Efficiency and productivity in software development have been a constant challenge, even more with the arrival of AI. Raúl Fernández, Director of Operations of LEDAMCanalyze how to measure the real value in software projects the same as that of the impact of AI that is transforming IT management.
Interview with Raúl Fernández, Ledamc Operations Director
The debate on efficiency and productivity in software development remains open, even now more than the use of AI has extended so much. Regarding this, Raúl What do you think of studies such as Denisov-Blanch, who try to measure the productivity of developers?
The Denisov-Blanch study puts the focus on a known problem: not all developers contribute the same value. But measuring productivity in software is not easy. Taking lines of code is useless, because we know that a quality code is usually more compact. In addition, the most experienced programmers invest time in meetings, problem solving and testing, key activities that are not always registered.
However, the AI is changing the rules of the game. While some fear that developers can replace, the reality is that those who know how to take advantage of generative tools can significantly increase their productivity and efficiency. According to Stack Overflow, more than 60% of developers are already using these tools, so in the end, the threat to a programmer is not AI, but another developer who knows how to use it better.
So can productivity be objectively measured?
The key, for us, is in the product. What matters is not how many lines of code are written, but how much useful and quality software is generated. This approach, based on ISO/IEC standards, is more reliable and is already used by governments and companies worldwide to identify real talent and optimize development.
And have problems arise arise to carry it out?
For years, poorly defined requirements have been a recurring problem in software development.
It is undoubtedly the paradox of this sector, in which we dedicate years to maturing processes, we documented, best practices … and, nevertheless, the poorly defined requirements continue to cause complications in the software projects.
And why is it so difficult to write clear and complete requirements, even in companies with mature processes?
Well, because, in the end, writing a good requirement is not only a technical issue, but also a matter of communication and understanding between the parties involved.
And how can it be solved?
One of the biggest problems is ambiguity. Many times, the problem is not the lack of information, but this is not structured or expressed clearly. And the same requirement can be interpreted differently by the business team, developers and tests, which generates problems from the beginning.
Also the business alignment fails. The requirements are written with too technical language or, on the contrary, too generic. And if there is no clear connection between what the business expects and what the technical team builds, cost overruns and delays appear. According to the last chaos report of the Standish Group, only 31 % of software projects end successfully in terms of time, scope and budget, and, finally, poor requirements are one of the risk factors.
It was all this that led us to see the need to incorporate in Quanter, our application of estimation and benchmarking of software projects, a functionality that helped improve the definition of requirements.
How does the improvement of requirements with generative artificial intelligence work?
In a tool like Quanter, having a good definition from the beginning is a necessity. The better structured the requirements are, the more precise the effort and cost of the project will be and the lower the risk of deviations in subsequent phases.
The new improvement of Quanter requirements is based on generative to optimize the drafting of the requirements before proceeding to its estimate. It consists of introducing the requirement in natural language and Quanter analyzes it and suggests improvements based on best practices, industry standards and/or criteria of the organization. It also explains each improvement, helping teams understand the impact of these adjustments.
