We recently had a couple different clients bring up the concept of “prompt engineers,” speculating on a future where engineering would be less about the creation of the code itself and more about the process of creating and curating AI-generated prompts.
The framing about what these prompt engineers might theoretically do felt remarkably similar to the stories and promises we’ve heard about the no-code/low-code space, which was interesting to me.
I can’t count how many developers and engineers over the years I’ve chatted with who actively distrust the outputs of no-code platforms, despite claims that it will speed and ‘democratize’ development. It’s notable that the idea of “building a website without ever needing to write source code” has taken the world by storm when it’s being done with AI/ML but is something we collectively tend to discount when it’s done with low-code.
So what’s the difference, and what does the introduction of LLMs mean for low-code platforms?
The primary justification for low-code platforms is allowing teams to move quickly from idea-to-deployment with no code input required.
Given what we’ve witnessed in the last few months, it’s inarguable that AI-code generation is going to have some impact on this space. The purported and demonstrated abilities of AI to generate code (in some cases from very abstract descriptions), create scaffolding, help brainstorm, test and prototype, and otherwise speed up the development process is well documented. See my colleague James Governor’s post on AI’s impact on the “revolution of idea-to-code.”.
LLM vs. low-code is not an apples-to-apples comparison. The primary difference between LLMs and low-code platforms is the output. Generally, when you tell an LLM to generate a website, it spits out actual code in an actual language that will run anywhere. When you tell a low-code platform that, it either won’t (black box) or it spits out incomprehensible and/or proprietary code that, often enough, can only be run on a single proprietary platform.
There are still plenty of use cases where low-code will still be the right choice. If you’re an organization that doesn’t want anything to do with infrastructure and don’t care about the underlying platform, then low-code might still be the best bet.
If you care about portability at all LLMs will probably be an appealing alternative.
LLMs won’t take the whole market, but it seems likely they’ll take some of the growth out of the market. Existing low-code customers and their apps are unlikely to migrate, but low-code platform growth is going to become more difficult to come by. Low-code providers have reason to be concerned about their future growth prospects.
No Comments