AI Can Now Handle Messages—But Should It?
As a Malaysian SME owner, you may spend part of every day replying to customer enquiries, following up on quotations, confirming appointments, and reminding staff about pending tasks. These messages may look simple, but they interrupt focused work and create delays when nobody is available to respond.
A new ChatGPT plug-in for Apple Messages shows where business communication is heading. According to TechCrunch, the plug-in can connect with a user’s Messages inbox, analyse conversations, suggest follow-ups, draft replies, search message history, delete messages, and send messages on the user’s behalf. OpenAI also warns users to review messages before sending them.
TL;DR: AI-assisted messaging can help you organise conversations and prepare replies faster, but sending messages automatically creates privacy, accuracy, and customer-service risks.
For most SMEs, the sensible starting point is not “let AI reply to everyone”. It is “let AI help you prepare better replies, while you keep the final approval”.
What This Means
The plug-in connects ChatGPT with Apple Messages so the assistant can work with conversations that would normally remain inside the messaging app. You could ask it to find messages from the previous day, identify people who need a follow-up, suggest a reply, or send a message after you approve it. The reported functions include sorting, analysing, editing, searching, deleting, drafting, and sending messages. These capabilities are described by TechCrunch.
For a business, this is similar to having an assistant review an overflowing inbox and prepare a short action list. The difference is that an AI assistant can process conversations quickly, but it may not understand the full relationship behind a message. It may miss sarcasm, cultural context, urgency, or an informal promise you made earlier.
That matters because a message is not just information. It represents your business. A careless reply can confuse a customer, expose private details, promise something your team cannot deliver, or make a sensitive situation worse.
The safest principle is simple: use AI to reduce typing, not to remove responsibility.
How This Applies to Malaysian SMEs
1. Retail and online sellers can manage follow-ups more consistently. If you sell through social media, marketplaces, or direct messages, customers may ask about stock, delivery timing, colours, sizes, or product suitability. AI could help you find conversations that have not received a reply and prepare responses using information you provide. You still need to confirm actual stock and delivery status before sending anything. A polished but incorrect answer is worse than a slightly slower honest reply.
2. Service businesses can reduce appointment administration. Salons, repair providers, tuition centres, clinics, consultants, and home-service businesses often exchange several messages before confirming a booking. An assistant could identify customers who asked for an appointment, draft a reminder, or summarise the details for your team. However, the final message should be checked against your calendar. A reply that offers an unavailable slot or sends the wrong address can create unnecessary disruption.
3. Wholesalers and B2B suppliers can organise quotation conversations. Your sales team may have many message threads involving specifications, delivery requirements, payment terms, and approval questions. AI can help summarise what each customer is waiting for and suggest the next action. This is useful when one person is on leave or when you need to take over an account. Before using such a system, remove unnecessary sensitive information and define who is allowed to access customer conversations.
4. Restaurants and food operators can improve enquiry handling. Catering requests may involve guest numbers, dietary requirements, event dates, delivery locations, and menu changes. AI can prepare a checklist from the conversation, but you should not rely on an automatically generated reply for allergy-related or special dietary matters. Those details require human confirmation from the person responsible for food preparation.
5. Small teams can create a clearer handover process. In many SMEs, customer messages sit on one owner’s phone. When that person is busy, the business appears slow even though other employees could help. A better system is to use AI to summarise conversations and identify pending actions, then record those actions in a shared customer-management system. The objective is not merely faster replies; it is making customer knowledge available to the right team member without exposing every private conversation.
Important Privacy and Accuracy Questions
OpenAI told Bloomberg that the plug-in runs locally on a user’s machine and does not create an index of all messages, as reported by TechCrunch. That may reduce some concerns, but you should still understand exactly what information the tool can access, what is processed, what is retained, and which account or device controls it.
Before connecting any messaging account, ask whether conversations contain identity-card details, addresses, health information, bank information, passwords, confidential quotations, or information about other people. You should also consider whether customers have agreed to their messages being processed by an external AI service.
Sending is the most sensitive function. OpenAI discourages persistent approval because it removes the final opportunity to review a message, according to TechCrunch. For business use, keep approval switched on. Your staff should see the recipient, message content, attachments, and any promises before the message leaves the business.
A Practical Risk-and-Use Guide
| Use case | Suggested AI role | Human approval |
|---|---|---|
| Finding unanswered enquiries | Search and summarise conversations | Recommended |
| Drafting routine replies | Prepare a response from approved information | Required |
| Appointment reminders | Draft based on a confirmed calendar entry | Required |
| Refunds, complaints, or disputes | Summarise the issue and suggest options | Required; manager review |
| Messages containing sensitive information | Avoid unless your policy permits it | Strict review |
| Automatic sending to customers | Use only for narrowly defined, low-risk messages | Do not remove approval casually |
The table is a practical control framework rather than a technical requirement. Start with low-risk tasks such as drafting internal reminders or summarising non-sensitive enquiries. Do not begin with complaints, legal matters, payment disputes, employment issues, or messages involving personal information.
Practical Takeaways for Your Business
- Start with drafting, not sending. Let AI prepare replies while a person remains responsible for approval.
- Create approved reply guidance. Give your team current information about delivery, booking, warranty, returns, and escalation procedures.
- Set a sensitive-information rule. Identify details that must not be entered into or processed by an AI tool.
- Use a shared record. Important customer actions should be recorded in your CRM, task list, or business system rather than left inside one person’s chat history.
- Check the recipient carefully. Similar names and old conversation threads can lead to embarrassing mistakes.
- Keep an audit trail. Record who approved important customer messages and when they were sent.
- Test with internal conversations first. Observe mistakes before allowing the tool to interact with real customers.
- Review the tool’s permissions. Understand which messages, contacts, files, and applications it can access.
The Bigger Picture
This development points to a broader change: business software is moving from showing you information to taking action for you. Instead of opening several apps, searching through conversations, and writing each follow-up manually, you may soon describe the outcome you want and let an assistant prepare the steps.
That convenience will make business processes more important, not less. AI can only produce reliable messages when your business has clear rules, updated information, consistent customer records, and defined approval responsibilities. If your delivery policy changes frequently but nobody updates the source information, AI may simply produce incorrect answers faster.
For Malaysian SMEs, the best advantage will come from combining automation with practical human judgement. Use AI for repetitive preparation, searching, summarising, and drafting. Keep people responsible for commitments, sensitive information, complaints, exceptions, and final approval.
You do not need to connect every message today. Choose one repetitive workflow, define the acceptable response, test it with your team, and measure whether customers receive clearer and faster service. If the process works safely, expand it gradually.
Ready to Streamline Your Operations?
Your business should run itself. AutoRunBiz deploys AI agents to automate your daily operations — WhatsApp orders, invoicing, customer follow-ups, and accounting. Book a free 15-min ops audit to see where automation fits your business →