Innovation

Artificial intelligence enters software houses: productivity grows

Assosoftware study of 73 companies reports widespread adoption: 78% already use AI tools and another 14% will introduce them within a year

by Andrea Biondi

10 February 2026

2' min read

Translated by AI

Versione italiana

2' min read

Translated by AI

Versione italiana

There is a silent change that is sweeping through Italian software houses. It does not take the form of a sudden rupture, but that of a progressive adaptation: new tools are entering processes, flanking developers, speeding up activities. Artificial intelligence, rather than replacing it, is becoming a constant presence in daily work.

The survey carried out by Assosoftware, conducted on 73 member companies, photographs this transition. The most evident datum is the diffusion: 78% of software houses state that they already use artificial intelligence solutions, while a further 14% plan to introduce them within the next twelve months. Only a residual 8% do not contemplate their adoption.

AI is now part of the development cycle, but not uniformly so. The presence is strongest in programming, where it reaches 90% of companies, and in documentation, at 69%. This is followed by requirements analysis (53%) and the modelling and design phase (51%). Where, on the other hand, support remains more limited is in the more delicate passages: revision and maintenance stop at 30%, testing at 17%.

The result is widespread but fragmented adoption. Automation, in most cases, concerns specific activities and not the entire process. By single step, the share of automated operations often remains below 15 per cent. Artificial intelligence analyses documents, suggests code, supports design, helps detect anomalies or vulnerabilities, but does not govern the entire cycle autonomously.

Pierfrancesco Angeleri, president of Assosoftware, summarises the picture as follows: 'Artificial intelligence is already transforming the way software is developed, improving the efficiency and quality of processes. However, our survey data clearly show that it is now a technology that complements and enhances human work, without replacing it. The added value remains in people's ability to interpret the context, understand the application domain and validate the results produced by the algorithms'.

It is precisely the ability to understand the context that remains the main critical point. In testing activities, 80% of the companies report too general or false positive and negative cases. In document analysis, problems emerge with incomplete understanding of the context (78%) and interpretation of the text (67%). Also in programming, in addition to limitations in effectiveness (50%), the risk of inserting vulnerabilities in the code weighs heavily, indicated by 57% of companies.

Despite these limitations, the internal climate within companies is overwhelmingly positive. Employees' reactions oscillate between curiosity (94%) and enthusiasm (54%), while resistance and fear remain low. The impact on day-to-day work is mainly perceived in terms of efficiency: between 45% and 70% of companies estimate time savings of up to 30% in operational activities. And in 44% of the cases, there is an improvement in the quality of development.

In 64 per cent of software houses, there is no AI specialist present and there are no plans to introduce one any time soon. The chosen path is above all that of the growth of internal skills, through training courses that are still heterogeneous between companies that have started structured programmes and others that leave more autonomy to individuals.

Andrea BiondiVicecaposervizio
Luogo: Milano
Lingue parlate: Italiano, Inglese, Francese
Argomenti: Finanza, Media, Telecomunicazioni, Economia dell'entertainment, Innovazione
Scheda autore
Trust project