Over the last five years, there has been increasing noise about whether the growing automation and orchestration of networks (and infrastructure in general) will lead to our jobs being eliminated. Concerns about mass layoffs are understandable given what happened when large scale automation was introduced to manufacturing.
What is left after automation has taken its toll on an industry? Presumably there is work for those who create and maintain the automation systems and there will be a need for workers to do the tasks which cannot be adequately automated, but the people who don’t fit into these categories might be facing a tough future. Some workers will retrain or adapt their skills to shift themselves into one of the “needed” categories, but since the idea of automation in most industries is to
reduce the need for salaried humans be more agile and respond faster to customer needs, the competition for those positions is likely to be strong.
Does Automation Mean A Bleak Future?
Up front let me say that I believe that the predictions of imminent doom are utter codswallop. In order for the jobocalypse to occur, automation has to be present across the every industry because IT infrastructure exists pretty much everywhere as a business enabler, but only in subset of industries (e.g. service providers) is it actually a product and a direct source of revenue. As with manufacturing, if a product can be made more cheaply and efficiently (thus implicitly maximizing profits), it is in a company’s interest to invest time and money in following that path; in other industries however, the motivation to do so may be significantly weaker.
I had a discussion (argument?) with Greg Ferro a year ago perhaps, where I suggested that white box switches would be attractive to enterprises in part because they are cheaper. Greg countered that this was not true because enterprise IT departments don’t care about saving money; his argument was that saving money on IT spending does not generate more revenue for the company. It may create a one-time short-term increase in profitability, but strategically an enterprise is better off simply shipping more units of whatever it is they make (or whatever service they offer, and so forth), as that’s where the revenue actually comes from. I’m paraphrasing so I hope I have not misrepresented Greg here (if so, I’m sure he will correct me and I will amend accordingly).
When you think about it that way, I have to ask why it is that every IT department seems to be asked to reduce their budget every year anyway? I think the truth of the situation a) varies depending on industry, and b) may lie somewhere in between these two positions. For example, if a company is staging itself to be purchased or for an IPO, every last penny apparently counts. At least, that’s the excuse we’re given when the company cuts the holiday party back to pizza and fruit juice in the office after work one day, and pushes to minimize unbillable expenses, reduce travel, save on IT costs, and so forth. For other companies it’s more important for them to buy equipment from the big name vendor because the sales droid has shinier shoes and, let’s be honest, nobody ever got fired for buying Cisco. If cost is not a driving factor for those companies, then how does automation benefit them?
Infrastructure at the speed of business—which I don’t believe is any company’s advertising tagline, but probably should be—is the potential benefit of all this automation. Putting aside the potential cost savings from the removal of meatbags from any process, the ability to deploy infrastructure at a higher speed may well be of benefit by removing the inevitable last mile dependency on infrastructure which seems to delay every project and deployment.
One Size Does Not Fit All
This is all very well, but will every company wish to eliminate all their engineers and replace them with automation? Maybe eventually, yes. But as with all other amazing technologies, the rate of adoption varies greatly from the early adopters through to those who will only adopt at the point at which they have no choice but to upgrade their existing infrastructure and an automated solution is the only option on the table. In other words, for a number of years yet to come, I believe there will be a similar range of jobs available to what we have today. Over time, however, the skills required from employees will split between the more hands-on work which is hard or impossible to automate, and the more abstracted configuration and design work which will shift towards support and development of automation, as is happening to some extent today. In other words, the “middle ground” where people currently design and execute changes manually, performing rote configurations like port turn ups and firewall rule deployments, will slowly disappear as they are replaced or are supplemented by automated processes and orchestration.
Consider the three roles I described as steps in a career development path, i.e. tier 1 is more passive involvement, tier 2 is more rote configuration and troubleshooting, and tier 3 is architecture and design. For networking that might be analogous to:
- Design Authority
The gradual elimination of tier 2 potentially leaves us in the position where engineers have to make the jump from tier 1 to tier 3, yet there’s no tier 2 in which to spend time developing the skills which lead to a role in tier 3. So something we need to keep an eye on as time progresses is the development of the tier 1 engineers to make sure we have tier 3 engineers in the future. We may also see tier 1 engineers leaving a Tier1/3 company in order to spend time as a tier 2 in a Tier 1/2/3 company, then moving back at a future date into a tier 3 role. Suddenly, far from tier 2 being a redundant position, companies still offering those roles may turn out to be very much sought after by those wishing to develop their career.
