It works and rather well, in fact. But not well enough to convincingly assure us that it can deliver on a promise: to take care of everything. And yes, “thinking” would be its reason for being, so to speak, considering it doesn’t have a real life, but it exists for that purpose nonetheless.
Globally, it’s AI (Artificial Intelligence); here in Italy, it’s IA (Intelligenza Artificiale), but the difference is minimal. It’s the increasing ability of computers to write documents for us – though ridiculous blunders do occur, requiring human oversight of everything the IA produces – and to perform tasks autonomously, particularly repetitive ones. Everything, in our place.
The use of artificial intelligence software serves, among a thousand possible uses, to grow the family – not ours, but its own. Yes, because the IA software also performs the task of writing software of the same “IA family,” which is banal but also a little unsettling. Software companies – in the broadest sense, extending to smartphones or car computers – have seen AI as an opportunity to develop and sell software written by the Intelligence itself.
Is this bad news? Does it mean that humans are left to wander aimlessly, since if you don’t work (because the machine does it for you), you don’t earn, and if you have no income, you can’t even pay your bills? Not exactly, not yet. That’s where we’re headed if we don’t come up with something, but it will still take time. Hopefully, more than a little.
Meanwhile, let’s look at how to frame the phenomenon, and that’s where a study by Assosoftware comes in. These are people who know what they’re talking about, as it’s the Italian Association of Software Companies, and this research – which its authors define as the first ever conducted on the sector – involved 73 associated software houses. The results reveal a silent advance, with AI already structurally present in the software development cycle of Italian companies in the sector.
Specifically, AI is most used and tested in writing code and technical documentation, because it reduces human labor and speeds up processes. To put it in numbers, the Assosoftware study indicates that 78 percent of software companies already use AI, and another 14 percent plan to introduce it within the next twelve months. Only 8 percent will continue to develop software using only the human mind.
However, the use of AI for writing new software varies considerably in the different phases of a program’s development. The programming phase is dominant: AI is used for 90 percent of it, and also for 69 percent of the documentation, 53 percent for requirements analysis, and 51 percent for modeling and design. In contrast, when looking at use in the review and maintenance phase, it drops to 30 percent, while for testing, humans still trust humans: AI accounts for only 17 percent of the work.
The contribution of artificial intelligence is growing steadily, that’s certain, but not uniformly: the level of automation of development processes remains limited and concentrated on specific tasks. For example, AI is used for analyzing documents and requirements, but also for other tasks: composing prototypes, reviewing code snippets, generating synthetic test data, and detecting bugs and vulnerabilities. However, it’s rarely seen taking charge of the entire end-to-end process. In other words that artificial intelligence is, for now, more “assistive,” integrated into human work, than “substitutive.” AI is, for now, more of an operational support with levels of automation that rarely exceed 30 percent. Few cases surpass 50 percent.
What are the biggest problems encountered by those who use AI to write software? The Assosoftware study says the main critical issues are related to understanding context and the ability of AI systems to interpret. Eighty percent of software houses report the generation of overly generic cases or false positives/negatives in tests, 78 percent report difficulties in documentary analysis and full understanding of requirements. There are also risks related to the introduction of vulnerabilities in the code and limitations in the abstraction of domain rules. But artificial intelligence helps productivity, requires personnel updates, saves time and therefore money. It also, unfortunately, allows for the development of a workload that would require a few more people without AI.
The Italian industrial landscape is witnessing a growing adoption of AI-enabled solutions, mostly driven by cost efficiency considerations, according to a 2022 study. There are early indications of leveraging AI to increase the offer portfolio, generating more revenues. New technologies supporting AI in small, embedded, and distributed AI are proving useful in leveraging data. A concurrent push by the Italian Government to foster AI and promote standardization and data sharing (the GAIA-X initiative – Italian Regional Hub) will provide further impetus to increased AI adoption.
The Italian Strategy for Artificial Intelligence 2024-2026 defines a new program that builds on previous experiences, sharing the principles of reference for an Italian Artificial Intelligence that fits into a European context and promoting the development of anthropocentric, reliable, and sustainable solutions. As of May 2025, 63 percent of large Italian companies have already adopted or are planning to adopt AI, a level of adoption that could translate into an overall productivity increase of 115 billion euros for Italian companies.
