We have an older production machine running CentOS 7, which is long past end of life. It currently has no internet access, although temporary access may be possible. Our in-house IT team may be able to obtain sudo or root privileges, but they have limited Linux experience.
Outside providers have quoted roughly $50,000 in labor for the migration, so we are considering whether our own team could handle it with guidance from ChatGPT, Codex, or another AI tool.
The proposed process would be to audit the CentOS system, back up configurations and applications, run compatibility checks, resolve dependency problems, migrate to AlmaLinux 8, reboot and test, then move to AlmaLinux 9 and restore and test the applications.
If an AI agent had terminal or SSH access, internet access, and sudo privileges, could it safely perform most of the work while our IT staff reviewed commands and approved important checkpoints? How realistic is it for an experienced internal IT team to do this themselves with AI assistance?
4 Answers
The biggest concern is not whether AI can type the commands; it is whether the migration plan is safe. Build or borrow a test environment first, document the services and dependencies, and make multiple verified backups. Test a fresh AlmaLinux installation and restore the applications there. An in-place upgrade across major versions is much riskier than a clean installation and controlled migration.
AI can be useful for explaining commands, researching errors, generating checklists, and reviewing scripts, but it should not be given unrestricted control over a production OS migration. It does not truly understand your environment and can confidently suggest destructive or incompatible changes. A human administrator should verify every command, especially anything involving packages, boot loaders, storage, networking, authentication, or application data.
Check AlmaLinux's supported migration tooling, such as ELevate, and confirm that the exact CentOS version, installed packages, kernel, storage layout, and applications are supported. Even with a supported path, take a full image or clone and have a tested rollback plan. Do not assume that a successful OS boot means the application, integrations, scheduled jobs, permissions, and security settings are working.
A $50,000 labor quote may be excessive for a simple, well-documented server, but the price could reflect discovery, downtime planning, backups, compliance requirements, application testing, and responsibility for recovery. Get several detailed quotes and ask what is included. An experienced Linux administrator can probably guide the work for far less than that, while your internal team handles supervision and business-specific testing.
The internal team should first inventory what the machine actually does and whether it has undocumented dependencies. If nobody can explain the services or restore the backups, that is a strong reason to bring in a qualified administrator before making changes.