Part one of this article explores why breakthroughs in no-code development, powered by generative AI and modern data-sharing technology, will revolutionize software development as we know it today.
This article was originally published on Medium.
Remember that Holiday Inn Express commercial a few years back where an operating room nurse asks the man acting as an emergency room doctor if he’s really a surgeon? His response: “No, but I did stay at a Holiday Inn Express last night.” Cue the catchy slogan about not needing to be smarter to feel smarter.
No-code offerings are a lot like staying at a Holiday Inn Express: sure, they might make you feel like you can tackle anything, but the reality is that learning how to be a developer and create (and test and maintain and monitor and support and document) business applications is hard.
Similar to the years of medical training and education required of a surgeon, becoming a software developer typically requires four years of college, possibly a master’s degree, and at least five or more years of on-the-job experience to build expertise and skills.
Even so, it’s not hard to see why the concept of citizen developers, empowered by no-code tools, holds such enduring appeal. Not only are there far more information workers—the actual users of business applications—than there are application developers, but these folks know exactly what the application needs to do because, well, they are its target audience.
If you could somehow unlock that potential and do away with developers in the loop, it would be a huge business opportunity: better business and user experience focus, more timely improvements with fewer delays, improved end-user outcomes, and the list goes on and on. Imagine 10x-ing your IT department overnight for no additional cost or hiring. So, what’s holding us back? Why can’t we do this already?
Let’s take a look at how no-code platforms work today, how GenAI removes the existing barriers, and why it will fundamentally change the way no-code tools are developed.
The curious paradox of “no-code” tools
Designed to empower anyone in the company with a business need—so-called “citizen developers”—no-code platforms were supposed to make application development easy enough to produce IT solutions without technical support…and without accidentally putting the business and its data at risk.
Unlike a “real” developer, whose code can do anything, good or bad, no-code platforms have to make sure that someone who doesn’t know anything about software development, data storage, or IT infrastructure can’t do serious damage. It’s sort of the moral equivalent of “safety scissors” for kindergarteners. No-code platforms try to make inherently dangerous actions, such as exposing the company’s data to outsiders or crashing mission-critical systems, safe in one of two ways:
- Limiting expressiveness: Restricting what citizen developers can ask for in the first place
- Adding guardrails: Using techniques that keep citizen developers on the straight and narrow (usually with a big “DANGER” sign should you choose to off road)
To give a non-technical example of how this works, consider traffic signals at an intersection. Sensors in the road detect when drivers are at a light. Pedestrians can manually press a “walk” button to indicate they want to cross. Lines on the street (and maybe even literal guardrails) tell both cars and pedestrians where they should and shouldn’t go.
This approach works well under normal conditions but it doesn’t allow for self-service handling of more complicated scenarios such as road construction, parades, or unusual traffic flow. For those problems, you’d still need a traffic cop (the equivalent of calling on a real developer when your no-code tools fail to deliver).
No-code platforms and tools work in similar ways. One of the easiest ways to make something safe is to simply remove capabilities—if you can’t do something dangerous, you can’t hurt yourself (or the company). By the same token, this approach also then limits capabilities.
A real-world example is popular SaaS-based recruiting platform Greenhouse with their simplified job page hosting. Anyone can set it up; no computer science degree is required. But it’s limited to selecting background colors, adding a company logo, and changing the job description text. To go further, which includes hosting on a custom domain and creating company-specific workflows, you’d have to give up on the no-code solution and turn the problem over to your IT development team.
Guardrails in no-code development often take the form of “drag and drop” components—if you stick to the existing palette, you won’t go too far astray. But often there’s some kind of “break the glass” option as well. If you pick that, then you’ve gone off road and all bets are off.
Intentionally going around the guardrails in a no-code platform can be as dangerous as driving your car around real guardrails on a steep road: A single mistake can expose company data or bring down a mission-critical system. Any problem involving security, privacy, uptime, accessibility, or compliance exposes the entire business to massive financial risk…risk that a citizen developer probably isn’t well equipped to evaluate or address without being trained in software development, SOC2 compliance, personally identifiable information (PII) handling, infosec regulations, and so forth.
Add in other “emergent” application challenges like performance, uptime, cross-geo resilience, multi-party data sharing, and so forth and we know have the perfect storm: Problems of unbounded complexity with unlimited potential for financial harm being pursued by citizen developers with almost no training or insight into the underlying infrastructure mechanisms. Yikes!
At least, this was the state of the no-code world up until 2023. Then, GenAI arrived and changed the rules entirely.The fundamental shift in no-code development: A GenAI expert on every desk
Think back to our traffic intersection example. What if you could put an experienced traffic cop in every car, watching every intersection, chaperoning every pedestrian? Of course, that’s not practical on the streets or in company IT settings if the cops have to be real people. But what if they could be machines instead?
No-code platforms have historically been unable to recreate the years of hard-won learning, professional training, experience, and acumen bottled up inside a human brain. Prior to GenAI, attempts to distill that knowledge into no-code platforms were largely unsuccessful because they had to be done through guardrails and limiting expressiveness…otherwise, you’re just asking a “citizen” developer to act like a real developer.
GenAI approaches things in a completely different way. Instead of a limiting, made-for-purpose no-code framework, GenAI can be automatically (and repeatedly) trained on a vast, domain-specific knowledge base. GenAI emulates what a great professional developer does: it reads all the good articles, catches up on all the explanations on popular developer ecosystem sites like StackOverflow, studies all the computer science textbooks and research papers…in fact, it can read and learn and comprehend and remember far more than any one human being is capable of and it can update that knowledge with every passing day to stay current (for instance, on zero-day exploit attacks and other fast-moving security concerns).
This is a serious game changer, and something no no-code platform on the planet has ever been able to pull off before!
On top of all this, GenAI fundamentally changes how no-code applications are created. Historically, the medium for a no-code development platform was a complicated, messy user interface (UI) designed to let non-coders drag and drop their way to an application. Since the 1980s, visual composition via UI has been touted as a way to let People Who Don’t Know What They’re Doing try to accomplish complex outcomes nonetheless.
Unfortunately, it’s a structurally unsound argument: Creating applications as if they were paint-by-numbers coloring books isn’t any more successful for software development than it would be for traffic management, medicine, or any other complex professional task.
GenAI creates code differently. Instead of a limited one-size-fits-all framework, GenAI works like a real developer: It can write any program, guided by its knowledge and industry-wide design pattern training.
Large language models (LLM) and modern machine learning combined are far more powerful an approach to “no code application development” than any drag-and-drop UI could possibly hope to be. They are as flexible and powerful as any developer’s brain while still offering security, performance, and other important outcomes in what they produce.
Like a real developer, they can be prompted to pay attention to important aspects like compliance, performance, and testability. And they can work from a natural language description rather than an artificial and constrained UI ‘picture’ of the code, allowing the business owner to express what needs to be built rather than doing a paint-by-numbers exercise.
That said, the potential to turn any information worker with basic Microsoft Excel proficiency into a full-fledged application developer turning out SOC2-compliant, privacy preserving, highly performance business applications isn’t quite here—yet.
Part two of this article outlines what progress is needed before we can fuel the coming generation of citizen developers.
How can your organization safely embark on this new GenAI frontier? Check out our eBook, “Unlocking the promise of generative AI,” for actionable measures you can take to keep sensitive data always secure.