End-to-end AI Solutions

I like a lot the slogan “AI is Just Software”.

I came to it some five years ago after meeting an obscure engineer—an extremely smart guy used to his job and never pompously writing on socials as I do 😊. Well 5 years ago his team created a solution to predict the price of renewable energy on the market, a formidable tool that helps energy companies scheduling maintenance on the power plants and planning budgets.

I was, in a way, honored to meet such a smart person and he was, in a way, eager to talk to another person (me) who had a good reputation in the software industry. “Yes” he said “we’ve done pretty good things with data and algorithms. We’ve arranged a pipeline that regenerates the models regularly always on fresher numbers, but in the end the hardest part for us was the software around it.” And then a few minutes of conversation later, the gorgeous slogan came out. “Man, at the very end of the day, AI is just software.”

What did he (and me) really mean by “AI”?

We actually meant an end-to-end solution, namely a product of its own or a module in a larger digital product. In the folds of such a solution, we inevitably find shallow algorithms, complex models like flavors of neural networks and plain pieces of code to calculate aggregates, check conditions, route messages to other participants in the overall product.

What did marketers really mean when they say things like “All companies must use AI”?

They’re probably unaware of it, but all they actually mean is that companies should become more digital and smart by implementing solutions based on advanced forms of data classification and prediction and, more recently, auto-generated material.

So, an AI solution is a pipeline of software modules, a few of which are black-box components internally made of sophisticated data models. ChatGPT and all cognitive AI components (i.e., those in Azure AI) are all such examples.

The rationale behind “AI Is Just Software” is that plain AI black-boxes alone are meaningless for customers. Customers, instead, need end-to-end solutions in which the power of AI black boxes is unleashed within the boundaries of a software-based pipeline that solves a business problem.