The title is obviously clickbait, but not quite an empty question.
There are many things happening with AI around us right now.
There is the wet dream of AGI that all the big names are dangling in front of our eyes to boost their company valuations. Empty promise? Hard to say, but I have doubts we can reach it with the current tech.
There are the companies that were quick on the trigger to fire people and said they would replace them with AI. There seems to be a wave that comes and goes around the industry. It feels a bit premature to me, but these are also companies that have bet big on AI and therefore need to show that they mean it. They might be on to something, but time will tell.
There is the urge to use AI for everything, to boost productivity, to make everything better, to support UBI (Universal Basic Income), so we can all slack ourselves into permanent leisure.
Just as everyone uses AI to automate different tasks in their daily work, there are now infra platforms that use AI to do all kinds of stuff, from incident management and monitoring to writing code and opening PRs to fix vulnerabilities and issues. Much of this is like the digitalisation process I wrote about in a previous post: we are just replacing human actions with AI (and, to be cynical, we later need to review the changes or fix them ourselves).
We need to rethink knowledge work
We need to rethink how we work when AI is a first-class citizen of a company. And we need to do this while taking into consideration what AI is actually good for, not what its makers say in their marketing posts.
AI is a revolutionary tool that muddies the borders between roles in a software company. With AI, any role that existed so far can extend its capabilities and take on a wider range of tasks, within limits. These limits stem from the current models themselves: though they’re very good at finding the most likely solution to a problem, they lack discipline, common sense, quality, and empathy. All of these are essential in a knowledge worker’s day, regardless of role or career ladder. This means AI has a hard time replacing roles where these qualities are required.
But it can extend human capabilities. One interesting move some companies seem to be making is to broaden the roles of the people who build their products while choosing architectures that keep infrastructure and platforms thinner. Yes, welcome back monoliths! Backend and fullstack engineers get to do some infra as well, all empowered by AI. As a result, these companies can have fewer platform teams and simpler platforms that don’t require as many people as before, while accelerating growth.
I’m sure we’ll see similar shifts in other functions too, not just engineering. Platforms for functions like finance or human “resources” are notoriously bad, and I wouldn’t be surprised if, as LLMs evolve further, formerly non-technical people start building their own mini-platforms on top of managed services from large providers.
We need to rethink what is possible and potentially re-evaluate the roles we have in a software-based company. There will be many learnings from the startups beginning today that take AI as a first-class citizen while keeping an open mind about how they organise themselves. It’s harder for established companies to do this rethinking while continuing to serve their customers, but it’s almost necessary for new companies to do it.
When it comes to platforms and AI, I don’t think internal platforms will disappear, but I do think they will be somewhat simpler. Platform engineers will become more architects than actual platform builders, setting the stage for how platforms should come together and leaving it to the AI-empowered engineers in each team to plug their products into, or build on top of, these platform foundations.