You don't have to look far to see the argument playing out in real time. Some of the biggest software companies in the world - Salesforce among them - have taken meaningful market hits in recent months, with AI increasingly cited as part of the reason, and what's notable is that this isn't a story about companies failing.
It's a story about the world around them changing.
'Do we still need this? Could we build it ourselves? Can we just ask for the outcome instead?' - these questions didn't really exist two years ago in any serious sense, and now they're present in almost every substantive conversation about software spend. The speed at which they've become normal is, in itself, significant.
What's revealing is how unevenly the pressure is landing. It's showing up most clearly in software that sits in the middle: tools built around structured inputs, predictable workflows, and relatively fixed outputs - the kind of software that made complete sense when interacting with systems meant navigating forms and fields and defined processes, but which starts to look optional the moment interaction becomes conversation. The interface was the product, more or less, and now the interface is compressing.
At the same time, the deeper systems aren't weakening - they're being pulled into focus, because once you remove the interface layer, the questions that remain are considerably harder. What's actually underneath this? Is the data consistent? Are the systems genuinely connected, or just adjacent? Can the output be trusted without someone checking it against something else? These questions don't get easier in a world with more AI capability - they get harder, and they get harder by orders of magnitude, not by degrees.
This is why what is happening in enterprise systems doesn't feel like a collapse. I t feels more like a separation between software that was valuable because it was necessary, and software that's valuable because it reflects the real complexity of how a business actually runs. A nd the market is beginning, without much fanfare, to recognise the difference. This isn’t happening with fireworks and fanfare, but it is happening steadily enough that the overall direction of change is clearly legible.
If the first shift was learning how to use software, and the second was software becoming easier to use, this one is structurally different from both of them. We're asking software to think with us - to handle ambiguity, to work across functions, to return something useful rather than merely accurate - and that changes what we actually need it to be, at a level that cosmetic updates to existing products cannot reach.
John Wills
COO, Shipnet