9 min, 1,980 words
Explore five key trends in how AI is transforming professional software development… and how it isn’t. You’ll understand where it adds value, where it falls short, and how to apply it intelligently in commercial-grade projects.
- Asking the right questions about AI coding at scale
- AI code generation and where it’s appropriate
- The place for low-code and no-code platforms
- AI-powered integration, deployment, and testing
Across the tech world, it’s obvious. LLMs and other “artificial intelligence” solutions are a part of our processes. But in the enterprise software space where Modularis works, this is not actually anything new. We’ve been in the software automation and code generation space for 25 years. I want to look at a series of five trends and frequently appearing topics around AI in commercial software development, and how it is (and isn’t) transforming the industry.
ASKING THE RIGHT QUESTION
Founders and tech leaders like you want to be smart about what these tools mean in your world.
But these are simply new tools for existing challenges. Imagine you’re in construction instead, and suddenly someone hands you a power tool to replace your manual screwdriver. Of course, you should use it. But you also must thoroughly understand its capabilities and limitations, and you should not assume that it can do everything alone. You wouldn’t abandon fundamentally sound construction techniques.
So instead of asking “Should we use AI?” the smarter question is, “How do we use AI?” What’s important in commercial software development is to understand that what you’re doing is much more than “writing code.” You’re building something that will underpin everything your business does. So, as you look at AI trends and opportunities in professional software development, every technology decision should be driven by a core business value and not by what’s hottest or newest.
The five AI software engineering trends we’re going to discuss:
- AI-Driven Code Generation and Autonomous Programming Are On the Rise
- Low-Code and No-Code Platforms Are Democratizing Software Creation
- Generative AI is Commoditizing Software Development
- AI-Powered Continuous Integration and Deployment Is Enhancing Stability and Speed
- Automated Software Testing is Reducing Errors and Improving Quality
Let’s dive in.
1. AI-DRIVEN CODE GENERATION AND AUTONOMOUS PROGRAMMING ARE ON THE RISE
Generative AI-driven coding might seem like a sudden “hot topic,” but Modularis has worked in commercial-level code automation for 25 years. It’s not new. It’s just being used in different and more broadly accessible ways. This is the question that each of these AI software trends is really trying to ask: “Should coding be automated?”
There’s one answer for DIY situations. And another for you. It lies somewhere in the middle of two extremes.
The debate between fully hand-coding software vs. using fully automated code is a question of ratios. It doesn’t make sense anymore to have people hammer out every line of code by hand; there are tasks where AI is more efficient, and those hours are unnecessary. Basic infrastructure setup is a good example.
But, especially in the commercial software product space, core software logic must be handled by humans. There are a few solid reasons for this:
- Ownership. If an LLM can build your core software, is it solid and unique enough to build a business on?
- Responsibility. You are responsible for the success and security of this software build. If it comes out of an AI black box, there’s not enough accountability.
- Serviceability. Commercial software must be serviceable and scalable to be stable for your customers and your business.
In short, LLM-based tools like GitHub Copilot add value and save time, but are not substitutes for architecture or senior engineering expertise. Best practices in software development automation still matter. “Vibe coding” and “trusting the AI” without the appropriate ratio of real engineers are dangerous in commercial software. Think of it this way:
“Just because you can generate the code doesn’t mean you should.”
2. LOW-CODE AND NO-CODE PLATFORMS ARE DEMOCRATIZING SOFTWARE CREATION
Low-code and no-code platforms aren’t new, but they are expanding programming accessibility just like LLM-based coding models. And they can work in the right context. These platforms are suitable for internal, low-to-moderate complexity projects. I say, for small-scale or internal projects, go to town with low-code platforms! These tools effectively enable quick delivery for enterprise apps not intended for commercial resale or scalability.
Problems with low-code and no-code arise when these platforms are used beyond their intended scope. Again, commercial-grade products require stability, scalability, profitability, and serviceability that “vibe coding” cannot provide. The democratization of coding is real, but limited by the complexity and commercial demands of the software being built.