I Go to Extremes
So let’s take the final step. What happens if we successfully demonstrate automation of our infrastructure to the extent that most of the roles in all tiers have been are redundant. Even in data centers, at some point I expect to see 19″ racks with backplanes to which server blades can be pushed home to connect to power and data without needing separate cables, at which point server insertion / removal / replacement could be automated in the same way we have been using tape libraries for years:
Echoing the tier 2 situation above, one of the characteristics of a post-industrial society as described by Alvin Toffler (of “Future Shock” fame) is that people will need to have multiple different professions over their lifetime because each of those professions will become outdated in a relatively short timeframe. This is in stark contrast to the previous generations who have enjoyed the infamous “job for life” if they wanted it.
If we take this post-industrial society to the next level, we could postulate that not only will professions age out at an acceleratingly rapid pace, but they will not being replaced by new, related professions, because automation is removing the need for jobs in entire areas across multiple industries. At what point do we have such a level of automation that only a small subset of people need to do any actual work in order to maintain the current productivity levels? In 1960, philospher Alan Watts predicted that when this happens we will end up generating many futile jobs; totally unnecessary “busy work” jobs, created only in order to keep people working. This is an ironic idea given that, as Watts points out, the idea of automation is to avoid having to do work!
Of course, if we get to a position where the majority of people don’t have jobs because we have automated them out of existence, then where will society as a whole get the money to purchase the goods and services created by the automated systems? Alan Watts has some eloquently-delivered thoughts on this subject and on the idea of the possible need for a Universal Basic Income:
If Watts is correct, we’re in for a pretty rough time in the future, despite all the supposed progress we have made.
There’s No Time Like The Present
Let’s drag ourselves back to the present time. In the short term I think it’s fair to say that we are already seeing the need for a skillset shift to include some kind of scripting or programming, though not everywhere; it very much depends how committed to automation the company is at this point in time. That trend is likely to continue for the moment, and will go hand in hand with the development of more tools to automate infrastructure operations, whether detecting and triaging problems, or automating configuration deployments.
Some companies will not want to own and maintain their own code, so will be looking to commercial software to provide that functionality to them; these will probably be the worst in terms of job availability because they need neither operations engineers nor network-savvy programmers. I think the majority of mid-sized enterprises will opt for a middle ground using commercial software to perform some part of the automation, but driving and wrapping those tools in bespoke code, and simply not automating some tasks because the effort to do so far outweights any benefits. And web-scale companies will continue developing their own solutions from soup to nuts.
Are we all out of a job? Not yet, at least. If the prognosticators about post-industrial society are correct, the time will eventually come when jobs become more scarce, but I don’t think that’s going to be in the next few years at least. What’s still important—in every job—is to maintain a broad vision of the industry, develop new skills, keep skillsets up to date with the industry, and be flexible and willing to change direction if needed. Smaller companies will most likely keep doing what they’re doing today, and they don’t need anything else.
Whatever else happens, while we may not have reached a societal state requiring a universal basic income, we have certainly moved from “a job for life” to “multiple jobs in our life”; it’s more important than ever to be able to adapt to our environment and reinvent ourselves as the opportunity arises. And let’s hope, selfishly, that the inexorable progress towards the futile jobs predicted by Watts is slow. Really slow…
Remember, Mr Bucket was devastated when he lost his job screwing caps on to toothpaste tubes, but he was willing to adapt, and later on found a new career as an engineer maintaining the very machine that took his job. Of course, his son Charlie was also given ownership of the largest chocolate factory in the world so Mr Bucket probably didn’t need to work any more in any case, but let’s not pick holes, eh?
I’d love to hear what you are seeing, and what you think the future holds. Please do take a moment to share your thoughts, even (especially) if you totally disagree with what I’ve said!
I was inspired to share some thoughts on this topic after a friend shared a link about Alan Watts’s 1960’s Predictions about Universal Basic Income“. That page looks at some commentaries on the Universal Basic Income (UBI) from people like Elon Musk and Martin Luther King Jr. It takes a different angle to my post, but makes great reading for anybody interested in how this aspect of the future might look.
Great post, John!
While a person may have had the same title of network engineer for 20 years that doesn’t meant that the work the person did is the same work the person is doing now. My point being that people in IT have always had to change. Before it was Frame Relay. Now we have DMVPN. Before we had a multitude of physical media. Now it’s 99% all Ethernet.
Some of the “router jockey” jobs and operations jobs will definitely disappear but I don’t see it as one big event where suddenly all of these jobs go away and it will be different for different industries.
There are still companies happily enjoying life running networks with STP and flat L2 networks etc. There are also companies that just recently moved away from leased lines based on HDLC and just came into the IP world.
It’s all about perspective. Like you said. Automate if it makes sense. And it won’t make sense for all companies.
What’s important is to stay updated and motivated and to try to stay ahead of the curve.
Thanks for writing!
