By Allen Houchins, CIO, Fleet Device Management
Enterprise IT has always built systems around people, such as interfaces people can navigate, identities tied to employees and devices assigned to individual users. AI agents are now challenging those assumptions. As they take on more work and dedicated AI hardware reaches the desk, IT teams will need to reconsider both how they manage technology and what they are managing.
In 2027, four connected shifts will put pressure on tools and workflows designed for a human-operated enterprise.
Point-and-click management will give way to agent-ready workflows
Point-and-click device management is coming under pressure as more organizational work becomes automated. IT has a significant opportunity to benefit, but realizing it will require tools that support agents as a core capability.
An interface that makes a task straightforward for a person can still create friction for an agent. When configuration depends on navigating menus and manually repeating changes, the interface becomes a constraint on how much work can be automated.
GitOps and configuration as code offer a more natural foundation. They express desired configurations in a form agents can work with, while giving practitioners a way to review changes and understand what happened. That creates room for people to focus on higher-level decisions instead of carrying out each individual action.
The competitive pressure will grow as practitioners use agents to accomplish substantially more. Tools built around manual interaction will fall behind workflows that allow agents to contribute directly.
Open source will become easier to interrogate
Open source has always offered a clear advantage: anyone can read the source code. In practice, however, having permission to inspect software and having the time or expertise to do so are very different things.
For many organizations, that transparency has remained a possibility rather than a routine activity. The code was available, but most users were unlikely to examine it themselves.
AI agents could change that relationship in 2027. They can help people investigate how software works, explore potential issues and answer questions without requiring every user to read code. Open-source software could face much more frequent scrutiny as the effort required to examine it falls.
That does not make an agent’s assessment definitive or remove the need for human judgment. It does make transparency more usable. With proprietary software, users still face a black box; with open source, they have underlying material they can question and investigate.
Agents will need identities and systems of their own
Most enterprise technology assumes a human operator who needs to open a browser, read a page and navigate an application. As agents take on more tasks, those assumptions will limit efficiency.
For instance, in a point-and-click workflow, a browser makes information accessible to a person, but an agent may be better served by direct access to the underlying data. Requiring it to follow the human route adds another step to the task.
Over time, enterprises will redesign workflows around how agents operate. That will also raise questions about identity, like how an agent is recognized, what information it can access and which actions it is permitted to perform.
In 2027, expect growing attention on hardware, identity frameworks and systems built specifically for agents. Early approaches may mirror familiar human workflows, but the direction will be toward infrastructure that supports agents’ own operating requirements.
AI hardware will become an endpoint management responsibility
AI infrastructure will also extend further beyond cloud platforms and centralized data centers. As dedicated AI machines reach desks and team workspaces, IT will inherit responsibility for managing them.
These systems will challenge assumptions about operating systems, ownership and use. AI workstations can introduce Linux and Arm-based systems into environments accustomed to other platforms. Some will run headless; others will be shared across teams. Workflows built around one employee sitting at a keyboard will not always fit.
Visibility will be the immediate challenge. Knowing the operating system is only part of the picture. IT teams will need to understand vulnerabilities across the AI stack, which models are running, where API keys are stored and how data moves between models, tools and cloud APIs.
Gaps in Linux coverage, purchases made outside IT’s oversight and tools that cannot inspect the AI software layer will leave organizations exposed.
Takeaways
Across these shifts, preparation starts with making AI part of everyday IT governance. Organizations that establish visibility, access controls and management expectations early will be better equipped to support both agents and AI workstations. The opportunity in 2027 is considerable, but it depends on giving IT the tools to manage an enterprise whose users, workflows and devices are evolving daily.
##
ABOUT THE AUTHOR
Allen Houchins is CIO of Fleet Device Management, bringing extensive experience in enterprise IT, security and digital employee experience. Previously, he served as Vice President of IT and Workplace Services at Jamf, helping scale the company from 200 to 3,000 employees and leading IT and security through its transition to a public company. Earlier in his career, Allen held engineering and systems leadership roles at Apple, including overseeing demo systems for more than 400 retail stores worldwide.
