Why Tesla’s Pothole Story Matters to Your Business
A promise that sounds small can reveal a much bigger technology problem. Tesla CEO Elon Musk has again said that Tesla’s Full Self-Driving software will soon be able to avoid potholes, even though a similar promise was reported in 2019. Read the source report
For you as a Malaysian SME owner, the important lesson is not whether a Tesla can detect a damaged road. It is how you evaluate ambitious AI claims before they become part of your operations. Whether you run a courier business in Shah Alam, a retail outlet in Johor Bahru, a workshop in Penang or a small professional-services firm in Kuala Lumpur, new AI tools are constantly described as “coming soon”, “fully automated” or “human-level”. Your business still needs reliable results today.
AI can help with customer replies, stock updates, document processing, scheduling and sales follow-ups. But a tool that works well in a demonstration may behave differently on Malaysian roads, with local language, mixed-quality data, unusual customer requests and manual processes. Tesla’s pothole story is a useful reminder to separate progress from promises.
What Happened
According to Electrek, Musk responded on social media to a request for Tesla’s driving software to avoid potholes by saying the capability was “coming soon”. The report notes that Tesla had made a similar statement about pothole avoidance in 2019, but the feature had still not clearly arrived by 2026. Read the source report
The report also acknowledges that Tesla’s software has improved. It says the system can navigate many roads, reach destinations and park in some situations, while Tesla’s autonomous driving development continues alongside its robotaxi plans. However, the article highlights a gap between broad claims about AI capability and the software’s performance on a very ordinary physical problem: identifying a pothole and safely changing position to avoid it. Read the source report
The timing is significant because the report says Tesla was preparing to launch its Cybercab later that week as a geofenced autonomous vehicle without a steering wheel. That raises a practical question: if an autonomous vehicle still faces common road hazards, how much confidence should businesses place in claims about wider autonomy? Read the source report
Why This Matters for Malaysian SMEs
Malaysian businesses operate in environments that are often more varied than a controlled technology demonstration. A delivery team may face heavy rain, faded road markings, motorcycles filtering between lanes, narrow shop-lot access roads, construction diversions and inconsistent addresses. A software tool may claim to automate dispatch, but you need to test whether it handles condominium access instructions, kampung roads, multilingual customer messages and last-minute order changes.
Consider a small food distributor. An AI system might predict demand and recommend replenishment. That is useful, but you should still check how it treats public holidays, school breaks, Ramadan, local festivals and sudden weather disruptions. Malaysia has 13 states and three federal territories, and official government information describes the country’s administrative structure accordingly. See Malaysia’s administrative information Local operating conditions can vary considerably, so a generic model should not be trusted without local testing.
For a service company, an AI chatbot might answer common questions. You should verify whether it understands Bahasa Malaysia, English, Manglish, abbreviations and industry-specific terms. It should know when to hand a conversation to a person instead of confidently giving an incorrect answer. The same principle applies to automated invoice extraction, HR screening, appointment booking and sales qualification.
Use AI promises as hypotheses to test, not as operating procedures to adopt immediately.
A practical pilot should measure the tasks that affect your daily work. For example, if you are considering AI for customer support, review a sample of real enquiries and track answer accuracy, escalation quality, response time and staff correction effort. If you are considering AI for delivery planning, compare suggested routes against actual driver experience during rain, peak traffic and restricted-access deliveries.
| AI decision question | What you should check |
|---|---|
| What exactly is being promised? | Define the feature in observable terms, such as “flags duplicate invoices” rather than “automates finance”. |
| Where does it work? | Test Malaysian locations, languages, documents, customer behaviour and operating conditions. |
| What happens when it is wrong? | Require approval steps, audit logs, correction tools and clear human escalation. |
| How will you judge success? | Set measurable targets for accuracy, turnaround time, exception rates and staff workload. |
| What is the fallback? | Keep a manual process available for outages, unusual cases and incorrect recommendations. |
The Bigger Picture
The Tesla example points to a wider shift in how businesses should assess AI. The most valuable question is no longer simply, “Can this system do the task?” It is, “Can this system do the task consistently, in my environment, with acceptable supervision?” A vehicle avoiding a pothole is a physical-world example. An SME approving a supplier invoice, replying to a customer or updating inventory faces the same reliability challenge in a business setting.
AI systems improve through data, software updates and feedback, but improvement does not automatically mean readiness for every use case. The Electrek report refers to Tesla’s large training-data claims while noting that the specific road-hazard capability remained unresolved. Read the source report For your business, a large volume of data is not enough if the data does not represent your customers, documents, products and exceptions.
You can apply a simple “promise-to-proof” process before adopting any AI tool:
- Write the promise down: Record exactly what the vendor says the system can do.
- Choose a narrow workflow: Start with one repetitive process, such as extracting purchase-order details.
- Use real examples: Test actual documents and customer messages, after removing sensitive information where appropriate.
- Measure exceptions: Count errors, missing fields, incorrect classifications and cases requiring staff intervention.
- Assign ownership: Name a staff member who reviews results and reports recurring problems.
- Set a review date: Decide whether to expand, adjust or stop the pilot based on evidence.
Malaysia’s Personal Data Protection Department provides guidance and information relating to personal-data protection, which is relevant when you place customer, employee or supplier information into digital systems. Visit the Personal Data Protection Department Before using an AI service, check where data is stored, who can access it, whether it is used for model training and how you can delete or retrieve it.
The practical takeaway is straightforward: do not reject AI because some promises take longer than expected, and do not accept AI because a prominent founder says a feature is coming soon. Start with a narrow business problem, test it using Malaysian conditions and keep human control over decisions that affect customers, compliance or operations.
For your SME, dependable automation will usually beat dramatic automation. A system that correctly processes routine work every day is more valuable than an impressive platform that claims to transform the entire business but still struggles with basic exceptions. Tesla’s pothole promise gives you a timely reminder to ask for evidence, define success and make reliability the centre of every AI decision.
Ready to Streamline Your Operations?
Technology moves fast. Your operations should keep up. AutoRunBiz builds AI systems that run your daily workflows — from WhatsApp order capture to accounting. Book a free 15-min ops audit →