Amazon is committed to implementing AI across the board, even if it hampers productivity.

Dina, a software developer residing in New York, embarked on her journey at Amazon two years prior, tasked with code writing. However, her role has swiftly shifted. Now, she spends a significant amount of her time rectifying the errors produced by artificial intelligence systems.

The internal AI tool in question, dubbed Kiro, frequently gives rise to errors and generates incorrect code. Dina finds herself in a constant loop of investigating and fixing the flawed code this AI produces, or often reverting to earlier versions to begin anew. “It feels like I’m trying to navigate a problem caused by AI, using AI,” she lamented.

“Many of us in my team don’t genuinely feel these tools enhance our speed,” Dina noted. “Yet, management insists we must accelerate our pace, proclaiming that this will facilitate faster delivery, emphasizing that speed is paramount.”

Shortly after sharing her thoughts with The Guardian, Dina received the news of her layoff.

Meanwhile, Lisa, a supply chain engineer with over ten years under her belt at Amazon, shared that her experiences with AI tools have proven beneficial in only about a third of her attempts. Even in those successful instances, she finds issues that require her to verify and discuss with colleagues, consuming more time than if she had executed the tasks without AI.

While she acknowledges the potential of AI tools, she critiqued the company’s insistence that all employees utilize them daily. “It’s not about thinking, ‘How can I use this hammer I have?’” she said. “It’s about assessing whether this is a job for a hammer or perhaps another tool entirely.”

More than a few current and former Amazon employees, ranging from software engineers to data analysts, have voiced their concerns regarding the company’s push to have all employees incorporate AI into their work routine. Many reported that this directive is, in fact, detracting from productivity. These employees described Amazon’s approach to AI use as disorganized and expressed their thoughts that the company may be utilizing their labor to enhance the capabilities of its future AI replacements, leading to significant feelings of demoralization. Due to concerns regarding potential professional repercussions, The Guardian honored their requests for anonymity.

Montana MacLachlan, an Amazon spokesperson, stated, “We have a large number of corporate employees in varied roles, each using AI in different contexts to discover what methodologies suit their specific needs best. While individual experiences can differ, the majority of our teams report significant value from the AI tools they integrate daily.”

This pressure comes during a time when Amazon has laid off 30,000 employees in just four months — a reduction of nearly 10% of its approximately 350,000 corporate workforce. These cuts follow a wave of tech layoffs associated with AI, impacting companies such as Block, Pinterest, and Autodesk. The extent to which these companies anticipate reliance on AI to replace human roles remains vague; various organizations have given conflicting explanations for their staff cuts. Jack Dorsey, CEO of Block, directly attributed his 40% staff cuts to AI, while Pinterest and Autodesk mentioned reallocating investments toward AI. Amazon’s explanations regarding how AI impacts its job cuts have varied, indicating both a potential for reductions paired with claims that recent layoffs were not AI-induced. The company did announce plans to allocate around $200 billion this year towards AI infrastructure and a separate $50 billion investment in OpenAI.

With the rising concerns regarding AI’s influence on employment, the strategies Amazon implements concerning automation — as well as how it communicates these changes — will significantly affect not just its vast workforce but likewise individuals across various sectors globally. Recognized as the second-largest employer in the U.S., Amazon has dramatically shaped workplace standards in sectors spanning both white-collar and blue-collar industries.

“There’s a significant discourse among corporate employees surrounding how such practices — related to performance monitoring and oversight — seem to originate from the warehousing and delivery spaces and whether Amazon aims to extend this labor model to professional employees,” commented Jack, a long-term software engineer at Amazon. “It feels like we’re entering a new chapter in employer-employee relations with the advent of AI.”

Although Amazon has gained a reputation for having a demanding work environment, the effects of its AI initiatives have notably amplified workplace tensions, as reported by employees. “It feels worse now,” expressed Denny, a software engineer in the retail division. “If we don’t adapt … we run the risk of being deemed unnecessary during the next round of layoffs.”

