On Tuesday, at an event called “What’s Next with AWS” in San Francisco, CEO Matt Garman took to the stage to announce that AWS is moving up the stack (for the seventh, eighth, or ninth time, depending on how you count) into the applications business.
Batting line:
- amazon quickformerly known as QuickSight, formerly known as Q Developer, formerly known as Q Business, formerly known as Q. This feels pretty close to their fourth rebrand in about 18 months. The press release states that it will be “generally available today.” Garman said on the same stage that the desktop app is in “preview stage.” In his keynote, Amazon Quick’s vice president said only that the company was “unveiling new technology that will change the way we work,” without specifying whether that meant GA or preview. Please select a lane. please?
- Amazon Connect decisionssupply chain planning products. Competitors include Workday, o9, Blue Yonder, Kinaxis, and SAP.
- Amazon Connect Talenta recruitment product that utilizes AI. Competitors include Workday, SAP SuccessFactors, Greenhouse, Eightfold, and iCIMS.
- Amazon Connect Healtha clinical AI product. As you may know, the leading incumbent is Epic, which has over 40% EHR market share and strong ambient partnerships with Abridge, Ambience, and Microsoft’s Nuance.
- Amazon Connect customersa rebrand of the original Amazon Connect contact center product. Existing customers who logged in this morning will notice that their console has been renamed without consultation. AWS described this on stage as “the birth of a new family.” Whether the existing Connect customer base interprets it that way is another matter.
- And here’s what grabs the headlines: Bedrock Management Agent Powered by OpenAI,plus Codex on AWSIn addition, a multi-year agreement has been signed for OpenAI’s AWS training capabilities. Sam Altman appeared in a pre-recorded video and stated (in his own words) that his schedule was “stolen from me today.” It’s a rare and candid look at the court date in the mask case that took up an afternoon. The video itself was about five minutes long and clearly a one-take, showing the OpenAI CEO tripping over the teleprompter at least twice while AWS announced a partnership that had just been treated as a major product announcement for the previous half-hour. The mood was, “I have a court date in 20 minutes. Please record it and send me the file.” Whose partnership is this exactly?
Did you understand everything? good. Now we can talk.
A quick tour of the application cemetery
To understand why this is important, it helps to recall AWS’s track record with applications, software that is intended for use by humans, rather than infrastructure intended for use by engineers who build software for humans.
Not complete tour:
- Amazon WorkMail: Announced in January 2015, support ends on March 31, 2027. A decade of answers to questions no one asked.
- Amazon Work Documents: Launched as Zocalo in July 2014, renamed in January 2015, and permanently shut down on April 25, 2025. The phrase “WorkDocs Drive” still gives a small wince to anyone who sees it.
- amazon honey code: Launched in June 2020, beta ended on February 29, 2024. Published for less than 4 years.
- amazon work link: Started in January 2019 and permanently closed on November 30, 2021. It took two years and 11 months from cradle to coffin.
- amazon chime: Launched in February 2017 and ends support on February 20, 2026. It was quietly phased out and then shut down in a big way.
- original amazon q: Announced in November 2023, version history is surprisingly difficult to enumerate without a whiteboard. Now reborn as Quick Suite (October 2025), it has absorbed Q Business and rebranded QuickSight in the process.
- amazon quick site: Announced in October 2015, generally available in November 2016, changed to Quick Suite on October 9, 2025. It has never visibly threatened Tableau or Power BI. Now it technically doesn’t exist by its own name.
One of the applications that AWS has shipped, continues to ship, and is actually purchased by customers at scale is Amazon Connect, a contact center product. Aubrey rightly pointed out on stage that last year, Connect processed 20 million customer interactions per day and 12 million AI minutes (probably similar to regular minutes, only much more expensive). Connection is good. Connect customers are happy. The connection works.
It’s worth asking why.
Why Connect was successful and the rest were not?
Contact centers are essentially infrastructure disguised as applications. This is a routing and recording system with a phone and a dashboard. Routing is where AWS excels. Records are where AWS excels. Once you frame it as “a call is a stream of packets that needs to be reliably processed and stored,” that becomes the kind of problem that AWS can bring its infrastructure power to and win.
WorkMail was different. You need to know not only how to extend an SMTP server, such as calendars, email rules, Outlook integration, address book synchronization, etc., but also how humans actually work. Microsoft is investing 15,000 person-years in Outlook. AWS launched WorkMail using the AWS Console as the management UI. The results were very predictable and very depressing.
This pattern repeats. Honeycode wanted the flavor of an application builder, like the one Glide and Retool spent years developing, not the flavor of an Amazon website. WorkDocs required us to use Google’s Google Docs. The chime required me to zoom out. Each time, AWS delivered infrastructure-grade capabilities for problems that required application-grade sense, and each time, AWS underperformed. Connect was an exception because contact centers have been infrastructure from the beginning.
Same bet with better wrapping
Connect Decisions wants to compete with 20- and 30-year-old domain leaders in supply chain planning. The press release’s headline customer metric is “$85,000 in excess inventory savings in the first quarter” for a company called Wells Vehicle Electronics. Context: SAP, o9, Blue Yonder, and Kinaxis regularly cite 7- and 8-figure savings per implementation. From a supply chain perspective, the $85,000 amount is well within the range where “the costs of running the planning meeting exceeded the savings.” On stage, Wells and TVS Motors, Decisions’ sole customer, were described as “partners” who “bring their products to meetings and discussions,” a frank expression that could be read as design partner language. You know, pre-revenue.
