Redmond, Wash. – If AI were a religion, this would probably qualify as a cathedral. On the edge of Microsoft’s headquarters, overlooking Lake Bill amid a stand of evergreens, a new four-story building has emerged as a destination for business and tech decision-makers. Equal parts briefing center, conference hall, and technology showroom, Microsoft’s “Experience Center One” offers a curated glimpse of the future — guided tours through glowing demo rooms where AI manages factory lines, models financial markets, and helps design new drugs.
It’s part of a larger scene playing out across tech. As Microsoft, Google, Amazon and others pour billions into data centers, GPUs, and frontier models, they’re making the case that AI represents not a bubble but a business transformation that’s here to stay. As the new center shows, Microsoft’s pitch isn’t just about off-the-shelf AI models or run-of-the-mill chatbots — it’s about custom agentic systems that act on behalf of workers to complete tasks across a variety of tools and data sources. That idea runs through nearly everything inside the facility, a glass-encased building featuring an elevated garden in a soaring open-air atrium, just across from Microsoft’s new executive offices on its revamped East Campus.
Experience Center One highlights what Microsoft calls “frontier firms” — ambitious companies using AI to push their operations to the edge of what’s possible in their industries. Agentic AI is “fast becoming the next defining chapter of a frontier organization,” said Alysa Taylor, Microsoft chief marketing officer for Commercial Cloud and AI, in an interview.
The underlying message is clear: get on board or risk falling behind, both competitively and financially. A new IDC study, commissioned by Microsoft, finds both opportunity in spending big and risk in not being bold enough. Companies integrating AI across an average of seven business functions are realizing a return on investment of 2.84 times, it says. In contrast, “laggards” are seeing returns of 0.84 times — basically losing money on their initial spend.
The divide extends to revenue, too: 88% of frontier firms report top-line growth from their AI initiatives, compared to just 23% of “laggards,” according to the IDC study.
And hey, somebody has to foot the bill for those multi-billion-dollar AI superfactories. For this second installment in our Agents of Transformation series, GeekWire visited the new Microsoft facility to see first-hand how the company is presenting its vision of the future. Here are some of the takeaways from the sampling of demos we saw.
These are not off-the-shelf solutions. Each demo reflects a custom deployment built with a major customer, showing how AI tools can be tailored to specific business problems. For example, one shows how Microsoft has worked with BlackRock to integrate a custom AI copilot inside the investment firm’s Aladdin platform to help analysts process large volumes of client and market data more efficiently. It helps reduce the manual work of gathering data and points analysts to potential risks sooner than they might have spotted it on their own. As another example of the customization, the system is trained to translate natural language requests into “BQL,” BlackRock’s proprietary programming language.
This deep level of integration tracks with the findings in the IDC report. It found that 58% of “frontier firms” are already relying on custom-built or fine-tuned solutions rather than generic models. This is expected to accelerate, with 70% planning to move toward customized tools in the next two years to better handle their proprietary data and compliance needs. “That’s a trend that we’ve seen even in the low-code movement — taking an out-of-the-box solution, extending it, and customizing it,” said Taylor, the Microsoft commercial CMO.
OpenAI integration remains critical for many Microsoft customers. Another demo focused on Microsoft’s work with Ralph Lauren, showing how the “Ask Ralph” assistant interprets a shopper’s intent and recommends full outfits from available inventory. Like many of the scenarios inside Experience Center One, this experience runs on Microsoft’s Azure OpenAI Service. It’s a reminder that Microsoft’s partnership with OpenAI — renewed and expanded in recent months — is still a key driver of commercial demand for the tech giant, even as both companies increasingly work with other industry partners. Teams of agents are starting to redefine industrial work. The clearest example of this was a digital twin simulation from Mercedes-Benz — essentially a virtual version of a factory that lets engineers anticipate and diagnose issues without stopping real production.
The demo begins with a production alert triggered by a drop in efficiency. In a real plant, tracking down the cause (something as small as a slight angle change in a screw) might take a team of specialists days of sorting through machine logs and sensor data. In Microsoft’s version, a human manager simply asks the system to diagnose what’s causing the problem, through a natural language interface. That question triggers a set of AI agents, each with a specific role: one pulls the right data, another retrieves machine logs, and a third interprets what it all means in plain language. Within about 15 minutes, the system produces a clear explanation of the likely cause and possible fixes, shortcutting a task that could otherwise stretch across most of a week. AI is compressing weeks or months of scientific research into days or hours. A demo focusing on Insilico Medicine’s work with Microsoft showed how AI is starting to significantly collapse the timeline for drug discovery. The process begins with a “digital researcher” that scans huge amounts of public biomedical data to surface promising disease targets. It’s the kind of work that would otherwise take teams of scientists months of reading and analysis. A second system runs simulated chemistry experiments in the cloud, generating and ranking potential molecules that might bind to those targets. These simulations can be completed in a matter of days, or less, replacing weeks or even months of traditional laboratory work. The demo follows a real example: Insilico used this workflow to identify a potential target for a lung disease and design molecules that could affect it. The company then synthesized dozens of these AI-generated compounds in the lab. One of them is now in Phase 2a human trials.
That’s a small sampling of the demos inside Microsoft’s Experience Center One. During our tour, we walked past displays for other major brands, including more iconic U.S. companies, but not everyone was on board to have the media spotlight cast upon their projects. As a condition of access, we agreed to stick to the examples cleared for public release. Of course, the demos are carefully curated, and it remains to be seen how broadly companies will deploy these kinds of systems in their real-world operations. In many ways, the facility is a successor to Microsoft’s longtime Executive Briefing Center and conference facility, which remains in use a short walk away on the Redmond campus. Experience Center One has been in operation for a couple months, hosting delegations of business clients and political dignitaries, including the prime minister of Luxembourg this week. It’s closed to the public, invite-only. Employees can request access to visit. Visitors arrive via a circular drive, with a plaque at the entrance dedicating the building to John Thompson, the former Microsoft board chair who led the search process that resulted in Satya Nadella’s appointment as CEO. There are private briefing suites on the upper floors, and a full cafe on the second. The building also includes a conference center with three auditoriums. But perhaps the most distinct feature is the interactive portal. As they leave the demos, visitors walk through an immersive digital corridor with scenes of nature on the virtual walls. Walking through the tunnel, motion sensors track their movement, causing digital leaves and particles on the wall-sized screens to swirl and flow in their wake. The audio consists of nature sounds (birds, wind, and rustling trees) that were recorded locally in the Redmond and Sammamish area. And in a fittingly Pacific Northwest touch, the visual display is connected to a weather API. If it has been raining outside (as it often has been recently) the digital environment inside the tunnel turns rainy, too. It’s meant to be a final moment of grounding — a programmed moment of Zen to help executives decompress and center themselves as they contemplate the frontier ahead.
