Partly, and this is one of the roles where the numbers are genuinely uncomfortable. Federal figures put network and computer systems administrators at a 4 percent decline from 2025 to 2035, taking the occupation from 323,600 positions down by about 13,200. Median pay was $99,130 in 2025.
A declining projection in a growing technology sector needs explaining, because the cause is not what most coverage assumes. Cloud migration and infrastructure automation did most of this work before generative AI arrived, and understanding that distinction changes what a system administrator should do about it.
Key points
- The decline is real but modest. Roughly 13,200 positions over a decade from a base of 323,600, which shows up as reduced hiring rather than layoffs.
- Cloud and automation caused most of it. Managed services removed the work of running physical servers, patching them and keeping them alive.
- Pay remains strong at $99,130, because the remaining work is more demanding than what was removed.
- The role is converting, not vanishing. Platform engineering, site reliability, security and cloud architecture all recruit heavily from this pool.
- Adjacent security work is booming, which is the clearest available exit route.
What the job used to be and what it is now
The traditional description involves racking servers, installing operating systems, applying patches, managing backups, adding user accounts, resetting passwords and responding when something stops working. A great deal of that has genuinely gone, and it went in stages.
Virtualisation removed the physical work of provisioning machines. Cloud platforms removed the data centre entirely for many organisations. With it went the hardware lifecycle, the capacity planning and the physical redundancy. Configuration management tools replaced manual server setup with code that defines the desired state. Managed services took over the operational burden of running databases, mail systems and application platforms.
What remains is different work with the same job title. A modern administrator defines infrastructure as code. They manage identity and access across many services. They handle security hardening and compliance, design and test disaster recovery, control cloud spending, and integrate systems that were never meant to work together. The manual tasks went. The architectural and security responsibility grew.
That is why pay held up while headcount fell. The organisations that removed administrator positions removed the ones doing repetitive operational work. The ones that remain are doing something closer to engineering.
The part that did not change
It is worth being clear about what survived all of this, because it is the answer to what a system administrator should be good at now.
Someone still has to know what the organisation actually runs. Not what the documentation says. Not what the architecture diagram from three years ago shows. The real set of services, dependencies and undocumented integrations that has piled up over time. In most organisations this knowledge lives in two or three people’s heads, and that is why those people are hard to replace.
Someone still has to decide what happens when things go wrong. Not the runbook step, which can be automated. The judgement call: whether to fail over, whether to accept degraded service, whether the fix might make things worse. That decision gets made under pressure, with incomplete information, and it carries responsibility.
Someone still has to say no. A great deal of infrastructure work is refusing requests that would create security exposure or operational debt, and explaining why in terms the requester accepts. No tool does that.
Where AI specifically fits
Generative AI has added a further layer, and it is worth being precise about what it does.
It writes configuration and scripts quickly, which compresses a task that used to take hours into minutes. It explains unfamiliar error messages and log output, which shortens diagnosis considerably for anyone working outside their core expertise. It generates documentation, which administrators have historically been poor at maintaining. It answers routine questions that used to arrive as tickets, which reduces the support load reaching a human.
Anomaly detection and predictive alerting have improved substantially, flagging failing hardware and unusual traffic patterns before they cause outages. Automated remediation handles a growing set of known problems without human involvement, restarting services and reallocating capacity according to defined rules.
The consistent limit is that these systems act within boundaries someone designed. Deciding what should happen when a service degrades, what the recovery objective actually is, and which trade-offs are acceptable remains architectural work. So does deciding what to automate in the first place.
There is also a failure mode worth naming. Generated configuration that looks correct but is subtly wrong causes outages that are hard to diagnose. The reviewer’s usual signals of carelessness are simply absent. Teams that adopted these tools without changing how they review have generally shipped more incidents, not fewer.
Why the decline is smaller than the automation suggests
Three factors slow the fall.
Complexity grew as much as tooling improved. Organisations now run more services, across more providers, with more integration points than a decade ago. Managing that estate takes skilled people even when each individual component is easier to run.