The product page for Connect Talent (the only one of the four new Connect SKUs actually live as of this writing) says, “Now available in preview.” The section explaining AI-driven voice interviews says nothing about New York City Local Law 144, the EEOC’s 2023 AI Employment Guidance, or the EU AI Act’s High Risk Systems Rule. For an AI-powered product launching in 2026, this is a notable lack.
To be fair, Connect Talent’s launch is clearly positioned for mass hiring in retail, hospitality, customer support, and logistics, in addition to Amazon’s own seasonal warehouse staffing scale. This is not intended to be an AI screening of engineering candidates, and it would be unfair to characterize it as such. But it targets precisely the amount by which bias is most effectively amplified, and the very scenario that Local Law 144 was written to address. “We’re just doing this at scale” is not an allegation of non-compliance. That’s why a compliance attitude is paramount.
The page promises that “all recommendations include clear, job-relevant rationale and consistent evaluation criteria to demonstrate best practices when making hiring decisions and help organizations comply with hiring requirements.” In other words, the customer is responsible for compliance. Vendors are not volunteering anything specific. This is a land mine waiting for someone to trip over, and the Terms of Service makes it clear that it’s the customer’s responsibility, not AWS’s. Connect Health took to the stage with one launch partner (NetSmart, a behavioral health and post-acute IT vendor) and a pre-recorded video of doctors complaining about the burden of documentation. If you’re a health system CIO and you watched the keynote expecting to hear a story about Epic’s integration, you saw what was left unsaid.
Who exactly is the customer?
The most powerful workflows demonstrated on stage all starred Amazon employees.
A quick desktop demo walked us step-by-step through preparing for a customer meeting, pulling in our roadmap files, account team emails, and internal Slack threads “from my machine.” It’s an AWS field representative preparing for a customer call. Connect Decisions’ pitch relied heavily on “Amazon’s 30 years of supply chain experience” and a team that manages 400 million SKUs. The traditional story of Connect Talent is Amazon’s unique seasonal hiring scale. Connect Health’s flagship deployment is One Medical, an Amazon subsidiary. Garman/White’s most powerful agent development anecdote is how Amazon’s Prime Video team rewrote its internal payment system in two quarters with a small number of people.
This is consistent with the explicit framework on stage. AWS built these tools to run Amazon’s own business and now provides them to customers. It’s an honest origin story. This is also the same starting point as WorkMail, Honeycode, and Chime. This tool was built for Amazon’s own use and projected externally in the hope that the rest of the market wanted the same thing, but it turns out that’s not the case at all.
It remains to be seen whether AWS will be able to capture external markets on its own with this generation of products. Current demos suggest that the most powerful customers are also vendors.
And then there’s something called “eumorphism.”
On stage, Connect’s new design philosophy was introduced as follows. humorism: A term coined as an answer to the “dynamics of human interaction” to the “physics of objects” of skeuomorphism. This is an analogy that, on its own merits, cannot survive exposure to sunlight. Skeuomorphism described a real industry-wide visual translation pattern that has been employed for decades. Humorphism describes a company’s UX intent and asks us to follow that analogy. Following AWS naming style, it starts with the root word “humor,” but it’s not funny at all.
Anyone who names this will, if they’re lucky, be promoted to a position with no naming responsibilities.
And on the other hand, humans
It would be a shame not to mention this. AWS invested billions and billions of dollars into Anthropic. Today’s keynote focused on a brand new, unique agent SKU for OpenAI using Codex on AWS, AWS training capabilities for OpenAI, and the aforementioned 5-minute Altman teleprompter output.
Anthropic’s “moment”, in case you missed it, arrived a week ago in the form of a quietly published AWS Machine Learning blog post announcing that Claude Cowork and Claude Code Desktop are now running on Amazon Bedrock. There are no keynotes or Dalio videos, just a blog. I’ve launched various crap posts with great fanfare.
AWS made a multibillion-dollar Anthropic investment and covered the corresponding Bedrock launch in a blog post on Tuesday. AWS spent eight weeks announcing a partnership with OpenAI, giving the company a keynote anchor, and a five-minute video from a CEO who clearly has a better place. This may be intentional news management. Anthropic may also have a great opportunity because OpenAI is a partner in a model that AWS customers have been clamoring for. Both articles are interesting, but neither makes it particularly easy for AWS marketing teams to keep a straight face for the rest of the year with their claim that “we are a place where you can run both major frontier models.”
kicker
AWS is great for infrastructure. AWS also believes in scale instead of domain depth in adjacent markets, and is increasingly becoming a $100 billion-plus business. Historical evidence says it cannot. Connect was an exception. Because contact centers have been infrastructure from the beginning. Today’s bet is on the Connect family of four products, the first Q3 rebrand in 18 months, the OpenAI anchor, expanding the Connect brand to three new buying centers, and assuming bets will become commonplace.
Sign up for Quick Desktop today. This could be really useful, and early adopters (some of whom I trust) claim so. I’ve tried it quite a bit and will tell you what I discover. But the price of this entire launch bet should be based on 20 years of AWS application history. The cemetery is full. New tenants come with impressive nameplates.
Place your bet accordingly. ®
