Writing Crystalized Thinking At Amazon. Is AI Muddying It?

Before AI, human writing at Amazon was sacrosanct. The company began each big meeting with six-page narrative describing the product or feature, typically written by the project lead, read in silence before anyone spoke. The writing’s purpose was to crystalize thinking and anticipate every scenario. Powerpoint, the enabler of logical leaps, be damned.

“The document should be written with such clarity that it’s like angels singing from on high,” Jeff Bezos once said. “I like a crisp document and a messy meeting.”

But now, Amazon is in its AI era, and its leadership is encouraging employees to let AI do the writing for them. The company’s internal marketing for Cedric, its ChatGPT-style tool, promises “six-page narratives in seconds.”

The implications for Amazon’s culture struck me as so profound that I reached out to over fifteen current employees, a mix of Bezos-era veterans and newcomers, to explore what the mandate says about their work lives, the company’s priorities, and what it means to be “Amazonian” today.

The worry within the company, I learned, is that Amazon is losing sight of writing’s centrality in its deliberative, thoughtful culture as it pursues powerful, new tools.

“Writing is thinking,” said one longtime company veteran. “That was the whole point of Amazon’s writing culture. I can’t tell you how many times I changed my mind when writing a narrative. And even when I didn’t, my arguments were more precise for having written them down. Now we have chatbots writing six-pagers to be summarized by other chatbots.”

Heavy Pressure to Use Inadequate Tools

Within Amazon, even those who expressed enthusiasm for generative AI described Cedric and other internal AI tools in terms like “comically inadequate.” For security reasons, Amazonians aren’t allowed to use external tools like ChatGPT.

“It has yet to return a single correct answer to me,” said one AWS employee, while another described it as “constantly hallucinating and then apologizing for being wrong.”

“I could use it to write docs, yeah,” said one self-described fan of human-centered AI. “But I don’t trust the output, so I’d rather just write it myself versus have to double-check every detail.”

Still, employees described heavy pressure to write documents via AI, whether or not the tools are up to par. “We get grilled in every team meeting about whether we’re using it,” said one junior Retail employee, while others receive weekly reminder emails from their VPs.

What they aren’t getting from leaders is an understanding of why machines should be doing their writing, beyond saving time or hand-wavy platitudes like “we should be using AI in everything we do.”

Employees are craving training and best practices for using it intelligently. In an open letter, CEO Andy Jassy encouraged Amazonians to attend company AI workshops and team brainstorming sessions. But when I asked employees about those, I got mostly head shakes or sardonic chuckles in response, suggesting that training is either spottily available or people don’t know it exists.

Amazon is a famously data-driven company, and the people I spoke with also expressed frustration that they aren’t hearing from leadership about how the impact of generative AI on the business is being measured qualitatively.

“There’s a communication vacuum,” said one Stores employee. “You can’t just say ‘Do it because I said so’ to hundreds of thousands of people with no real direction or training and expect to succeed.”

Amazon did not return a request for comment.

Resistance and Resignation

During these conversations, I began to mentally split employees into two general camps, the Resisters and the Resigned. The Resisters, many of them old-timers, see the AI push as both a symptom and an accelerator of declining standards. “I’m appalled at half-baked products being passed off as ‘polished’ work,” said one company veteran. “Trainings with incorrect information. Business reviews submitted with no narrative to put the data in context. [Documents] that the owner can’t even speak to. The list goes on and on.”

The Resigned tend to be newer and more junior, with less autonomy over how they work. “My manager is asking for longer documents now, because he says he can use AI to summarize them instead of reading the whole thing,” said a Retail employee who joined Amazon last year. “How is that an improvement over writing a quick one-pager that people actually read?”

Like other Resigneds, she’s complying for fear of falling behind if she doesn’t. “Expectations for volume of work keep rising,” said a product manager, “especially after these last few years of layoffs. If my teammates are using AI for their docs and I’m not, I’ll just end up in Focus [Amazon’s performance management tool] for failing to keep up.”

Other Resigneds say that expectations for quality of thought and communication simply aren’t as high as they used to be. One longtime product manager said it’s “screamingly obvious” when a document was AI-authored because it’s full of buzzwords and the kind of “excited marketing intern” language that used to earn a slap on the wrist from hype-allergic leaders. “A [written document] is basically just a box to check now. I’m not sure how much the quality really matters.”

Story continues below…

Promevo CEO Karthik Kripapuri on Google’s AI Comeback, Enterprise Agents, The Real Path to AI ROI (sponsor)

Karthik Kripapuri, CEO of Promevo, joins Big Technology to discuss how companies are finding real ROI today from cutting-edge AI models. Tune in to hear what’s changed on the ground with Gemini and Vertex AI, how companies find ROI by starting small with clear KPIs, and why data integrity is still the biggest blocker to agents delivering value. Hit play for a practical roadmap to what’s working now in enterprise AI and what it’ll take to scale it.

Watch Now

A “Grand Experiment”

As someone who found Amazon’s intellectual rigor rewarding and exhilarating in my twelve years at the company, all of the above made my head spin.

Worried that I might be sentimentalizing the past rather than understanding today’s Amazon on its own terms, I turned to two long-tenured executives for their perspective. ...