Each time a task arises, the predominant inquiry from management pertains to whether AI can expedite the task at hand. This has led employees to incorporate AI tools simply for the sake of it. Recently, a colleague revealed that an internal AI agent helped him save around a week’s worth of development time for a feature. Yet upon reviewing the code, Denny identified numerous comments from peers indicating basic flaws in the AI-generated code — the quality was notably poor.

“Ultimately, I suspect that the developer cycle may not change, and could even lengthen,” Denny noted. “The pressure to utilize AI has led to inferior quality code and increased workload for everyone.”

Denny was among numerous employees who informed The Guardian of the pressures to utilize an overwhelming array of AI tools that were hastily created during internal hackathons, alongside the need to complete surveys regarding their experiences with these tools.

“My manager frequently presented me with these random tools from hackathons, insisting I give them a try,” Denny shared, describing the tools as “half-baked” and counterproductive, often adding to his responsibilities as he had to assess their functionality.

Amazon routinely conducts quarterly hackathons to stimulate engineers in creating new projects. Denny recalls that the company pivoted primarily towards generative AI hackathons last year, resulting in the majority of projects focusing on enhancing developer productivity.

“We don’t mandate teams to use AI tools,” MacLachlan stated. “However, we believe these tools can significantly assist employees in working more efficiently and automating tedious, repetitive tasks.”

Public setbacks related to Amazon’s AI adoption have also surfaced. A report from February in the Financial Times revealed at least two outages linked to its internal AI tools, including a significant 13-hour interruption affecting customer-facing systems after an engineer permitted an AI tool to make certain changes. Amazon later clarified that an employee, not the AI, was responsible for the disruption. The Financial Times also reported that Amazon is set to gather engineers to address the “spate of outages, including incidents related to AI coding tools.”

“If the push to utilize AI tools spans every aspect of work, errors like this will likely increase,” Sarah, an Amazon software engineer, expressed.

While Sarah sees the value in AI, she believes that engineers should dictate its utilization. Unfortunately at Amazon, she has found herself in a role of training the AI. “We have to delineate detailed processes for the AI to comprehend and yield improved outputs,” she explained. “My role increasingly feels like one of training AI that could ultimately replace our jobs.” Being early in her career, she also fears that relying on AI to facilitate her tasks stunts her professional development.

Ifeoma Ajunwa, the founding director of the AI and Future of Work Program at Emory University, commented on the trend of mandating tools that often leads to backlash. “Employees are typically in a better position than management to determine which tools bolsters productivity,” she said.

Many Amazon employees find they have to seek additional training for AI best practices independently.

Will, a user experience researcher, reported that although Amazon provides an extensive collection of AI training videos on learning portals, most are optional. While attending training sessions, he found the emphasis consistently centered on “how to build something quickly.” Trainers, usually skilled peer employees, cautioned participants to meticulously evaluate each aspect before permitting the AI to proceed. Although he has encountered sessions where the trainer instructed to let AI check its own outputs, he remains hesitant, knowing that its accuracy isn’t infallible.

“The degree of success in adopting AI and employee perceptions of its productivity enhancement are significantly influenced by management’s encouragement and training availability,” Alex Imas, a behavioral science and economics professor at Chicago Booth noted.

MacLachlan confirmed that Amazon does provide varied training resources across the organization, reflecting a structured approach. “We encourage employees to use these tools as a learning instrument, employing a learn-as-you-work strategy which has proven to be one of the most effective methods of adopting AI throughout the company,” she stated.

An AI-fueled shift to surveillance

In addition to the productivity challenges posed by Amazon’s AI initiatives, employees have reported a rising sensation of being monitored.

Historically, Amazon employees have begun their workday greeted by mechanisms like Amazon Connections, which solicit feedback on teamwork dynamics and overall job satisfaction. However, over the last year, these inquiries have increasingly addressed AI usage, moving away from human-centric topics.

