Building a trusted IT asset inventory with Fabric and AI at Microsoft

For an enterprise company like Microsoft, understanding exactly which devices are connected to our corporate network is essential.

As the company’s IT organization, we in Microsoft Digital manage millions of network-connected assets. An asset could mean an employee device, network infrastructure, pieces of equipment that sit inside the data center or buildings, or anything else that gets connected to our network.

Information about these assets previously lived within disparate systems, processes, and teams scattered across the company. This fragmentation created a barrier to the establishment and maintenance of a trusted IT asset inventory.

For example, if a security team needed information such as device ownership, location, and the lifecycle status of a device involved in a potential security incident (which remains rare at Microsoft), the dispersed nature of the previous system meant a lot of manual outreach across the organization. This was time-consuming and frustrating for our IT team, and it also delayed security incident response times.

This limited visibility also made it more difficult to manage the company’s hardware investments in an efficient way, something that we knew needed fixing. And because we realized this is a challenge our customers are also having, we—the company’s Customer Zero—agreed to share what we learned with you along the way.

To get started, we launched a multi-year effort to inventory our enterprise IT assets. This inventory brought together previously disconnected sources, improved data quality, and created the foundation for future AI-powered asset management experiences across the enterprise.

Why asset inventory matters

At Microsoft, we prioritize security over all other business outcomes. Our security teams depend on accurate, up-to-date information about network devices to protect the company’s virtual environment, detect any issues promptly, and respond to security incidents efficiently. Any time spent tracking down essential details like device ownership and network identity can lead to delays in incident response.

“One of the early challenges we faced was that we weren’t operating as a single organization—we were a collection of businesses. Each business leader had developed their own governance frameworks, operating practices, and approaches to managing assets, making consistency and alignment difficult to achieve.”

Debashis Sahoo, principal group manager, Microsoft Digital

We on the IT team don’t control the acquisition process of assets across every business unit of the company, which meant asset inventory was constantly changing and there was no direct oversight for when new devices entered the network.

“One of the early challenges we faced was that we weren’t operating as a single organization—we were a collection of businesses,” says Debashis Sahoo, a principal group manager in Microsoft Digital. “Each business leader had developed their own governance frameworks, operating practices, and approaches to managing assets, making consistency and alignment difficult to achieve.”

Another important element of IT asset inventory is the financial aspect.

We inventory all high-value devices in our network to determine our financial footprint in the IT asset arena, which includes both devices that are being actively used—including laptops, printers, and deployed IoT devices—and devices that aren’t currently being deployed on our network.

“For employee devices in Microsoft, 60% of our employees own more than one device,” says Aniruddha Das, a principal product manager in Microsoft Digital. “For developers, typically they have a primary device. They might also have a backup workstation, but a lot of times what we have seen is these devices sit in inventory and they don’t get deployed, which means we have our capital tied in but not being utilized. We want to know what’s tied in so that we have good information about what to purchase next year.”

These undeployed devices were a prime focus of our inventory efforts. We also prioritized finding which devices had reached their end of service life limits and needed to be replaced. Older devices don’t get patched as often, which can expand the attack surface of IT infrastructure.

The scale of these different challenges meant we couldn’t work on everything at once. Our team worked with security stakeholders to identify the devices and attributes that were most critical to the company’s security on a daily basis. We determined that network devices, lab devices, and Internet of Things (IoT) devices should take priority.

Driving alignment through iterative leadership check-ins

Our answer to this challenge was to build a process that connected to any system that might have information about a security-related asset inside the company.

To create this new system, which we call the Enterprise Asset Data Platform (EADP), we brought together stakeholders from across Microsoft in a Kaizen continuous improvement initiative.

When we waded in, we found the task of standing up a new IT asset inventory proved more wide-ranging in scope than we initially anticipated. Asset data was peppered across myriad systems, each of which served a different business need. The networking, security, real estate, lab, and enterprise asset management teams all had stakes in the outcome of this initiative. We ran a two-day program to identify focus areas and divide them into different segments, such as the IoT and lab segments.

The Kaizen process helped us establish common goals across departments, define what success would look like, and create a roadmap. We used this exercise to create a charter defining what issues we wanted to address and get universal alignment around the shared goal of improving inventory quality and completeness.