Great thoughts. I’m sitting here thinking: network management tools are dysfunctional and fail to automate things that should have been automated 10-20 years ago. So what’s going to be different about network configuration tools? I suspect there’s a fallacy there, and it may be the perceived value is more immediate. With network management tools, the things they do NOT do aren’t visible to management or to people who haven’t a fair amount of experience PLUS the contrariness to say “I want more”. I end up concluding: caveat emptor — or all hail the educated buyer?
For someone who is only starting in this field this article is really depressing… but i may be in time to shift.
I think we’re all mostly safe for as long as computers lack the ability to interpret & express language (and emotions to a lesser degree) and ‘think’ & act (‘work’) creatively. I’m not aware of any mathematicians worried about the calculator, or computer.
The day computers can do those things will come perhaps, but I hope that in the very long time it takes to get to that point, society, philosophy and the human race as a whole will be very different to what it is today and ready.
A couple of points… I think the future prospect is mixed, and automation by itself not major except for many-sited organizations, like retailers, hotel chains, due to inherent costs.
1) The complexity moves from the network to the automation tool. Maintaining the latter currently is often more complex. There’s also troubleshooting problems with the tool. Job time spent: shifts a bit.
2) Most sites are understaffed, and there’s a bad shortage of skilled people. Skills are key! So automation helps you find more time and get more done.
3) Most small nets don’t have enough repetition and standardization for automation to matter much.
4) Having said that, SD-WAN may make it easier and cheaper for SP’s to provide your WAN service than doing it yourself. So the change may be MSP’s for datacenter, etc. Tech and troubleshooting skills are a partial answer to that.
Where I see automation potentially having most impact (and not yet on the horizon) is network management and troubleshooting. I claim the tools are about 20% of what they should be. Slow, lacking data, expensive, hard to operate, and they don’t Just Work, they take a lot of fiddling. But that’s another discussion.
I always value your responses, Pete. My responses to your responses, in short:
1) Yes, though that will change over time I suspect, as we get better about it.
2) True enough, and that’s why – short/medium term at least – the people with skills have little to worry about.
3) At the moment. If it becomes advantageous to them to buy an off the shelf standardized block solution, that will change. Think hyperconvergence, if you will.
NM/Troubleshooting> Oh goodness, yes. We don’t have good tools.
Interesting thoughts, Steven. In general, I agree about computers thinking and working creatively. That said, AI/ML are huge research areas right now, and within narrow constraints, are making enormous progress. I emphasize the narrow constraints because typically when something gets automated or gets artificial intelligence, that “skill” is very specific and only represents a small part of the larger picture.
However, let me take your example about calculators and mathematicians and reframe it. While researching mathematicians probably welcomed the calculator as it sped up a part of their job and enabled them to do more with their time, the field of accounting was changed dramatically. The days of a pool of people ka-chinging on adding machines were no more. As computers came in to accounting, that entire room full of people becomes two people entering numbers on a keyboard. If you were an adding-machine operator, your job no longer existed. Of course, that doesn’t mean there aren’t other jobs within the field that those people could seek out, but it did change the landscape. Just like the accounting software, you will most likely never totally remove the need for humans, but you can sure as heck reduce the number who are needed, and change the remaining skills that are required.
Thanks John. I was expecting a mention of AI/ML in due course, which was very much what I was referring to with “will come perhaps”. It can be impressive but as you rightly point out, it’s very tightly focused on small, discrete specific tasks. Duplicating the abilities of a 1yr old baby entirely (which might sound simple enough until you really think about it) is many decades away I’m sure. Considering it’s cost to develop and research I also imagine it’ll be the preserve of the few who can afford it for a long time, allowing plenty of time for us/workers/society to adapt to the changes wrought by this new ‘intelligence’.
That last point applies to your second one around the calculator/computers. What you describe didn’t happen overnight or over a year. It took place over a much longer time frame, giving everyone with any sense time to plan and prepare in whatever way they saw fit, as you suggest in your piece.
Agreed about AI/ML; we won’t be replaced any time soon I suspect. And as you say, these changes are neither immediate nor total; I’m sure there are still some small business that do their books by hand, even today.
To put that in context for Javier (since it’s his comment that sparked this discussion), that doesn’t mean that anybody should run away from networking as a career or be afraid of it. Just be aware that like most industries, it’s always changing, and you need to keep your eye on where it’s moving, and drive your professional development accordingly.
There was a WEBIT event recently in Sofia, Bulgaria. One of the topics was about AI and future jobs. The proposed ideas were very similar to the ones presented in this post – artificial intelligence cannot substitute just yet the people. A lot of things can be automated but doing so flexability is being lost and this is the biggest problem.
Futhermore, automating tasks means creativity is taken out of the picture and we know how often it is crucial to business’ success.