It’s impossible to predict with absolute certainty which jobs will grow over the next decade or so, but we can make some pretty good guesses based on current trends. For example, it’s a near-certainty that positions will only grow in the fields of artificial intelligence (A.I.) and IoT (Internet of Things). But which jobs might decline in the long term? That’s also difficult to determine, although we tried our hand at some guesses. Bottom line: A.I. is going to end up challenging a lot of categories, in terms of human employment. And whether or not the following predictions actually pan out, one thing is clear: tech pros need to stay current on the latest and greatest technology if they want to stay employed.

Programmers

The rise of machine learning and artificial intelligence may put a subset of programming jobs at risk. Imagine a future in which apps automatically adapt to user input, building out new logic without the need for a human programmer. That’s not a science-fiction scenario: a sufficiently trained A.I. can already beat humans at chess and Go without too much trouble, which is significant when you consider the amount of creativity and improvisation involved—traits that could easily port over to programming. That’s not to say we’re facing a programmer-less future. But if A.I. indeed assumes a greater role in software development, those programmers who survive will need to do more than simply translate requirements into code; they’ll have to display enough imagination and innovation to differentiate themselves from the machine options out there.

Architects

Earlier this year, David Foote, chief analyst of Foote Partners LLC, explained to Dice that the value for 33 certified and noncertified architecture skills declined in 2016. “It’s not that companies don’t value architects or need their expertise,” he said at the time. “But they’ve staffed up, which limits demand and the need to offer premium pay, at least temporarily. That’s why I’m putting architects on the list of roles that may experience flat growth and possibly a small decline in market value in 2017.” The key word in Foote’s evaluation is “temporarily.” Systems architects pulled down an average of $125,946 in 2016, according to Dice’s most recent Salary Survey—good for second place on the list of top-paying jobs, despite a year-over-year decline of -4.7 percent. With that sort of pay, and a clear need on the part of companies for architects of all experience levels, it seems a fairly safe bet that the job market for this category will rebound. In the long term, architects (and other tech pros) must stay aware not only of new technologies, but also methodologies. Every aspect of the architect’s job—from user interfaces to system requirements—will likely evolve in coming years. Those that don’t stay aware will risk obsolescence.

Datacenter Jobs

But wait, you protest, there are more datacenters than ever. And that’s certainly true: thanks to the ubiquity of cloud services, datacenters have sprung up all over the place. But therein lies the rub: as datacenters evolve, the technology inside of them becomes more sophisticated—which means the automation of jobs once held by human beings. This doesn’t mean that datacenter jobs such as security specialists, backup and storage administrators, and storage engineers will disappear completely; rather, there will be fewer of these roles, as hardware and code assume an ever-increasing number of datacenter responsibilities. What does this mean for tech professionals who already work in datacenters and want to stay there? Specialize, specialize, specialize. Automation initially takes out the positions that involve routine processes; people whose roles involve critical thinking, creativity, and management (i.e., troubleshooting and systems design) have a better chance of surviving.

Legacy Languages

Legacy languages are a sensitive topic among tech pros. For obvious reasons, many of them feel a great deal of affection (or at least affectionate hatred) for languages they’ve used for years—if not decades. Suggest to them that a language is bound for the dustbin of tech history, and they’ll point out all sorts of places where that language is very much in use. But none of that stops many languages—even really popular ones—from seeing reductions in use. Take Objective-C, for example, which dominated iOS and macOS development for years. When Apple decided that it would replace Objective-C with Swift, a language with a variety of improvements over its predecessor, it put Objective-C on the curve to eventual obsolescence. Objective-C still places highly on programming-language rankings such as TIOBE, but its use will end up restricted more and more to the maintenance of legacy apps. A similar paradigm applies to COBOL and other decades-old languages that continue to power legacy systems (especially mainframes). Tech pros who specialize in these languages can command eye-popping salaries from companies desperate to keep their ancient systems clanking along for another record-breaking year. But sooner or later, those systems will migrate to the cloud (or newer infrastructure) after a long and painful process, and then those languages will become strictly hobbyist. (This is why it pays to stay aware of a programming language’s popularity.) In the meantime, there’s still a specialized market for legacy software experts. If that interests you, consider boning up on COBOL, Fortran, AS/400, UNIX (and C), and older versions of Windows. You should also know a variety of debugging tools.