Software industry shift: three signals for Italian SMEs

Software industry shift is happening while many Italian small businesses still debate whether to invest in digitalisation. A SaaS service priced at 120 dollars a year can be replaced in 20 minutes with code generated by a language model. A company with 10 developers competes with organisations of 100. China delivered 13,000 humanoid robots in 2025, five times the previous year, with autonomous coordination systems already in production. These are measurable dynamics rather than forecasts, and they reshape costs, skills and competitive advantage.

Why a 120-dollar service disappears in 20 minutes

An engineer documented an experiment: replacing a SaaS subscription costing 120 dollars a year with code generated entirely by a language model in 20 minutes. The result was working software that replicated the essential functions of the paid service. For SMEs that sell software or digital services, the message is direct: margins on simple functionality are collapsing. When a technically skilled customer can replicate the service in less than half an hour, value moves to reliability, ongoing maintenance, integration with existing systems and qualified support. Those elements require verifiable human skills rather than automatically generated code.

For SMEs that buy SaaS subscriptions, the build-or-buy decision becomes more complex. Developing internally with AI looks cheaper than paying annual subscriptions, and the operational reality differs. Rapidly generated code accumulates technical debt at an unprecedented rate. AI writes 80% of the code in a fraction of the usual time, while the final 20% demands more refined skills. Testing, validation, integration, security and long-term maintenance do not automate with a prompt. Research on AI-assisted development shows that the challenge lies in the ability to maintain, integrate and validate the result rather than in initial generation. Strong engineers delegate code writing to AI and concentrate on architecture, quality assurance and strategic decisions, while less prepared teams produce fast, fragile code that creates problems requiring expertise to solve. For companies with revenue below one million euro, replacing SaaS with AI-generated code makes sense only with internal skills to manage the full software lifecycle. Otherwise the initial saving turns into hidden costs: unmanaged bugs, security vulnerabilities, incompatibility with updates and missing scalability.

How 10 people beat 100 developers

37signals, a company that builds project management and collaboration software, operates with about 10 people and competes against rivals employing 80 to 150 developers. The advantage comes from operational discipline built on three pillars: cutting scope without mercy, keeping the codebase clean and hiring by strict criteria. The strategy produces a virtuous cycle. With less code to maintain, developers apply more review and refactoring. A cleaner codebase reduces bugs and eases future changes. That keeps the team small, which avoids communication overhead and maintains a high average quality. Many companies assume that more features mean more customer value and higher margins. 37signals shows the opposite: reducing complexity generates measurable value, with a smaller attack surface, faster response on critical problems, onboarding of new developers in days rather than months and lower operating costs. The framework extends beyond software. Small firms in legal services, architecture and engineering consulting can focus on a few high-value services, automate low-value activities and invest in distinctive skills rather than expanding headcount. Hiring well matters more than hiring many: one senior professional who masters the whole stack is worth more than three juniors who need constant supervision.

What ignoring China costs

LimX Dynamics, a Chinese robotics company, demonstrated 18 humanoid robots emerging from crates and running coordinated routines in full autonomy. The system, a Cognitive Operating System for Agents, manages perception, memory, basic emotional states and multi-robot coordination without human supervision during execution. While this happens in China, many Italian manufacturing SMEs still debate whether to adopt a single traditional industrial robot. The competitive gap is strategic rather than technological. Data from the Digital Innovation Observatories of Politecnico di Milano shows that about 60% of digital implementations in Italian SMEs fail because the existing operational flows are not understood, not because of technology limits. Companies try to automate disorganised processes and obtain digitalised chaos. Autonomous coordination of multiple systems requires sophisticated software architectures, and when 18 robots work together the value lies in the system that orchestrates them rather than in the individual machines. For small manufacturers, the critical investment lies in the skills to integrate robots into existing processes rather than in robotic hardware. Businesses in construction, logistics and component assembly should ask whether they collect data on their production processes and whether they have visibility on cycle times, bottlenecks and waste. Without that baseline, advanced automation becomes a blind bet. China is not winning because it has superior technology alone; it invests systematically in the digital infrastructure that makes the technology usable. Italy remains the second European market for installed robots and the fifth worldwide, with 8,783 new units in 2024, and 90% of manufacturing SMEs with 50 to 249 employees report using robots. The gap is in developing and deploying advanced autonomous systems.

What to do without improvising

The three signals converge on one lesson: the cost of producing software is falling, and the complexity of using it well is rising. The opportunity is to build or buy software capacity that limited budgets could not previously reach. The risk is accumulating technical debt, vulnerabilities and operational fragility at an unprecedented rate when skills to validate and maintain the output are missing. The build-or-buy evaluation with AI requires honesty about internal skills. A company without someone able to read code, identify vulnerabilities, manage integrations and test systematically should prefer verified SaaS subscriptions to fragile software when reliability matters. Reducing complexity, as 37signals does, means cutting scope before automating. Many SMEs carry redundant processes layered over the years, and the useful question is which work genuinely serves the customer and which exists to cover earlier inefficiencies. Automating the wrong process produces faster errors. Closing the gap with advanced automation requires preventive analysis: mapping where time, margin and opportunity are lost today. That mapping benefits from specialist skills, because a professional systems integrator sees common patterns and errors that a first attempt misses. Hands-on support pays off when it accelerates learning and reduces risk, identifying in three weeks problems that an internal team would discover after months of costly attempts.

One of those signals is uncomfortable to name: the decline in software quality is now visible in what companies buy every day.