Example · Systems Administrator
70% of your week as a systems administrator is still yours.
AI now writes scripts and routine fixes. Owning the environment when something breaks stays with you.
Make my planA typical week at the middle pace. Your plan uses your own tasks.
Is sysadmin a dying career?
In a typical week at the middle pace, 45% of this work needs your judgment and 25% is work you do using AI. AI could do about 30%. A model does well whatever part of your job it can do, and if you have watched Copilot write your PowerShell you know it. That is about 31% of your week.
Read more
It cannot do the part your company actually pays you for. That includes being the named person when a business-critical system stops at 6 a.m., knowing which Group Policy breaks which line, and being able to bring the ERP, the core business system, back from a backup.
For many sysadmins the bigger force this year is the move to the cloud, often pushed by a virtualization licensing bill. Moving most of your workloads into Azure or another cloud removes or reshapes a large part of what you do today, whether or not anyone ever gives an agent your credentials. That is both the main risk and the main opening, because a migration resets who knows the environment. Whoever builds the new one knows it. This report helps you become that person, instead of the reseller or consultancy doing the design.
The week, task by task
Who does each hour of a typical 40-hour week.
See every task
Needs your judgment.
- Incident response: outages, the on-call page, deciding to fail over, hands on the rack12%
- Migration project work: dependency mapping, wave planning, cutover testing10%
- Environment-specific troubleshooting: GPO, DFS, hybrid identity sync, print, "why does that site do that"8%
- Vendor, licensing and maintenance-window coordination with operations7%
- + 2 smaller tasks8%
AI does the slow parts. You still own it.
- Monthly patch cycle: WSUS and Intune rings, Linux hosts, the reboot order, what broke on Wednesday9%
- Backup job monitoring, alert-threshold tuning, routine file-level restores for users8%
- Building VMs and servers to standard: image, hardening baseline, domain join, agent install7%
AI could do most of it.
- Routine joiner-mover-leaver provisioning: AD account, groups, mailbox, licenses, offboarding10%
- PowerShell and scripting: reports, cleanup, bulk changes, scheduled jobs8%
- First-pass triage of escalations and alert noise: read the log, propose the cause7%
- Ticket notes, change records, KB articles, runbook text6%
Why some tasks still need a person
The 6 a.m. outage page, the file server down, the "everything is slow" call from a site
A model can propose causes, but it cannot decide to fail over the ERP during a production shift and own the consequence, and it does not know which switch stack reboots on its own every third Sunday. Someone with an ID badge and a rollback plan is answerable. When you are on call, that is you.
Knowing where the bodies are buried in Group Policy (GPO), shared file folders (DFS) and print
Years of Group Policy is undocumented history. A model reading your GPOs sees what they do, but it cannot tell which ones were a workaround for a scanner driver years ago. This knowledge has a shelf life. A cloud migration and Intune will retire a lot of it.
Being the person who can restore the ERP
A restore is the hardest task on your list to check. You find out it was wrong when the company cannot ship. The person named in the DR plan is doing protected work. If you are the only one, it is also single-point-of-failure work.
Access provisioning under a compliance regime (for example CMMC, the Defense Department's security rules)
Under the DFARS CMMC rule (effective November 10, 2025) a named affirming official affirms continuous compliance annually. The access-control and identification-and-authentication families under NIST SP 800-171 are what an assessor tests. The affirmation is signed above you, but the evidence is yours, and an assessor asks a person for it.
Maintenance-window and vendor coordination with operations
Which line or team can lose a system for 40 minutes, and whose weekend that ruins, is not written anywhere.
Good to know
What is hard about this field, and what you can do about it.
The hard truth
The platform you know best can be repriced overnight.
Your deepest skills may sit on one vendor's platform. After Broadcom bought VMware, costs rose and most surveyed IT leaders said they are shrinking their footprint.
Sources (2)
- 86% of surveyed North American IT decision-makers are actively reducing their VMware footprint. CloudBolt survey of 302 IT leaders, via CIO Dive, 2026-02
- Among those planning migrations, almost three-quarters are moving to public cloud. CloudBolt survey, via CIO Dive, 2026-02
What's next for a systems administrator?
Plan A is the best move inside the work you do now.
Grow in the job you have
The main focus. Most weekly sessions build toward it.
