Artificial intelligence has become part of the daily routine of software development teams. From code and documentation generation to debugging and automating repetitive tasks, AI-powered tools are now present in almost every technical team.
A survey conducted in May among Essensys’ technical team reveals an interesting paradox: the more developers use AI, the more aware they become that real impact comes not from the tool itself, but from the way it is integrated into well-defined workflows.
AI is already part of everyday work
Approximately 70% of Essensys developers have incorporated AI-powered tools into their daily routine.
While most acknowledge that these tools help them work more efficiently, they believe AI delivers real impact only when used within a well-defined framework, supported by clear processes and human validation, according to an internal survey conducted in May among Essensys’ technical team.
This conclusion is particularly relevant in enterprise environments, where quality and security remain essential.
“AI has already changed the way technical teams operate, but its real value no longer lies solely in the speed of code generation. It lies in people’s ability to validate, interpret and make sound decisions. The more teams use AI-powered tools, the more aware they become of the need for control, verification and human accountability,” says Mihai Matei, CEO of Essensys, a technology company specializing in enterprise software solutions and complex digital transformation projects.
Almost half (48%) of respondents say they are curious to better understand how these technologies work and how they can use them more effectively in their day-to-day activities.
For 40% of respondents, AI represents a professional development opportunity or a competitive advantage, while 19% view the evolution of these technologies with enthusiasm and believe they can bring significant changes to the industry.
Enthusiasm and concerns at the same time
“Perhaps the most relevant finding of the survey is not a single percentage, but the combination of them: 62% of respondents selected at least one positive and one negative sentiment about AI simultaneously. We’re not talking about blind enthusiasm or outright rejection, but both at once. Developers clearly recognize the usefulness of these tools, while also having genuine concerns about them,” adds Mihai Matei.
Main Concerns: subtle errors and data security
The most common concern identified is the risk of subtle, difficult-to-detect errors in AI-generated code.
Nearly half of respondents (48%) say this is the primary risk they associate with using AI tools. The next most frequently cited concern is data security and confidentiality, mentioned by 43% of participants.
The survey results also show that developers have become more cautious when reviewing AI-generated outputs.
Most respondents say they always or usually assess suggestions provided by these tools with a critical mindset, a behavior that directly reflects their concerns about code quality.
In Enterprise Projects, processes make the difference
In enterprise projects, development based solely on AI-generated code does not create meaningful impact.
Architecture, security, integration and quality control remain responsibilities that cannot be delegated to a tool.
The Clarity Framework by Essensys, a methodology refined over the company’s 21 years in the market, provides exactly that foundation: clear processes and the structure needed to turn the benefits of AI into real business value.
This trend is not unique to Essensys. According to a recent Stack Overflow survey conducted among nearly 50,000 developers, 84% of respondents either use or plan to use AI tools in their work, and more than half use them daily.
However, as adoption increases, trust in AI-generated results declines, only 29% of developers say they trust the accuracy of the outputs generated by these tools.
For Essensys, the internal survey results confirm a direction the company is already following: AI is a valuable tool, but its impact in complex projects depends on the people and processes surrounding it, not on the speed at which it generates code.
