When Technology Promises More Than Your Business Can Verify

You may have seen the latest enthusiasm around artificial intelligence and automation. Software can draft emails, analyse documents, write code, summarise meetings and answer customer questions. For a busy Malaysian SME owner, that sounds appealing: fewer repetitive tasks, faster responses and less time spent chasing information.

But the practical question is not whether AI looks impressive in a demonstration. The question is whether you can safely check its work before it affects your customers, staff, stock, compliance or reputation. That is the concern behind the current technology backlash discussed in The Verge’s Decoder discussion.

TL;DR: Automation works best when the result is easy to verify. Use AI for repeatable, reviewable tasks first, while keeping people responsible for decisions that involve uncertainty, judgement or customer impact.

For a small business, this distinction matters more than technology hype. You do not need to reject automation. You need to choose processes where the software can help without quietly creating new problems.

What This Means

The central idea is simple: some work has a clear right or wrong answer, while other work requires judgement in the real world.

For example, software can check whether an invoice total matches the listed quantities and tax fields. It can identify duplicate customer records or alert you when stock falls below a set level. These tasks are relatively easy to verify because you can compare the result against a rule, calculation or source document.

Other tasks are much harder to verify. An AI system may produce a confident supplier recommendation, interpret an unclear customer complaint or suggest wording for a sensitive employment matter. The output may sound professional while still being incomplete, unsuitable or factually wrong.

The Decoder discussion makes this distinction through software engineering and mathematics. In those areas, results can often be tested. Code can be run, and mathematical answers can be checked. Outside those areas, verification becomes harder. A proposed business strategy, hiring recommendation or medical claim cannot be approved simply because the wording sounds convincing.

The useful question is not “Can the system produce an answer?” It is “How quickly and reliably can you check whether the answer is safe to use?”

This changes how you should evaluate automation. Instead of asking whether a tool appears intelligent, ask whether it fits a controlled workflow. Can the input be trusted? Can the output be checked? Is there a person responsible for approving it? Can you reverse the action if something goes wrong?

How This Applies to Malaysian SMEs

Consider a Malaysian trading or distribution company managing orders through WhatsApp, spreadsheets and accounting software. An AI assistant could extract product names and quantities from incoming messages, prepare a draft sales order and flag missing delivery details. That is a sensible starting point because your staff can compare the draft with the original customer message before confirming it.

However, you should not allow the same assistant to automatically approve unusual credit terms or promise a delivery date without checking stock and transport availability. Those decisions depend on local conditions, supplier reliability and customer history. A polished automated response can create an operational commitment that your team cannot fulfil.

For a restaurant, café or food wholesaler, automation can help organise daily purchasing. A system could compare recent sales with current stock, prepare a replenishment list and highlight items nearing their use-by date. Your manager can review the list before placing an order. This creates a clear approval point and helps reduce forgotten tasks without pretending that demand can be predicted perfectly.

For a professional services firm, AI can summarise meeting notes, classify enquiries and draft follow-up emails. That saves time, especially when several team members handle leads. Yet the final message should be reviewed by someone who understands the client’s situation. A summary may omit an important qualification, misread a deadline or make a commitment that was never agreed.

Construction, renovation and maintenance businesses can use automation to turn site updates into structured job records. Staff might submit photographs, voice notes and material requests through a simple form. The system can organise these entries by project and alert the office when a task is overdue. But safety instructions, variation approvals and disputes should remain under human control because the consequences of an incorrect interpretation can be serious.

Human resources is another area where caution helps. Automation can remind you about interviews, gather documents and prepare onboarding checklists. It should not independently reject applicants or make sensitive decisions based on incomplete information. You remain responsible for fair, consistent processes and for protecting personal data.

A practical verification test

Business task	Suitable automation role	Human check
Invoice processing	Extract fields and detect duplicates	Confirm supplier, amount and tax details
Customer enquiries	Draft replies and classify urgency	Review complaints, promises and exceptions
Stock management	Flag low stock and prepare reorder lists	Check demand, supplier timing and substitutions
Meeting administration	Summarise notes and assign action items	Confirm decisions, owners and deadlines
Staff onboarding	Send reminders and organise documents	Review access rights and personal information

The table reflects a useful operating principle: let software prepare, sort, compare and alert; let people approve decisions that are difficult to reverse.

Practical Takeaways

• Start with repetitive work: Choose tasks involving data entry, reminders, document sorting, status updates or basic calculations.
• Define the source of truth: Decide which system contains the approved customer, stock, order or staff information.
• Require review before external action: Keep a person in the loop before sending sensitive messages, changing records or confirming commitments.
• Use small pilots: Test one workflow with a limited group before applying it across the company.
• Record approvals: Keep a simple history of who reviewed an automated recommendation and what was changed.
• Set exception rules: Send unusual orders, angry complaints, large variances and missing information to a responsible employee.
• Protect confidential information: Do not place customer identity documents, payroll records or contract details into a tool without checking its privacy controls.
• Measure useful outcomes: Track processing time, correction rates, missed follow-ups and customer complaints rather than focusing only on how advanced the tool sounds.
• Train staff on limits: Make it clear that a fluent answer is not automatically a correct answer.

As a simple starting exercise, choose one process and map four steps: what enters the system, what the software does, what a person checks and what happens after approval. If any step is unclear, the workflow is not ready for full automation.

The Bigger Picture

The long-term lesson is not that AI has no place in business. It is that trust will become a more important part of automation. As tools become better at producing convincing text, images, recommendations and software, businesses will need stronger internal checks to distinguish useful assistance from unsupported confidence.

This may favour SMEs that build disciplined processes early. A small company does not need a large technical department to benefit. You need clear records, defined responsibilities and sensible approval points. These foundations also make ordinary automation more reliable, whether the tool uses AI or simple rules.

The current backlash is partly a reaction to technology being presented as a substitute for judgement. That approach is risky for any company, but especially for an SME where one incorrect invoice, missed customer promise or poorly handled complaint can consume valuable management attention.

Use automation to help your team see work sooner, organise information better and complete routine steps consistently. Do not hand over responsibility simply because a system can produce an answer quickly.

For your next automation project, ask one question first: “What evidence will tell us that this output is correct?” If the answer is clear, you may have found a good process to automate. If the answer depends on experience, context or a difficult-to-reverse decision, keep a capable person involved.

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 →