- Each week
- 4 sessions
- Ready in
- About 4 to 6 months
Why it fits
Why you. You may be the only internal person with both the on-prem context the migration needs and its failure history. The vendor and the security analyst have neither, and senior colleagues leave.
The gap. You have no cloud credential, no infrastructure-as-code in production and no written designation. Your manager may see you as the person who patches, because that is what you have asked to do.
The trade-off. You are volunteering for the work that replaces the work you are good at, and you will spend a year as the least experienced cloud person in the room. If the vendor's contract already names them operator of record, this is a fight over a contract you did not sign, and the incident-owner hedge matters more.
2 other moves that fit
- Own identity and access governance, end to endReady in about 3 months to a working pilot
- Incident owner, and the portable version of itReady in about 2 months to five runbooks
What's a plan B for sysadmins?
Plan B is meant to be different from this job. It was picked because skills from this role carry over.
Get ready for a new kind of job
A second option. A newer job AI is creating that these skills already fit.
Your on-call duty, failover calls and incident write-ups map onto running AI incidents, though policy work would be new.
- Each week
- 1 session
- Outlook
- Emerging
- Pay
- $116K to $375K a yearTwo 2026 postings: a Google analyst role and an Anthropic manager role (May); AI lab pay runs high.
2 other options
- AI workflow builderYou already script away repeat admin work in PowerShell, so building AI-driven automations is a short step with a small portfolio.
- AI governance analystRunning identity and access reviews gives you the control mindset that AI governance needs, plus some compliance training and the AIGP.
In your own plan you can switch to any of the 47 roles.
The 13 weeks
Small steps that add up. One plan for both: most sessions build Plan A, one a week builds Plan B.
By week 13 you will have
AFor Plan A
By week 13 you'll have an ERP restore runbook two people have tested, a dated Server 2016 plan, five on-call runbooks, and a memo asking to own identity, backup and DR.
BAnd for Plan B
- A timed record of a practice drill
- An AI incident runbook and after-action report for your portfolio
- 1Measure your starting point
- 2Ask to own the cloud
- 3Map account changes
- 4Test the ERP restore
You finish withA first ERP restore runbook
- 5Own compliance by name
- 6Plan the Server 2016 exit
- 7Train a second restorer
- 8Try a new access review
You finish withTwo numbers side by side, old way and new way
- 9Check the job market
- 10Lead one cutover
- 11Write down on-call
- 12Send the memo and ask
- 13Decide your next move
You finish withAn AZ-104 exam date, or a written reason to wait
Three courses from the plan
- Microsoft Certified: Azure Administrator Associate (AZ-104)Microsoft · $165 USD in the United States (priced by country)
Yes, if your migration is on Azure. Sit it after Week 13.
- Windows Server Hybrid Administrator Associate via AZ-800 + AZ-801Microsoft · Regional
Don't buy this. It is being replaced.
- Exam AZ-802: Administering Windows ServerMicrosoft · $165 in the US (varies by country)
Maybe in 2027, if your remaining on-prem estate is large. Take AZ-104 first.
Checked 2026-09-22.
The fuller picture
Risks beyond AI: moderateMigration risk
Cloud moves this number more than AI does. If your employer is moving workloads to the cloud, a large share of your week ends or changes hands in the migration before any AI is involved, and it is the work the AI score treats as safe because it needs your site knowledge or you on site: patching, builds, backup checks, vendor coordination, the rack. Afterwards, whoever designed the landing zone knows the environment. If that is a vendor, an MSP retainer can take the escalations. Headcount also tracks your industry's budget cycle.
Where this could be wrong
- We may be over-weighting the migration's reach. If core plant or business systems stay on-prem for latency or data-boundary reasons, your context keeps its value longer. Ask in Week 1 what the migration actually contains.
- Incident response may be less protected than we scored. The score assumes a person must decide to fail over. A mature cloud estate with automated failover and an MSP on retainer moves that decision to a runbook and a contract. The need to be there in person ends the day the last rack is empty.
- Compliance accountability may protect the security analyst more than you. The affirmation is the affirming official's; the program is the analyst's. If they want access-control ownership, the plan names them and the identity move becomes a support role.
- The shares are guesses until Week 1. If your log shows much more provisioning and escalation work, the share of your week AI can do rises into the mid-40s and the identity move stops being optional.
Also matched: sysadmin, systems administrator, system administrator. Research checked 2026-09-22. Page updated .