Artificial intelligence (AI) can be a hero or an anti-hero in the development of software, depending on your implementation approach. According to the 2024 Dora accelerate the state of the DevOps report, The adoption of AI is promising with 75% of respondents reporting positive productivity gains, but also reveals unexpected challenges, including a 7.2% reduction in software delivery stability.
The key is to position AI as a support partner. AI must be used to improve human expertise rather than replace it.
AI is promising, but let’s stay suspicious
The AI quickly infiltrated software development for almost two years now. Coding assistants, automated test tools and AI analysis promise to transform how we create applications, and some early adopters have shown impressive results. When the AI seized the development workflows for the first time, it addressed well -defined problems such as code completion, simple test generation, comments and documentation project with reasonable success.
But this initial enthusiasm has given way to more complex realities. The development of modern software involves complex systems, multi-cloud deployments, complex safety requirements and accelerated delivery expectations. The AI has shown both remarkable capacities and limitations concerning this environment.
After hearing stories of customers on dozens of AI implementations, I found that the story was not as binary as “hero” against “bad guy”-he looks more like “hero” compared to “anti-hero” erratically-trobbled-si-victoire “. The reality of AI is nuanced – and much more interesting, according to many of the implementation approach, the governance scheme and where organizations choose to apply AI technologies.
The work problem is real and measurable
Each developer knows the frustration of “work” – these manual and repetitive tasks that consume time without adding a lot of value. Work is not only a minor discomfort. Dora 2024’s report confirms that this challenge remains persistent, the developers continuing to fight with the time allowance between precious work and repetitive tasks.
Recently, I had the chance to discuss the heroism of AI and anti-heroism with the product manager of Cloudbees, Shawn Ahmed. Cloudbees is a supplier of Enterprise DevOPS solutions and was co -founded by the creator of Jenkins Kohsuke Kawaguchi, he was therefore in good position to observe the thousand ways whose AI was injected into the development of software. Their internal research indicates that the developers spend about half of their time triacing the failures of the tests, pending construction times or solving safety problems – and as little as 20–30% of their time writing code. As Ahmed pointed out to me, this ineffectiveness seems to not be tolerated in most other industries; It’s like accepting a manufacturing line that has only 20–30% efficient.
An example of the Cloudbees approach to incorporate AI in a productive and useful way is their continuous integration capacity (CI). It analyzes the historic data of the pipeline to identify the models in successful deployments and uses AI to provide suggestions for future deployments. This eliminates configuration working hours, thanks to AI’s ability to switch the recognition of models over thousands, or even millions of data points.
From the point of view of Cloudbees, AI in software development is not a replacement technology in Dev but rather what Ahmed characterizes as a “companion”. The best role of AI is to manage routine tasks to preserve and extend the time of the human developer to carry out creative and strategic work.
Three implementation models, three different results
It is clear that all the implementations of AI are not created equal. I see three emerging distinct approaches, each with radically different results:
- Replaceers: Some organizations approach AI to reduce workforce by automating development tasks. This model generally provides poor results, fundamentally including the capabilities of AI and the creative nature of software development. Dora’s research supports this concern, showing that although the adoption of AI is widespread, 39.2% of respondents said they had little or no confidence in the code generated by AI, which indicates that successful implementations require human surveillance.
- Tools first!: These organizations adopt individual AI tools without coherent strategy or governance framework. This approach gives mixed results – productivity improves in specific fields, but in many cases at the cost of quality, security or long -term maintaability. The results of the Dora highlight this challenge, showing that despite the positive impact of AI on individual productivity and the quality of the code, it can lead to a decrease in software delivery performance.
- Augmentators: The most successful approach treats AI as an amplifier of human capacities rather than a replacement. Dora research validates this model, showing that the adoption of AI leads to significant improvements: when the adoption of AI increases by 25%, individual productivity increases by 2.1%, flow improves by 2.6%and work satisfaction increased by 2.2%.
Cloudbees is built on this philosophy of increase. According to Ahmed, AI manages the recognition of models and repetitive tasks while humans make decisions on strategic issues. Their DevSecops solution demonstrates this approach using AI to identify security vulnerabilities while preserving human authority on correction strategies. Cloudbees’ value flow analysis uses AI to identify the ineffectiveness of the workflow while leaving decisions to improve processes to teams that include the organizational context.
Make the hero of your Dev Story application
Here are five critical success factors that determine if AI becomes a hero or an anti-hero for you:
- First target high altitude activities. Successful implementations begin with well -defined repetitive tasks where AI can offer immediate value. Cloudbees’ recommended targets include the generation of tests, optimization of pipelines and code magazines, including areas where AI can considerably reduce manual effort.
- Establish governance first, deploy the second AI. Organizations seeing positive results implement governance executives before the generalized AI adoption, the definition, for example, the appropriate use cases, the verification requirements or the clear limits for the autonomy of the AI.
- Invest in the education of developers. Unlike certain stories, AI’s success requires more qualified developers. The most efficient organizations invest in training programs that help developers understand the capacities and limitations of AI.
- Integrate rather than isolate. Rather than adopting unlocked ad hoc solutions, effective organizations implement AI capabilities that integrate throughout their delivery pipeline. The CloudBees platform adopts this approach, connecting AI’s capacities from the analysis of requirements through production monitoring.
- Measure the significant results. The most effective implementations have exceeded the measurement of code generation measures. Instead, they focus on quality improvements, improvements in safety postures and commercial impact indicators that reflect the true value of AI adoption.
AI is your enthusiastic partner, not your replacement
The AI successful implementations in App Device seem to share the common characteristic of using it as a complementary force rather than replacement. AI is suitable for banal and repetitive tasks that drain motivation and attention among developers and increase their work. It allows humans to focus on innovation, customer experience and, ultimately, commercial value.
According to Ahmed, Cloudbees plans a future where AI becomes more and more personalized for individual developers. He predicts that within two years, developers will probably work with several AI companions adapted to their specific preferences and their specific work styles. This aspect of personalization represents an important evolution beyond the more standardized AI tools today.
A particularly remarkable perspective that I obtained from research and discussions on sellers is the relationship between the adoption of AI and the team scale. While some organizations initially address AI to reduce the workforce, the data suggests the opposite effect. As Ahmed noted in our discussion, the most effective approach is not to use AI to reduce a team of 10 people to 5, but rather by taking advantage of AI to help this team of 10 people to achieve what would have previously had 15 people. This is the proactive perspective and creation of companies “on a scale, and not to lower” which can help to propel an organization to its objectives.
I concluded that today, the hero of software development is not AI. Not alone, but it is no longer a purely human expertise either. The real hero is the partnership between them, working in concert to create results, neither could achieve independently.