Remember, you’re not really “writing software,” you’re building a product. Would you rest your business on a foundation cobbled together by best guesses? Would you drive a car designed by ChatGPT?
3. GENERATIVE AI IS COMMODITIZING SOFTWARE DEVELOPMENT
Generative AI tools have an incredibly low barrier to entry. They’re available to anyone, any time, anywhere, and promise to write usable code to help solve any problem. That’s why “vibe coding” is such a prevalent term right now. If you can think it… AI can code it.
Right?
For DIY, sure. But there is a large expectation gap between what Gen-AI coding promises and what it delivers, and that matters at a commercial scale. By its nature, this code isn’t proprietary or built for your exact purposes. Raw AI output lacks uniqueness, ownership, and business defensibility. Personalized, structured use of generated code is the way this technology fits into commercial software development.
The Spray Foam Analogy
Picture spray foam insulation. If you take a can of spray foam and press the trigger to dump out its contents, what do you get? A big pile of goop, taking up space. It’s difficult to remove. It hasn’t contributed meaningfully to your project, and it’s actually adding negative value. But… if you need to fill a gap in a structure and you’ve built the constraints around it, spray foam is incredibly useful.
This is how we think of AI in software development. The LLM-driven commoditization of code is real, meaning it only becomes meaningful and marketable in the hands of experts and when constrained by solid structure.
4. AI-POWERED CONTINUOUS INTEGRATION AND DEPLOYMENT IS ENHANCING STABILITY AND SPEED
Continuous Integration and Continuous Deployment (CICD) is a natural fit for AI automation and one of the areas where I consider AI to be a powerful enhancement. Say you’re practicing incremental software modernization: we can stand up CICD pipelines and validate them with less effort using AI tools, and significantly reduce the manual “housekeeping” of these ongoing tasks.
Once again, however, this automation doesn’t absolve anyone on the team of the fundamental accountability for the quality of each deliverable. So even in something like CICD, where you can significantly increase the ratio of AI to human work, you cannot hand it off to the machine and hope for magic. Human/machine collaboration is essential to avoid over-automation that risks failure and rework. You don’t need a massive team anymore, but you do need an expert team.
5. AUTOMATED SOFTWARE TESTING IS REDUCING ERRORS AND IMPROVING QUALITY
Across enterprise companies and heavy engineering companies like ours, the trend is towards a strong AI presence in the QA process. These models are an excellent addition when used correctly.
Generative AI software helps:
- Filter information and results at a scale that humans can’t handle
- Analyze huge amounts of data to find issues, trends, and red flags
- Add ongoing intelligence to products that help owners improve usability and quality
But, once again, automated, AI-driven software testing isn’t something that can run independently. It must always be backed by a deterministic architecture and expert-driven design process that ensures reliability and accountability for the owner. If you don’t understand how issues are found or how your QA is run, how can you claim that you know what’s happening in your software? Automated quality assurance should be used to increase test coverage and speed while preserving accountability and engineering oversight.
Once again, use the power tool… but use your unique business intelligence to build a structure that will stand the test of time.
QUESTIONS TO ASK YOURSELF ABOUT AI IN SOFTWARE DEVELOPMENT
At the end of the day, AI in commercial-grade software development will continue to grow as a tool, as a resource, and as a hot-button issue. Each of the five trends I’ve talked about today is only going to get more complex as the tools grow. So I challenge anyone investing in AI software design for their company to ask themselves a couple of key questions.
Is it “yours” enough?
If an LLM can handle your software task, is it really unique enough to underpin your entire project? At the enterprise level, anything that can be “vibe coded” can be “vibe coded” by a competitor, too.
Are you really just writing software, or are you building a product?
If you answered “building,” make sure that your decisions around AI technology reflect your core business values and deliver reliability and profitability for which you are responsible.
After answering these questions, consider reaching out to Modularis. We’ll help you leverage automation intelligently to extend your R&D dollar in ways that are fast, safe, and effective. Contact me here to start the conversation about how AI fits into your project vision and how the Modularis team can help you build ownable, reliable, enterprise-level products.