.

One policy to know first. Meta's business platform terms, last modified September 23, 2026, bar AI providers from the platform when a general-purpose AI assistant is the main product on offer. In my reading, a support or booking bot for your own business is incidental to your service, but ask the developer which of the two they are building.

Questions about the connection and the account

1. Which connection will carry my customers' messages?

A good answer sounds like: "Meta's WhatsApp Business Platform. Your number lives in your Meta business account, and I connect through the Cloud API or a provider you hold the account with. I can show you the connected status."

A bad answer sounds like: "I use a QR-code gateway. It's cheaper and there's no approval."

WhatsApp's terms of service list "bulk messaging, auto-messaging, auto-dialing" among impermissible communications. If a gateway gets the number banned, you carry the cost, so I keep unofficial gateways away from any number customers depend on.

2. Who owns the Meta business account, the number and the templates?

A good answer sounds like: "You do. I get a user role, and at handover you hold every login and every template."

A bad answer sounds like: "I'll set it up under my agency and transfer it later."

Ownership decides what happens when a relationship ends. A portable storage owner I worked with was losing access to a franchise account, and the franchise would not move it, so I had to clone everything into a new account. Templates are worse: they cannot be carried from one business account to another at all.

3. What happens outside the 24-hour window?

A good answer sounds like: "Free-form replies go out only while the window the customer opened is still active. After that, a workflow sends an approved template. In GoHighLevel I put the window check, with its Open and Closed branches, in front of every send."

A bad answer sounds like: "The bot just keeps sending."

A good builder treats the window as a gate, not a guideline. On a facial spa's chat bot, which answers on Facebook and Instagram, the reactivation tagger only reaches a lead while Meta's 24-hour messaging window is open, and it caps each lead at four touches. On WhatsApp, a late follow-up written as free text simply fails.

4. What will the messages cost at my volume?

A good answer sounds like: "Meta charges per delivered message, by category and country. Its pricing page, updated September 30, 2026, gives each number 1,000 free service replies a month and charges after that. Utility templates are charged inside the window too since October 1. Here is your monthly estimate by message type, and we check it against your first invoice."

A bad answer sounds like: "WhatsApp is free."

Meta's own explainer on the October change, updated September 28, does not mention the free tier, which is one more reason to check the billing in WhatsApp Manager after the first month. The rates and a worked month are in WhatsApp chatbot cost.

Questions about duplicates and consent

5. What happens when two workflows touch the same contact within a second?

A good answer sounds like: "The guard tag goes on before the stage moves. Every send checks the stage and a sent-tag, the workflows remove the contact from each other, and a reply stops all of them."

A bad answer sounds like: "That shouldn't happen."

6. What happens if the same inbound message arrives twice?

A good answer sounds like: "Each message is claimed in a database before anything replies, and each message has exactly one entry point."

A bad answer sounds like: "WhatsApp only delivers once."

I have fixed both causes in production: overlapping scheduled polls, and two webhooks firing for one message. The details are in n8n WhatsApp integration.

7. How do opt-in and opt-out work?

A good answer sounds like: "I store where and when each person opted in, honor stop requests on every channel, and keep volume low enough that complaints never cap the number."

A bad answer sounds like: "I'll upload your contact list."

WhatsApp's business messaging policy says you may contact people only if they gave you their number and opted in to your messages. It also says WhatsApp may limit how many messages a business can send if it gets a lot of negative feedback.

Questions about proof

8. How will I know a message was never delivered?

A good answer sounds like: "I watch delivery status, not only the send step. Critical staff alerts have a second channel, and an alert fires when expected messages stop."

A bad answer sounds like: "The log says sent."

At a mortgage brokerage in Florida, the WhatsApp staff alert kept reporting "accepted" for days after GoHighLevel had lost its connection to the business account. After I moved that alert to SMS, the next real lead reached four staff phones.

9. How does the bot prove it booked something?

A good answer sounds like: "A booking counts only when the calendar tool reports success, and the bot's reply is built from what the tool returned."

A bad answer sounds like: "The prompt tells it to confirm bookings."

For an insurance agency in Puerto Rico, whose bot kept telling leads they were booked, the fix was a real booking action on the correct calendar plus a rule to confirm only after the reservation exists.

10. What does the bot do with "mañana a las 4"?

A good answer sounds like: "Code turns relative dates into real dates. The tools return a weekday table, and the model never works out today's date."

A bad answer sounds like: "GPT knows what tomorrow is."

A facial spa's bot once booked "tomorrow (Saturday)" for the same day, because its sense of now came from a server clock in UTC.

11. How do you test without messaging my real customers?

A good answer sounds like: "A test contact, a whitelist and a test calendar. GoHighLevel's trial chat can book real appointments, so I never point it at a calendar real people use."

A bad answer sounds like: "I'll try it on a few real leads first."

I learned that one when a QA run booked a real slot on the insurance agency owner's calendar.

How do I score the answers?

Give each answer 0 for a bad one, 1 for a vague one and 2 for a good one backed by a real story. The maximum is 22. A zero on question 1, 5 or 8 rules the candidate out whatever the total, because those three decide if your number survives, if customers get one message each and if you ever hear about failures.

Ask every candidate the same questions in the same order. Finish by asking each one to describe the last time a WhatsApp send failed in production and how they found out. Someone with no such story has either never run one or never noticed, and both should lower the score.

For the build side, read WhatsApp CRM integration. The general interview for automation builders is in hire an n8n developer, and the workflow traps are in GoHighLevel workflow mistakes. Clinics have extra rules, which the healthcare page outlines. You can also book a call and put the same eleven questions to me.

About the author

Carlos Familia, CRM & AI automation expert. GoHighLevel, n8n, AI agents, servers. Full-stack developer.

I build CRM systems in GoHighLevel, n8n automations on my own servers and AI voice and chat agents for service businesses. Full-stack developer: I write the workflow, the API integration, the database and the server it runs on, and I test every build with real journeys before handoff.