I was chatting to a friend last week, and we got onto the topic of where sysadmins come from. "When two sysadmins love each other very much…" - no, that doesn't bear thinking about. BRB, washing out my mind with bleach.

But seriously. There is no certification or degree that makes you a sysadmin. Most people come into the discipline by routes that are circuitous, sideways, if not entirely backwards. The one common factor is that most people scale up to it: they start running a handful of servers, move or grow to a 50-server shop, build out some tools and automation to help them get the job done, then upgrade to 500 servers, and so on.

The question my friend and I had was, what happens when there are no 10 and 50-server shops around? What happens when all the jobs that used to be done with on-premises servers are now done in SaaS or PaaS platforms? My own employer is already like that - we’re over a hundred people, and we are exactly the stereotypical startup that features in big infrastructure vendors' nightmares: a company that owns no physical compute infrastructure, beyond a clutch of stickered-up MacBooks, and runs everything in the cloud.

The 90s and Naughties, when I was cutting my teeth in IT, were a time when there was relative continuity between desktop and enterprise computing, but that is no longer the case. These days you’ve got to be pretty technical as a home user before anything you’re doing will be relevant at enterprise scale, because those in-between cases have mostly gone away. I got my start in IT working at the local Mac shop, but neighbourhood computer stores have gone the way of the dodo. There simply are not many chances to manage physical IT infrastructure any more.

Where Are Today’s On-Ramps?

There is one part of that early experience of mine which remains valid and replicable today. My first task was pure scut-work, transferring physical mail-in warranty cards into the in-house FileMaker Pro "database". After two weeks of this, I demanded (and received) permission to redo the UI, as it was a) making my eyes bleed, and b) frustrating me in my data entry. Once I’d fixed tab order and alignments, I got ambitious and started building out data-queries for auto-suggestions and cross-form validation and all sorts of other weird & wonderful functions to help me with the data entry. Pretty soon, I had just about automated myself out of that job; but in doing so, I had proven my value to the company, and received the traditional reward for a job well done - namely, another job.

That is today’s path into computing. People no longer have to edit autoexec.bat on their home computers just to play games, but on the other hand, they will start to mess around behind the scenes of their gaming forum or chat app, or later on, in Salesforce or ServiceNow or whatever. This is how they will develop an understanding of algorithms, and some of them will go on from there, gradually growing their skills and experience.

A Cloudy Future?

To be clear, this cloud-first world is not yet a reality - even at Moogsoft, only a fairly small percentage of our customer base opts for the SaaS deployment option. More use it for the pilot, though, and interest is picking up, even in unexpected quarters. On the other hand, these are big companies, often with tens or hundreds of thousands of servers. They have sunk costs that mean they lag behind the bleeding edge of the change.

Even if someone does have 50 servers in an in-house server room today, as the hardware reaches its end-of-life date, more and more organisations are opting not to replace them. I was talking to someone who re-does offices, and a big part of the job is ripping out the in-house "data closet" to make more work space. The migration to the cloud is not complete, and won't be for some time, but it has definitely begun, even for existing companies.

What will save human jobs in this brave new world will be "intersection theory" - people finding their niches where different sub-fields and specialisations meet. Intuitive leaps and non-obvious connections between widely separated fields are what humans are good at. Those intersections will be one of the last bastions of the human jobs, augmented by automation of the more narrowly-focused and predictable parts of the job.

There will be other hold-outs too, notably tasks that are too niche for it to be worth the compute time to train up a neural network. My own story is somewhere in between the two, and would probably remain a viable on-ramp to IT - asssuming, of course, that there are still local firms big enough to need that kind of service.

Constant Change Is The Only Constant

To be clear, this is not me opining from atop an ivory tower. Making those unexpected, non-obvious connections, and doing so in a way that makes sense to humans, is the most precise definition I’d be willing to sign up to of the job I expect to have twenty years from now.

As we all continue to reinvent ourselves and our worlds, let's not forget to bring the next generations in. Thinking that being irreplaceable is an unalloyed win is a fallacy; if you can't be replaced, you also can't be promoted. We had to make it up as we went along, but now it's time to systematise what we learned along the way and get other people in to help us cover more ground.

See you out there.