Security expanded enormously. Identity management, access control, patching cadence, compliance evidence and incident response have all grown into substantial workloads that fall to administrators in most organisations.
Cloud costs became a discipline. Someone has to understand why the bill went up, which resources are idle, and how to reduce spending without breaking things. That role barely existed fifteen years ago.
Migration is slow and incomplete. A great many organisations still run substantial on-premises infrastructure, in regulated industries, in manufacturing, in the public sector. That work continues to need administrators for years yet.
There is a fifth factor that rarely gets mentioned. A meaningful number of organisations moved to cloud, discovered the cost was higher than expected, and moved some workloads back. Repatriation is a real trend rather than a rhetorical one, and it needs exactly the skills the previous decade was busy declaring obsolete. That does not reverse the overall direction, but it does mean the decline is lumpier and slower than a straight line.
What to know before deciding
| Measure | Network and computer systems administrators, 2025 |
|---|---|
| Median annual pay | $99,130 |
| Number of jobs | 323,600 |
| Projected change, 2025 to 2035 | -4 percent (Decline) |
| Projected employment change | -13,200 |
| Typical entry-level education | Bachelor’s degree |
The comparison across IT is the most useful thing here, because it shows exactly where the work moved.
Information security analysts are forecast to grow 21 percent with median pay of $129,180. Computer network architects grow 8 percent at $134,050. Meanwhile computer support specialists decline 3 percent and computer programmers fall 7 percent.
The pattern holds across the whole sector. Roles defined by executing known procedures are declining. Roles defined by designing systems, and by owning their security and reliability, are growing and pay more. System administration sits between the two. That is why it declines slowly rather than sharply.
For context on how exposure is measured, the Bureau publishes AI exposure categories for 831 occupations and states plainly that exposure “does not imply job loss, productivity gains, automation probability, or wage effects.”
What actually changes over the next decade
- Scripting stops being a bottleneck. Configuration and automation code gets written far faster, which shifts value toward knowing what to write.
- Diagnosis gets faster. Log interpretation and error explanation compress the time spent understanding an unfamiliar failure.
- Review practices have to change. Generated configuration needs verification against intent rather than a skim for obvious errors.
- Security consumes more of the role. Identity, access and compliance work grows in almost every organisation.
- Cost management becomes a named responsibility. Cloud spending is now a discipline with dedicated tooling and, increasingly, dedicated people.
- Titles shift. Platform engineer, site reliability engineer and cloud engineer are absorbing work that used to sit under system administration. The work did not disappear so much as it was renamed and repriced, which matters when you are searching for jobs.
Decision framework
Five questions if you administer systems now.
- What share of your week is repetitive execution? Patching, provisioning, account management and routine ticket handling are the exposed categories. Measure it rather than estimating.
- Do you understand the systems or only operate them? Being able to explain why the architecture is the way it is, and what breaks if it changes, is the difference between the roles that survive and the ones that do not.
- Can you move toward security? It is the fastest-growing adjacent field, it pays substantially more, and administrators already have most of the underlying knowledge.
- Is your employer on-premises, cloud or hybrid? This determines your local exposure far more than any national figure, and hybrid environments need administrators for a long time yet.
- Will you learn to review generated work properly? Someone has to catch the configuration that looks right and is not. That person is more valuable than the one who wrote it.
That last skill applies well beyond infrastructure. Knowing where a model’s confident output fails, and building a habit of verifying it against intent, transfers to any technical work. If you want a structured route in, explore Coursiv AI lessons and check current plan details on the official site.
Your next step
If you administer systems now, spend a week categorising your own work into repetitive execution, incident response, security and design. That split is the honest measure of your exposure, and most people find it differs substantially from what they assumed.
Then pick the adjacent field that fits what you already enjoy. Security suits people who like adversarial thinking, site reliability suits people who like systems under load, and cloud architecture suits people who like design. All three are growing, all three pay more, and all three recruit administrators specifically because the underlying knowledge is hard to teach from scratch.