“Hey, we have a new feature requirement, how fast you can deliver this?” Me, “It will take couple of sprints.” “Why don’t you use AI, it should help you delivering the feature in not time.” To tell the truth, I'm a skeptic. As a software engineer for over 25 years, I've seen witnessed the changing landscape of software technologies. Every new tool has the potential to be the next big thing, the game changing revolution. While generative AI and tools like ChatGPT were gaining traction, I initially discounted it as another shiny object. My first few attempts with Copilot gave me mixed results. When I asked it to write a new script with a very clear objective, it was fantastic. In a matter of seconds, it gave me a solid code, it was a nice jump start. Then trouble started when I noticed a small, trivial bug and trying to fix it by talking to the AI was often more work than just fixing it myself. Frustrating at times, I felt as though I was correcting my new "helper" more often than it was saving me. But I kept experimenting, some of that was forced by new mandates at work and that's when my viewpoint started to change. I started realizing that I was expecting the GenAI to perform at least at the same level as I am implicitly. I was expecting a perfect, senior level expert who can just read my mind with couple of sentences as prompt. Instead, what I discovered was a valuable junior teammate who is fast, sometimes incompetent, but always available. Most of software development community is navigating this journey from skepticism to practical application and make sense of all the hype. As these AI tools are integrated more into our daily workflows, we will able to see the outcomes the good, the bad, and the truly surprising. I genuinely believe that we are not being replaced but our roles are being augmented with a new partnership which will change how we build, learn, and lead. The real discussion of this article isn't about how fast AI writes code, but about how we, as engineers, create real value for our businesses in this new era. GenAI Autocomplete or More: A Tour Through the Workflow Real impact of GenAI is impacting all stages of the software development lifecycle (SDLC), not just code development. Let's break it down with practical examples. Phase 1: The Vision (Design & Planning) Phase 1: The Vision (Design & Planning) Before writing code, we need a clear architectural design, technical specifications and manageable chunks of work i.e. user stories. Creating all these documents is usually a laborious manual process. So, the use case we have for GenAI is to act as a product manager and brainstorming user stories and requirements. How GenAI helps? Imagine you're building a new e-commerce feature for "Wishlist” and you have to write the user stories. You can prompt an AI tool with, "Act as a product manager. Write 5 user stories for a customer Wishlist feature, including one for a guest user." The AI will instantly generate well-structured stories giving you a great starting point, saving hours of work. This doesn't replace the product manager, as you still have to make sure the stories are in-line with your feature and they work with overall product. The other use cases we have here are generating or adjusting architectural design and technical specification, where architects and engineers can utilize AI tools in similar way to achieve their goals. Phase 2: Accelerating Development (Coding) Phase 2: Accelerating Development (Coding) Utilizing AI in development is not about the AI replacing us, but it is about assisting us. In the development phase, you are either building something completely new or implementing a new feature in existing code. Can I use GenAI to "Jump Start" on the development of a new feature? As I discovered in my own experience, GenAI does a good job in providing initial implementation. If I need a Python script to read data from a CSV and call an API, I can write prompt explaining this, and a tool like Copilot will generate the initial code. What I do then has been always a senior developer's job: to review, test, and refine that code. This speeds up the development allowing faster releases. A study by GitHub found that developers using Copilot complete their tasks up to 55% faster. Another use case is of the GenAI as "Pseudo Code Expert" for existing codebases. Given a code repository, AI was able to quickly scan, analyze and tell me what is happening in the code. If the code repository follows good readability standards, you will be amazed with the output. This was a great benefit. How it works? When you are trying to understand a large, unfamiliar codebase, you don’t even know where to start. With tools like copilot, I was just able to give it whole repository and ask the questions like, "What is the purpose of the UserService.cs file?" or "What type of database this solution utilizes?" It was able to provide me direct answers, acting as a powerful search engine and interpreter for the code itself. Within minutes, I was able to understand general structure of code and helped me becoming productive. The business value is clear: faster developer onboarding and increased team efficiency. Phase 3: Accelerating Quality (Testing & QA) Phase 3: Accelerating Quality (Testing & QA) Testing is one area where I've been not so skeptic of GenAI. Writing tests is critically important, but let’s be honest, it can be tedious and often we all miss test scenarios due to one reason or other. The speed and power of GenAI are not just helpful but it’s truly mind blowing. This directly results in faster and more confident shipping of higher quality products. The AI can generate unit tests quickly for any given code or even for give user story. With simple prompt like, "Write unit tests for UserService.cs. Include the edge cases, like null inputs and empty arrays," I will get the complete test suite and all I have to do is review and refine. Another game change is Creating Rich Test Data. I can’t even count how many times have I needed to create a fake user profile with realistic looking names and emails? Now, it’s like a magic. Give prompt, "Generate a JSON array of 50 user objects, defined as in class User.cs and voila you have the test data delivered by AI instantly. The Always Available Mentor: GenAI for Learning I've always found that the best way to solidify my understanding of a new concepts is to discuss it with someone who is also interested in the subject. But finding someone available is harder than ever as I work with globally distributed team. So, I turned to GenAI and thought to use it as that partner. This is where I found another unexpected role for GenAI: a tireless, knowledgeable discussion partner. Tools like ChatGPT are helping me solidifying my understanding. I asked questions and let it ask questions. I discussed scenarios that required me to research about the concepts and validated my understanding.  