Maria, a product manager let go in January, noticed the shift toward questions such as: “Are you using AI in your daily activities? How frequently? Do you regard yourself as a power user? Is AI a focus within your team?”

Further, employees expressed concern about obvious surveillance indicators. Managers have access to dashboards that keep track of their subordinates’ AI usage, monitoring what tools they utilize and their frequencies (reported first by The Information in February).

Jack, another long-standing software developer, has witnessed the launch of new dashboards specifically focused on assessing generative AI adoption, engagement, and usage levels within teams. “Approaches vary widely across teams,” he noted, with specific groups aiming for a 80% utilization rate of AI tools on a weekly basis.

Sarah relayed that her principal engineer expressed he reviews this monitoring dashboard daily, consistently pushing for increased AI utilization within the team.

“Understanding the tools our teams are using and how well those tools aid their productivity is critical,” MacLachlan emphasized.

Nick Srnicek, an expert in digital economy and author of Platform Capitalism, argued that the extensive deployment of AI often leads to heightened surveillance. “The rapid rollout of AI leads to an uncritical enhancement of oversight, as these technologies require comprehensive insights into personal workflows,” he cautioned. “Improving AI capabilities necessarily means providing management with greater insight and authority over workers’ daily functions.”

Moreover, many employees are increasingly convinced that their career progression relies heavily on their enthusiastic adaptation to AI.

“Our promotion documents now include a segment querying, ‘How does this person leverage AI?’” stated Lisa. “They seem to be inclined towards retaining individuals who embrace this AI initiative and distancing themselves from those who have reservations.” Reports indicate that at Amazon, “managers regard AI engagement when considering promotions,” as highlighted by a report from the Wall Street Journal.

Contrarily, MacLachlan explained that while the company anticipates employee engagement with available resources — including AI — to cultivate a more enriching work environment, they have not directed managers to assess AI usage as part of their evaluative processes. “Our focus remains on fostering AI adoption and disseminating best practices to champion innovation and efficiency across the organization,” she clarified.

However, Amazon’s CEO, Andy Jassy, has been vocal regarding his expectations surrounding AI for his employees. In a company-wide email last June, he foretold that AI-enhanced productivity gains could lead to a reduction in the corporate workforce, urging employees to integrate AI into their daily responsibilities. “Educate yourself, attend workshops, and leverage AI whenever feasible,” he advised, emphasizing the need for teams to brainstorm innovative approaches to better serve customers while optimizing productivity.

The unspoken math

Jassy’s company-wide email incited notable internal dissent last summer, as employees criticized his leadership and expressed frustrations over the demoralizing effects of Amazon’s AI initiatives, according to Business Insider. Months later, a petition was signed by over 1,000 workers raising concerns about what they termed the company’s “aggressive rollout” of AI tools.

In the backdrop of Amazon’s decision to lay off thousands, the company eagerly shares its increasing revenue figures each quarter. While Jassy has frequently asserted that these layoffs are neither driven by financial necessity nor AI initiatives, for employees like Maria, these developments make sense.

“If you automate away two hours of a person’s job, you must find a way to translate that into cost savings for that position,” Maria expressed, summing up what she described as the underlying logic motivating job reductions. “That’s the unspoken math informing their decisions.”

Jack continues to reflect on remarks made by Jassy during a company-wide gathering last spring. A Business Insider report on that meeting noted that when questioned on operating Amazon as “the world’s largest startup,” Jassy emphasized that they aim to be “scrappy” and to “achieve greater outcomes.” He also cautioned that competitors are the “most technically proficient, the most ambitious” companies, including startups working tirelessly each week.

“Those comments combined carried an implicit threat that employees remaining must dedicate more effort and longer hours,” Jack remarked. “It struck me that should Amazon fail to amass profits through continuous growth, they would further extract from the existing workforce.”