How AI enhances our Software Development workflow

Artificial Intelligence is rapidly transforming industries, and software development is no exception.

At Flock., we actively integrate generative AI and Large Language Models (LLMs) into our daily workflows where possible. This not only enhances productivity, but also redefines how we approach coding, testing, and even translating software.

Here's an inside look at how AI tools are revolutionizing our work.

Coding assistance: a smarter way to write code

One of the most significant changes AI has brought to software development is its role as a coding assistant. Tools like GitHub Copilot and JetBrains IntelliJ AI Assistant blend directly into our Integrated Development Environments (IDEs), offering real-time suggestions as we type.

• Mind-reading precision: These tools seem almost magical. As you type, they suggest the next line of code, sometimes so accurately that it feels like they’ve read your mind.
• Reduced repetitive work: Whether it's auto-completing an if statement, adjusting copied code for a new context, or even anticipating the content on your clipboard, these tools handle repetitive tasks with finesse.
• Efficiency boost: This feature eliminates mundane tasks, enabling developers to focus on high-impact coding. For example, adapting copied code to fit new contexts now happens in seconds, saving precious time.

However, there’s a learning curve. For some tasks, the traditional Google-and-Stack-Overflow approach still feels quicker.

Generating test data: automating a tedious process

Another area where AI shines is in generating test data. Building realistic datasets for testing features like scheduling or reporting can be time-consuming, especially when dealing with edge cases or large volumes.

AI tools allow us to generate diverse, high-quality test data programmatically. Recently, while working on a feature, we used AI to create datasets that mimicked real-world scenarios, including overlapping shifts and timezone differences. This saved hours of manual setup and provided better coverage, helping us identify potential bugs earlier in the development cycle.

Translating code between languages

One of the standout ways AI has revolutionized my work is in translating code between programming languages. For instance, when working with a function written in TypeScript that needs to be converted to Kotlin, AI can handle the initial heavy lifting. By simply selecting the code and prompting the AI to "convert to Kotlin," developers can generate an equivalent function in the target language. This capability isn't just limited to small snippets; entire libraries can be transposed from one language to another.

A practical example to highlight the impact: I created two Kotlin libraries based on TypeScript counterparts that were unavailable in the Kotlin ecosystem. While the AI-generated output wasn’t perfect and required additional refinement, the initial conversion process—which might otherwise take a day—was completed in a fraction of the time. Subsequent manual adjustments were also accelerated with the AI's assistance, offering suggestions for constructs and translations. This process ultimately reduced the total effort by approximately 30% (shortened from 3 to 2 days). AI was the best in facilitating a solid first iteration and saved me valuable time.

If you’re interested in the details, here are the links:

• https://github.com/kogosoftwarellc/open-api/blob/main/packages/openapi-types/index.ts
• https://github.com/flock-community/kotlin-openapi-bindings

Productivity gains for all, but especially for seniors

Senior software engineers tend to benefit the most because they have the expertise to steer AI effectively. They can quickly recognise when AI suggestions align with best practices and when they require adjustment.

Junior developers, on the other hand, may need more guidance to use AI effectively. While these tools help them learn and accelerate their output, their lack of experience can make it challenging to assess the quality of AI-generated code. This makes mentorship and collaboration even more important in an AI-enhanced workflow.

Shifting focus to requirements

One of the most profound changes AI has brought is how it shifts our focus. Traditionally, a significant portion of our time was spent on repetitive, mechanical tasks like writing boilerplate code, creating test cases, or refactoring existing code.

With AI handling these, we can now dedicate more energy to understanding and refining requirements. This shift aligns our work more closely with business goals and user needs. Instead of getting bogged down in low-value tasks, we’re able to explore creative solutions and ensure our products solve the right problems.

The challenges: hallucinations, security and understanding

While AI has immense potential, it’s not without its challenges.

• Hallucinations: AI occasionally suggests incorrect or non-existent functions, especially when dealing with niche libraries or APIs. Senior engineers can spot these errors quickly, but less experienced developers might mistakenly trust the output.
• Security concerns: Organisations like ING prohibit using tools like ChatGPT on proprietary code due to data privacy concerns. AI tools process and potentially store snippets of input data, raising questions about data ownership and confidentiality.
• Understanding AI-Generated Code: Sometimes AI suggests solutions that work perfectly but are hard to explain. For example, it might optimize a SQL query in a way that improves performance significantly, but understanding why it works requires careful study. This “black box” aspect can slow things down, especially when the code needs to be reviewed by the team.

What’s next?

The role of software engineers is evolving. As AI tools handle more coding, our focus will shift towards:

• Requirements engineering: Understanding business needs and translating them into high-level instructions for AI to execute.
• Architecture validation: Ensuring that AI-generated code integrates seamlessly with existing systems and meets business objectives.

AI tools are only getting better. As models improve, their ability to handle complex tasks and deliver contextually accurate suggestions will grow. At Flock, we’re excited to embrace this revolution and continue exploring ways to maximize productivity while maintaining high-quality software solutions.

#softwaredevelopment #softwareontwikkeling #ai #innovation #flockarticles