I predicted Meta would fail in its VR venture, and Apple would struggle to sell its headset. I said Amazon would get nowhere quickly with its drone delivery plans if it gave the task to its logistics division. I predicted healthcare aspirations would not go well for Amazon, Apple or Google, and that John Lewis would fail in property development. I predicted trouble for Ineos with the Grenadier, and there are more.
I don’t try to predict often, but when I do, it’s because I have a model on which to base that prediction, or criteria against which to validate the outcome. Only then do I predict. People fool themselves into thinking they’re good at it because they get lucky once or twice, but as predictions are mostly two way choices, 50/50 is coin toss territory.
That’s why I’m ignoring most predictions about what AI will do to software development. It’s safe to say it’ll be used by developers as a tool. It’s also safe to assume it’ll provide productivity gains, but predictions of 2x or more are just guesses. What definitely isn’t clear is whether AI will increase or decrease jobs for engineers.
For coders who implemented simple tickets written by others, the writing is on the wall. But then, there was never a place for them in software development anyway. They were just a product of the flawed mass production approach to programming.
The demand for true software engineers is as likely to rise in the future as it is to fall. Some people see a dip in the jobs market and assume it’s a long term trend. They predict a continued decline into extinction. That’s one of the most common mistakes in the world of prediction: assuming a recent trend continues forever in its current direction. Generally the opposite is true.
Most “trends” turn out to be cycles, alternately rising and falling around a predictable mean. These oscillations are inevitable, because when humans see a trend they act on it, providing positive feedback to drive it beyond its natural position until an unsustainable extreme is reached. Then they realise their mistake and hurriedly work to undo the damage. The pendulum swings to the other extreme, and the pattern repeats.
That’s why any attempt to predict the future for AI in software engineering and its impact on the jobs market is pure speculation. The safest bet is to assume it has kicked off a new phase in the oscillations. Any dip will be followed by a surge, which will then give rise to a dip. This, at first wild, oscillation will settle down to a gentler pattern until something else gives the pendulum a push.
So, having said predictions are hard, I’m going to make one. In the near future there’s going to be a course correction. Companies will need more software engineers and demand will outstrip supply.
My advice?
Make sure you acquire skilled software engineers as they enter the market. Look after them and the loyalty you get in return will pay dividends, when your competitors are struggling.
As a wise colleague once said “when the tide is out, gather shellfish”