Recently, leaked information revealed that Amazon's key AI product, Amazon Q, faces numerous challenges in its early stages after launch and risks losing customers to Microsoft Copilot. This highly anticipated product encountered a series of issues after its public release in April.

Internal memos indicate that in August, some Amazon Web Services (AWS) employees expressed their views on Amazon Q in internal Slack channels, highlighting issues such as insufficient functionality, high costs, and difficulties integrating with other software. Some employees are concerned that these issues may drive customers to Microsoft's Copilot instead.

One AWS employee stated on Slack, "In my opinion, Q is currently only suitable for demonstrations and very controlled, simple application scenarios."

The difficulties encountered by Amazon Q reflect the challenges faced by commercial software developers in the era of AI. While AI technology brings much promise, it has also sparked fierce market competition, leading tech companies to rush to launch new products that often lack full functionality, have flawed internal tracking systems, and may have unintended IT impacts. Amazon's competitors like Microsoft and Google have also encountered similar issues during the launch of their early AI products.

For Amazon, this is a test of its AI technological prowess. Although the company expects to generate hundreds of millions of dollars in revenue from AI services this year, AWS's AI chip, which competes with NVIDIA GPUs, has seen low adoption rates. Additionally, attempts to apply ChatGPT-like technology to Alexa, the voice assistant, have also encountered obstacles.

Competition with Copilot

Internal information indicates that employees are concerned that Amazon Q's shortcomings could lead to customer loss to Microsoft Copilot. A common customer complaint is that Amazon Q cannot process images embedded in PDF files, making its performance "inferior to Copilot." Furthermore, customers will have to wait until AWS re:Invent in December to see product improvements.

Another issue is image generation. Employees have noted that Q lacks "multimodal capabilities," i.e., the ability to process both text and image files simultaneously, making it less appealing. One customer chose Copilot after discovering that Q could not create marketing campaigns involving images and text.

Moreover, some customers are dissatisfied with the cost of data integration with Amazon Q. For instance, one customer found that when integrating with Microsoft Exchange to enhance email search, Amazon Q lacked tools for selective import of email content. This resulted in an initial cost of approximately $400 per user per mailbox. In contrast, Microsoft Copilot offers "out-of-the-box" functionality to support this use case.

Ready or Not?

This question was raised on Slack.

Prior to Amazon Q's launch, some employees had warned that the launch process seemed rushed, with insufficient testing and over-reliance on manual review. This reflects Amazon's corporate culture of encouraging employees to be frank about issues to resolve them quickly. Other significant tech projects at the company have also faced similar challenges during their initial stages but eventually improved and succeeded. Amazon Q may follow a similar trajectory.

AWS spokesperson Patrick Neighorn told Business Insider that open discussion is part of Amazon's culture, encouraging employees to engage in "loud self-criticism" about their work. He added that Amazon Q is experiencing robust growth across various aspects.

"Loud self-criticism is central to earning our customers' trust, and we actively seek and value feedback from builders," Neighorn wrote. "We are seeing major customers like Bayer, Smartsheet, and National Australia Bank using Amazon Q, and we are also using Amazon Q extensively across our own business, including the AWS sales organization and Prime Video."

Fire Drill

Some AWS employees on Slack channels noted that the marketing language for Amazon Q was overly exaggerated and questioned whether the company needed to "tone it down" to avoid losing customer trust.

"This is a fire drill waiting to happen," one employee wrote on Slack. "All we can do is clean up the mess after the marketing department pushes customers towards tasks that Q cannot accomplish."

Neighorn stated that Amazon provides clear pricing for Amazon Q, and customers can use the AWS pricing calculator to estimate costs. He also mentioned that Amazon Q offers customers options to control which documents are indexed and how they are imported.

"Amazon Q was only launched in April, and we have been continuously adding new features based on customer feedback," Neighorn said. "Amazon Q has a rapidly growing customer base using Q for various purposes."

Poor Early Sales Data

Internal data also indicates poor sales performance for Amazon Q.

According to internal data obtained by Business Insider, one branch of the AWS sales team, comprising over 3,000 people, failed to meet Amazon Q's sales targets in July. These data, tracked by Amazon's senior leadership team, showed that Q's sales fell short of expectations across all regions, particularly in North America.

Moreover, Amazon has struggled to track the actual sales of this new AI product.

Neighorn explained that the internal sales data obtained by Business Insider came from "inaccurate preliminary numbers based on incomplete methods, which we corrected a few weeks ago." Business Insider is aware of specific sales figures and targets but has not disclosed them due to Amazon's revised methodology.

"Customers have shown great interest in Amazon Q, and it is being rapidly adopted since its launch four months ago, nearing our ambitious sales targets," Neighorn said.

$260 Million in Efficiency Gains

AWS CEO Matt Garman instructed part of his sales team of approximately 6,500 employees to undergo a half-day mandatory training on Amazon Q earlier this year. This training, aimed at building "confidence and skills" necessary for selling Q, covered everything from Amazon Q's branding and security features to supply chain and coding overviews.

Neighorn stated that sales training is "standard practice" for enterprise tech companies, and any claims to the contrary are "incorrect."

Garman holds high expectations for AI assistants like Amazon Q. In an internal speech in June, he mentioned that software engineers might soon no longer need to program due to advancements in such AI tools. Amazon is also developing another AI chatbot internally codenamed Metis.

Currently, Amazon CEO Andy Jassy is perhaps Amazon Q's biggest promoter. Last month, he posted on LinkedIn that Amazon's internal use of Q has helped the company achieve "significant efficiency gains," adding that it has saved "4,500 years of developer effort" and an estimated $260 million in "annual efficiency gains."

"This is a game-changer for us, and not only does our Amazon team plan to make greater use of this transformational capability, but our Q team also plans to add more transformations for developers to leverage," Jassy wrote.