Am I setting unrealistic expectations by using AI to design critical infrastructure?

November 18, 2026 4 answers
Asked by 5k+vordali_628

I'm new to IT and the only admin for our infrastructure. I've built high availability and disaster recovery plans with a lot of AI help on the architecture, scripts, and documentation.

My manager is pleased, but I'm worried about handling an unexpected failure without senior support. How do I use AI responsibly and set realistic expectations?

Answers

_g31+claspmaraca.8

Using AI doesn't make you a fraud. Shipping something you can't explain or troubleshoot is the risky part. Go back through the design and make sure you understand what each component does, what it depends on, and what happens when it fails.

Treat generated scripts as untrusted code. Review them, test them somewhere safe, and keep a way to undo changes. Document the awkward dependencies and recovery steps too, not just the happy path. The goal is for you or another admin to be able to follow the plan under pressure.

Norlixa+5d:

Keep a copy of the recovery documentation somewhere you can access if the affected systems are down.

Reply to this answer
5p+quorble_917

The bigger issue is that the business has one relatively new person carrying the risk for critical systems. That's a staffing and management decision, not something you can fix by working harder.

Tell your manager what has been implemented, what has actually been tested, and where you still need help. Ask for a senior review or outside support for the architecture and recovery plan. Be clear that a successful build isn't a guarantee against every failure.

5k+vordali_628:

That helps. I can frame the review as reducing a single point of failure rather than just asking for reassurance.

Reply to this answer
6h!zindaro_394

Backups and high availability solve different problems. Make sure there are off-site backups, then practice restoring them and document the results. A recovery plan on paper is a lot less useful if nobody knows whether the restore works or how long it takes.

Test failure scenarios carefully and agree with management on what downtime and data loss the business can tolerate. If the current setup hasn't been tested against those expectations, say so plainly.

Reply to this answer
_puzzlecricket.8

Keep a record of what you've built and what you've learned. Praise is nice, but make sure your responsibilities and compensation get discussed too. If the company won't provide support or a path to learn from more experienced admins, looking for a team with mentorship is a reasonable option.

Reply to this answer

Add your answer

Answers are reviewed before publication.

Related Discussions