Amazon Starts 90-Day Reset To Stop Constant Tech Outages

Amazon has begun a 90-day internal review of its coding processes after several system outages affected customer orders on its platform. The move is being described inside the company as a “code safety reset” and will apply to some of its most important engineering systems.

The step comes at a time when the company is making wider use of artificial intelligence (AI)-assisted coding tools to help engineers write software faster. While these tools have increased development speed, they have also created challenges in maintaining stability across large systems.

Dave Treadwell, Amazon’s SVP (Senior Vice President) of e-commerce services, told employees that such incidents have emerged since the third quarter of 2025.

New Checks For Critical Systems

As part of the reset, Amazon plans to introduce additional checks before software changes are rolled out to production systems.

The measures will focus on around 335 “Tier-1” systems. These are systems that directly affect customer activity, including placing orders, browsing products and checking delivery timelines.

According to the internal communication, engineers will now need two reviewers to approve any major code change. They must also record details of the update using Amazon’s internal documentation and approval tools.

Testing requirements are also being strengthened. The goal is to identify problems earlier and prevent disruptions that could affect customers.

Amazon has increasingly allowed engineers to release more code because AI tools can generate large sections of software quickly. However, this has also made traditional review processes harder to manage.

AI systems are not always perfectly consistent. The same request can sometimes produce slightly different results. In complex platforms such as Amazon’s retail system, even small variations can create unexpected issues.

Outages That Triggered The Reset

The review follows a number of recent disruptions. One of the major incidents happened on 2 March, when customers were shown incorrect delivery timelines during checkout. Internal data suggested the issue generated around 1.6 million system errors and led to roughly 120,000 lost orders across Amazon’s marketplaces.

Another outage occurred on 5 March. That disruption reportedly resulted in about 6.3 million orders being lost.

Internal records cited in the report indicated that one of the incidents was linked partly to Amazon’s AI coding assistant called Q. In another case, a production change was deployed without going through the full documentation and approval process.

Amazon said the reset is intended to stabilise its systems while it continues to expand the use of automation and AI tools in software development.

Also read: India To Launch $11 Billion Fund To Accelerate Chipmaking Push

Impact On Technology Companies

For people who invest in tech stocks, this is a warning. Many companies are trying to replace human workers with AI to save cash. Amazon’s recent trouble shows that AI cannot work alone yet. It still needs human eyes to prevent massive losses.

If other tech firms face similar bugs, they might also have to slow down. This could be good for traditional IT service firms that use skilled human engineers.

Investors should look for companies that use AI carefully rather than those that just try to be the fastest. In the long run, firms that balance technology with human oversight can build more reliable systems. That approach can matter more than rushing to adopt every new AI tool.

Sources:

NDTV

The Times of India