Artificial Intelligence

Software development, with AI efficiency growing between 25 % and 70 %. Productivity doubled by 2027

Buildo estimates: productivity is growing, but the risk of technological stagnation is just around the corner. For managing partner Luca Cioria, AI lowers access barriers to customised software, but disrupts training paths

by Letizia Giostra

27 May 2026

3' min read

Translated by AI

Versione italiana

Key points

Enterprises
The opportunities for SMEs
Today's entrants to the tech job market do not find a path
Systemic consequences

3' min read

Translated by AI

Versione italiana

In companies, the systematic introduction ofartificial intelligence tools in software development produces efficiency gains of between 25% and 70%, depending on the nature of the activities. The analysis by the Italian tech boutique Buildo monitored several projects on the subject in Italian companies in the first four months of 2026.

But for the company, this result is only the starting point: their forecasts indicate that by the end of 2026, overall productivity growth in software development could reach a range between 100 and 200%.

The enterprises

According to Buildo, the cost of customised software is significantly reduced by making it accessible also to Italian SMEs, which account for more than 90% of enterprises and of which only 19% adopt advanced technologies in a structural manner.

A study by the Stanford University documents a 20 per cent drop in the employment of developers between 22 and 25 years from the peak in late 2022. Today, the simplest operational tasks are handled by AI.

The result is a labour market in which companies look for senior skills right away, but without the pathway through which those skills were built.

The opportunities for SMEs

The cost reductions estimated by Buildo could close part of the gap, also reducing dependence on third-party vendors and standardised solutions designed for the international market. The signal can also be read on the financial markets: in February 2026, several stocks of subscription software manufacturers recorded sharp falls on the stock exchange, interpreted by analysts as an expected sign of an imminent migration towards more customised and less expensive solutions.

Today's entrants to the tech labour market do not find a path

AI handles the operational component of development with increasing effectiveness. In the past, junior profiles learned by writing code, making mistakes and comparing themselves with seniors. Today, that training phase is compressed or eliminated. Now people are looking for the Agentic Engineer, a new professional figure who is in great demand, but who already requires the experience that junior profiles can no longer acquire through traditional routes.

"It is as if automatic helpers had been introduced into a restaurant kitchen. Experienced chefs are able to handle them. But those who now start on a path to become chefs no longer have the opportunity to learn the fundamentals,' explains Luca Cioria, managing partner of Buildo.

Systemic consequences

On the computer security front, the massive production of AI-generated code introduces real risks: according to a Veracode report from 2025, 45 per cent of the code generated by AI models contains vulnerabilities, with a specific flaw rate 2.74 times higher than code written by human developers. OWASP has already published a specific list of the main vulnerabilities of AI agent-based systems.

On the technological innovation front, if AI systems learn to use existing technologies and these work, the incentive to develop new ones is reduced. With fewer developers grappling with day-to-day code practice, the engine that historically drove the evolution of tools may slow down.

"In our industry we are perhaps realising this sooner than others. AI is lowering access barriers to customised software, and this is a real opportunity for Italian SMEs. But the same change is disrupting professional training paths for those entering the market today, and increasing the risk for companies that produce code without adequate security safeguards. These are faces of the same phenomenon, and it is useful to read them together,' says Luca Cioria, Managing Partner of Buildo.