I have been using AI as sounding board if no one else available. This doesn’t eliminate the need of other humans around me as the responses from these tools is still limited by the input I provided and others can help me look beyond the scenarios I thought of. I've been using these tools more recently to help me with my tasks like peer reviewing journal articles. I just ask the AI, "Explain the concept of Retrieval Augmented Generation (RAG) in simple terms," when I come across a concept, I am not familiar with. It gives me a nice summary, to which I can keep drilling down. It helps me perform my job better and makes me a more knowledgeable engineer. Beyond the Code: Growing Significance of AI Other than improving software development lifecycle, AI is showing promises in engineering operations. The observability platforms are using to correlate alerts across multiple services which quickly helps in identifying likely cause of outage, drastically reducing the time it takes for resolving issue. Cost Management is another area where AI powered tools can monitor cloud usage costs. Tools can be easily built adjust the resources to optimize the cost, without impacting the state of a system. It is impossible to cover each and every use of AI for engineering. Some uses worth mentioning are Advanced Security, Documentation and Knowledge Management, and Governance and Compliance. The Necessary Friction: Real Challenges and the Human Role As I am realizing the value of GenAI as a tool, my skepticism isn’t gone completely. It has simply shifted the focus on the new challenges it introduces. 1. The Hidden Cost of "Good Enough" Code 1. The Hidden Cost of "Good Enough" Code My biggest fear with AI generated code isn't a dramatic security breach, but the slow decay of code quality. I've seen developers introduce the code that breaks the established patterns of our repository while using AI. They don’t look at is as critically as they would when the code is written by their other peers. This blind faith in AI generated can have direct financial costs. In one case I came across, an engineer implemented a suggestion from Github Copilot that solved the immediate problem but introduced a new, AWS resource that added to our monthly cost. 2. The Evolving Role of the Engineers 2. The Evolving Role of the Engineers If the AI is the tireless junior teammate, the engineer’s value shifts toward a higher level of oversight: The Quality Assurance: In a world where anyone can generate volumes of code using AI tools, the engineer becomes the guardian of the codebase's quality, consistency, and maintainability.The Architect: The engineer has to strike a balance between enthusiasm with prudence. This means protect from costly suggestions and, above all, establishing data security governance. Protecting Sensitive PII data. We saw the risk when employees at Samsung accidentally leaked proprietary source code to ChatGPT. The architect’s role is to establish clear guardrails and push for secure enterprise solutions.The Mentor: The engineer's role as a teacher is magnified. They must actively mentor others on how to use AI tools responsibly, ensuring they don't just copy and paste but truly understand the code. The Quality Assurance: In a world where anyone can generate volumes of code using AI tools, the engineer becomes the guardian of the codebase's quality, consistency, and maintainability. The Architect: The engineer has to strike a balance between enthusiasm with prudence. This means protect from costly suggestions and, above all, establishing data security governance. Protecting Sensitive PII data. We saw the risk when employees at Samsung accidentally leaked proprietary source code to ChatGPT. The architect’s role is to establish clear guardrails and push for secure enterprise solutions. The Mentor: The engineer's role as a teacher is magnified. They must actively mentor others on how to use AI tools responsibly, ensuring they don't just copy and paste but truly understand the code. Conclusion: The Human in the Loop is The Most Important Feature When GenAI started gaining popularity, I was looking for a reasons to dismiss AI instead of looking for ways to use it. I was asking the wrong question, "Is this tool good enough to replace me?" What I should have been asking was, "How can I use this tool to become better at my job?" The answer is clear to me now. GenAI is not a self-sufficient specialist. It is a strong force multiplier. The tireless assistant can produce a thousand lines of test data, and the junior teammate can manage the first draft. The experienced tutor is the one who can explain a novel idea at two in the morning. It definitely needs a human in the loop to be genuinely effective. It requires our experience to prevent its costly errors and our skepticism to challenge its output.  It takes our expertise to create software that is reliable, secure, maintainable, and functional. Software development is not going to be a human versus machine contest in the future. It's a collaboration. Engineering is a changing craft. Engineers who can not only write code but also expertly coordinate this potent new human-AI collaboration to create the incredible things that will happen in the future will be the most valuable engineers of the future.

The is an opinion piece based on the author’s POV and does not necessarily reflect the views of HackerNoon.

Samsung

GitHub

The AI Writes Code, But the Engineer Creates Value

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Cloning Environments on AWS Beanstalk: A Practical Fix for Zero-Downtime Patching

HackerNoon is Upgrading Its AI Detector 🕵️

What MCP Means for Secure AI Integrations and the Future of Agentic Workflows

Cloning Environments on AWS Beanstalk: A Practical Fix for Zero-Downtime Patching

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

Darwin's Hybrid Intelligence to Align AI & Human Goals for Startups & VCs

Cloning Environments on AWS Beanstalk: A Practical Fix for Zero-Downtime Patching

HackerNoon is Upgrading Its AI Detector 🕵️

What MCP Means for Secure AI Integrations and the Future of Agentic Workflows

Cloning Environments on AWS Beanstalk: A Practical Fix for Zero-Downtime Patching

The Noonification: How Often Do NFTs Pass The Howey Test? (1/13/2023)

Darwin's Hybrid Intelligence to Align AI & Human Goals for Startups & VCs

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps