@tjrobertson52

It's clear that AI is quickly changing the way we work, even if we can't see exactly how that will happen in the next few years. But looking at the advancements we've seen in just the last two months, I think it's finally starting to take shape. So I'd like to look back on how knowledge work has evolved so far and then predict what's coming next.

Everyone likes a good speculation video.

The story of how computers have evolved has been a story of abstraction. Machine code was abstracted away by human readable code and then we got the command line interface, and that has been abstracted away by modern operating systems. Nowadays to use a computer, you no longer need to know how to program it or even how to use the command line interface.

Of course, all that code is still being run, all those commands are still being executed. But unless you're doing something highly technical, you don't need to be aware of any of it.

Well, I think the same thing has already started happening with modern software and it's only going in one direction. We're quickly moving into a world where you don't even need to use the software. AI will just handle the software layer and unless you're doing something especially technical, you won't even have to deal with a modern operating system.

If that seems hard to imagine, let me point out how this is already happening. One of the biggest early use cases has been research. It used to be if you wanted to research essentially anything on the internet, you would go to Google, perform a series of searches and look through a dozen or more web pages.

Nowadays, unless you absolutely need to verify the facts yourself, it just makes way more sense to have the AI do this research for you. That's why website visits are way down right now. And of course, the better the models get, the less sense it'll make.

The same thing has also already happened with tools like Photoshop or PowerPoint. AI is just better than 99% of us at this kind of work now. Unless you're doing something very specific, it just doesn't make sense to use these tools yourself anymore.

And again, the better the models get, the less sense it'll make.

So my prediction is one or two years from now, most software won't be made for humans, it'll be made for AI. But just like the command line interface still exists, we'll still need some interface for humans. The websites aren't going away.

It's just that fewer and fewer humans are going to be using your website and more and more they'll be used by AI.

So it might be about time to start thinking about your website as a database and a tool for AI. Is all the information anyone could possibly want on your company, somewhere on your website, clearly organised and labelled as someone who reviews company websites on a daily basis? I can say almost certainly it isn't.

And honestly, it hasn't made sense until recently. The average human is not gonna look at more than 2 pages of your website. They're not gonna read thousands of words about your company, but AI will.

Can AI easily make all the transactions a customer might need to make with your business through your website? It looks like web MCP will probably be the standard for these kind of interactions. If AI can't easily do business with you through your website, there's a good chance of using your competitor website instead.

Because in this future, I'm imagining, unless the human is really motivated, they're not going to the website themselves.