AI agents walk on the file system using standard tools to navigate directories and read file paths.
The challenge, however, is that there is a lot of enterprise data in object storage systems, especially Amazon S3. Object stores provide data through API calls, not through file paths. Bridging that gap requires a separate file system layer with S3, deduplicated data, and sync pipelines to keep the two aligned.
TeaThe rise of agentic AI has made that challenge even more difficult, and it was impacting Amazon’s own ability to get the job done. AWS engineering teams using tools like Kiro and Cloud Code faced the same problem: Agents defaulted to the local file tool, but the data was in S3. Downloading it locally worked until the agent’s context window was collapsed and session state was lost.
Amazon’s answer is S3 Files, which mounts any S3 bucket directly into the agent’s local environment with a single command. Data remains in S3, no migration required. Under the hood, AWS integrates its Elastic File System (EFS) technology with S3 to provide full file system semantics, not a workaround. S3 files are now available in most AWS regions.
"By making the data immediately available in S3, as if it were part of the local file system, we found that we had a really big acceleration with the ability for things like Kiro and Cloud Code to be able to work with that data," Andy Warfield, VP and distinguished engineer at AWS, told VentureBeat.
Difference between file and object storage and why it is important
S3 was built for durability, scale, and API-based access at the object level. Those qualities made it the default storage layer for enterprise data. But they also created a fundamental incompatibility with the file-based tools that developers and agents rely on.
"S3 is not a file system, and does not have file semantics on many fronts," Warfield said. "You can’t do any moves, atomic moves of an object, and there aren’t actually directories in S3."
Previous attempts to bridge that gap relied on FUSE (Filesystem in Userspace), a software layer that lets developers mount a custom filesystem in userspace without altering the underlying storage. Tools like AWS’s own mount point, Google’s gcsfuse, and Microsoft’s blobfuse2 use FUSE-based drivers to make their respective object stores more like file systems.
Warfield notes that the problem is that those object stores were not yet file systems. Those drivers either simulated file behavior by stuffing extra metadata into the bucket, thereby breaking the object API view, or they refused file operations that the object store could not support.
S3 files take a different architecture altogether. AWS is connecting its EFS (Elastic File System) technology directly to S3, offering a full native file system layer while keeping S3 as the system of record. Both the file system API and the S3 object API remain accessible simultaneously against the same data.
How S3 files accelerate agentic AI
Before S3 files, agents working with object data had to be explicitly instructed to download the files before using the tool. This caused a session state problem. As agents narrowed their context windows, records of what was downloaded locally were often lost.
"I had to remind the agent that the data was available locally," Warfield said.
Warfield went before and after for a general agent task involving log analysis. He explained that if a developer was using Kiro or Cloud Code to work with log data, in the case of just objects they would have to tell the agent where the log files were located and go and download them. Whereas if the logs can be immediately mounted on the local file system, the developer can easily identify that the logs are on a specific path, and the agent has access to go through them immediately.
For multi-agent pipelines, multiple agents can access the same mounted bucket simultaneously. AWS says thousands of compute resources can connect to a single S3 file system at the same time, with total read throughput reaching several terabytes per second – something VentureBeat was not able to independently verify.
Shared state between agents works through standard file system conventions: subdirectories, notes files, and shared project directories that any agent in the pipeline can read and write. Warfield described AWS engineering teams using this pattern internally, in which agents log investigation notes and task summaries to shared project directories.
For teams building RAG pipelines on top of shared agent content, S3 Vector – launching on AWS re:Invent in December 2024 – layers on top for similarity search and retrieval-enhanced generation against the same data.
What analysts say: It’s not just a better FUSE
Positioning S3 files against FUSE-based file access from AWS, Azure Blob NFS, and Google Cloud Storage FUSE. For AI workloads, the meaningful difference is not primarily performance.
"S3 files eliminate data shuffling between object and file storage, turning S3 into a shared, low-latency workspace without copying data," Gartner analyst Jeff Vogel told VentureBeat. "The file system becomes a view, not another dataset."
With the FUSE-based approach, each agent maintains its own local view of the data. When multiple agents work together, those views can potentially become out of sync.
"This eliminates a whole range of failure modes, including unexplained training/inference failures caused by outdated metadata that are extremely difficult to debug," Vogel said. "FUSE-based solutions remove complexity and problems for the user."
The agent-level implications go further still. The architectural logic matters less than what it opens up in practice.
"For agentic AI, which thinks in terms of files, paths, and local scripts, this is the missing link," IDC analyst Dave McCarthy told VentureBeat. "This allows an AI agent to treat an exabyte-scale bucket as its own local hard drive, enabling a level of autonomous operational speed that was previously bottled up by the API overhead associated with approaches like FUSE."
Beyond agent workflows, McCarthy sees S3 files as a broader inflection point in how enterprises use their data.
"The launch of S3 Files isn’t just S3 with a new interface; This aims to remove the last friction point between huge data lakes and autonomous AI," He said. "By combining file and object access with S3, they are opening the door to more use cases with less rework."
What does this mean for enterprises
For enterprise teams that are maintaining a separate file system with S3 to support file-based applications or agent workloads, that architecture is now unnecessary.
For enterprise teams consolidating AI infrastructure on S3, the practical change is concrete: S3 stops being the destination for agent output and becomes the environment where the agent’s work happens.
"All these API changes that you’re seeing from storage teams come from direct work and customer experience using agents to work with data," Warfield said. "We’re really focused on removing any friction and making those interactions work as well as possible."
<a href