We aimed to cut down the time it takes to address and remediate security incidents and stand up weekly, biweekly, and monthly operating reviews around governance for the 12 months following the Kaizen to monitor the progress of the initiative. This was to remove any blocks that might arise and make sure teams were aligned throughout the process.

Modernizing our inventory solution

Expanding our IT asset inventory capabilities meant we needed to modernize the tech that underpinned them. Our device data was initially collected from over 70 sources aggregated into a central inventory, which meant we encountered a wealth of issues with duplication, data quality, and clashing governance.

In ingesting all this data, we took over data management from the teams who distribute devices here at Microsoft, which cleaned up our data because it more accurately reflected what was really being used by our employees and contingent staff. This allowed us to reduce our per-person spend by  22%, a significant savings.

And we’re making more improvements. If one of our business units is planning to buy a new set of laptops or lab devices, we can now “see that,” and can mine our data at that point, instead of hoping to learn about the acquisition from a secondhand source.

The complexity of standing up and maintaining this process demanded that we create our data platform, which we built on Microsoft Fabric. With this system, data is ingested, enriched, reconciled, and surfaced through curated datasets designed for operational use, removing the need to manually review each device and compare it against what’s already on record.

Accelerating our progress using AI

Building a trusted inventory delivered immediate security and operational benefits to our organization. But we quickly realized that the same foundation could support a broader goal: Using AI to help our employees and contingent staff find, understand, and act on asset information. With consistent data now available across the enterprise, we could begin building intelligent experiences on top of the inventory, rather than asking users to navigate dozens of disconnected systems.

As a platform provider, Microsoft Fabric holds all the device data across the company; the majority of our AI workload is on this platform. If a Microsoft team needs to create an AI experience for their business unit and wants to use those capabilities, they can do it by connecting to our Microsoft Digital suite of AI tools.

“AI has improved our engineering efficiency drastically and reduced our time to value significantly. We’re able to deliver value and identify gaps much sooner than before.”

Ashwin Kaul, senior product manager, Microsoft Digital

We’re constantly looking for new ways to create valuable experiences for different domains and business units at Microsoft. We’re currently in the process of creating agent-based experiences that streamline tasks like lab operations, device tracking, and asset updates. Our operators will eventually be able to interact with our AI-powered assistants using natural language, streamlining the process of finding and using the information they need.

Because inventory records are now standardized and trusted, teams no longer needed to spend time reconciling data across systems. That consistency also created an opportunity: using AI to help employees interact with asset information more naturally.

When it comes to selecting which devices a user needs, AI can understand the user’s persona, their role, and make recommendations for which devices would best suit that user’s experience. From there, the user can choose from the AI’s device recommendations. This device selection process, which used to take anywhere from 15–20 days, can now be completed in minutes.

“AI has improved our engineering efficiency drastically and reduced our time to value significantly,” says Ashwin Kaul, a senior product manager in Microsoft Digital. “We’re able to deliver value and identify gaps much sooner than before.”

AI also helps us track and trace and then remediate security issues within our extensive network of IoT devices, which contain thousands of lights sensors, temperature controls, and other device types across the campuses of our offices globally. We partnered with our real estate team to use Microsoft Copilot studio to stand up agentic AI capabilities for data quality checks to make sure that all these devices are being accurately documented.

At our Kaizen event, we chose a goal of a 90% improvement for inventory completeness and accuracy. We’ve made a 74% improvement compared to our baseline as of summer 2026, and we’re on track to end the year with an 85% improvement.

Our AI-powered foundation for the future

For our team in Microsoft Digital, the journey began with a simple objective: Gain a trusted view of the company’s technology assets. The result is a stronger security posture, improved operational efficiency, and a data foundation that can support increasingly intelligent experiences.

One great example of these kinds of new experiences: We’re currently in the process of developing a Labs Asset Management Agent, or LAMA. LAMA will be a human-led, multi-agent Frontier Firm experience for our employees that will hugely simplify Microsoft Labs operations.

Our labs have a huge amount of operational processes to manage and run. Once we simplify these processes and reduce the amount of manual intervention needed to manage lab devices, we anticipate reducing vendor and hardware costs using this agentic AI. Our goal is to speed up the pace at which we can do lab deployments by 50%.

As organizations look for new ways to apply AI, our experience demonstrates an important principle: the quality of AI outcomes depends on the soundness of the underlying data. Building a trusted asset inventory may not be the most visible part of an AI strategy, but it is often one of the most important.