AI Didn’t Supercharge Low-Code—It Made It Redundant
AI-driven low-code platform replacement is the shift from visual drag-and-drop builders aimed at non-developers toward AI agents that author clean, explicit source code which enterprises can run, audit, and govern like any other production software, making traditional low-code and no-code abstractions increasingly obsolete and less competitive.
Low-code and no-code tools were supposed to free non-developers from source code; AI has flipped that logic on its head. Large language models now write readable HTML, workflow code, and glue logic on demand, making the visual layer feel more like friction than empowerment. Tines’ CEO says it plainly: “There is a sell-by date on low-code, no-code as a category.” The category hasn’t failed; it has been outpaced. When AI can generate production-ready code, the hard problem moves from assembling colorful blocks to governing what the machines write. That’s why the most interesting activity in automation is no longer happening inside visual builders, but in platforms that treat code as the first-class output of AI—and treat runtime as something that must stay deterministic, testable, and under IT’s control.
Shopify’s Back-to-Code Moment Shows Why Visual Abstractions Break
One of the clearest signs of no-code obsolescence is unfolding in storefront customization. Shopify is rolling out a new base theme that returns to mostly HTML plus its Liquid templating language, in a structure that “anyone with even a smidgen of HTML knowledge” can understand. The new theme has 93% fewer lines of code than the JSON-heavy Horizon template it is set to replace, turning page structure back into plain text instead of serialized configuration blobs. This is not nostalgia; it is AI pragmatism. Shopify’s Sidekick assistant has already driven 25 million theme edits this year, used by 20% of merchants. To serve those agents, the company found that “agents want the same things as humans: code that is easy to read, explicit contracts, and constructive feedback.” When the AI wants clean code, the JSON gymnastics that visual editors introduced stop looking like democratization and start looking like noise.
In chasing flexibility, Shopify once converted template instructions into sprawling JSON so every knob and button could be adjusted in a visual editor. That power came with a cost: “you could no longer understand a page by reading one file.” Visual abstraction quietly turned into opacity. By reorganizing themes so both humans and AI can read them, Shopify is acknowledging what many enterprises are now discovering: the more AI participates in daily development, the more valuable explicit, human-readable source becomes. The company is even adding 20 new rules to constrain complexity—governing contracts, nesting, structure, and file sizes—so generated themes remain predictable and auditable. That looks a lot less like low-code magic and a lot more like old-fashioned software engineering, optimized for agents and people together.
Tines 3B: AI Writes the Workflow, IT Owns the Runtime
Where Shopify is simplifying templates, Tines is aggressively turning its back on its own no-code origins. After eight years of building a no-code automation platform, the company has launched 3B, which uses AI to author enterprise workflows while executing them as conventional code. In 2023, it raised a USD 125 million (approx. RM575 million) Series C at a USD 1.125 billion (approx. RM5.18 billion) valuation, and its original system now runs 1.5 billion automated actions per week—yet the team still decided “no code, low code… was clearly not the future.” With 3B, an employee describes an application, agent, or workflow in natural language; the platform writes the code, deploys it, and lets IT and security teams stay in control of credentials, integrations, and permissions. According to CEO Eoin Hinchy, “Building is now commoditized,” so the value shifts from click-together UIs to managing AI-generated code as a first-class asset.
The crucial design choice is that AI is involved only at authoring time, not at every execution step. “We’re using AI in the creation of these workflows, but that’s it.” Once generated, each step becomes code that runs in its own ephemeral Docker container, destroyed and recreated on each run. The result is deterministic AI workflows: the system does not ask a model to reason about every step at runtime, which makes common flows faster, cheaper, and easier to test. Admins see dashboards of which workflows are running or failing, which touch the public internet, and which depend on particular systems, with suggestions for fixes they can approve. Employees in finance or HR who “don’t know anything about Git” can still describe what they need, while IT pre-defines what systems and data each workspace may reach via access maps, with isolated execution, credential protection, and full logging for every flow.
From Hiding Complexity to Exposing Clean, Governed Code
Both examples tell the same story: the low-code platform replacement underway is not about better drag-and-drop. It is about giving up on the idea that complexity should be hidden from users behind a visual interface. Shopify’s re-architecture shows the benefits of code that is “simpler for both people and machines to digest and hack against,” backed by a forever API and guardrails for contracts, structure, and complexity. Tines’ 3B shows how AI code generation enterprise platforms can let anyone describe work in natural language while enforcing deterministic execution under IT governance. In this world, code is no longer the scarce, scary resource reserved for engineers. AI turns clean, production-grade code into the default interface. What now matters is whether that code is explicit, testable, and auditable.
Enterprise IT teams have learned the hard way that opaque visual canvases age poorly. When workflows are critical, they want code-level control and observability, not a maze of UI blocks no one wants to reverse-engineer during an incident. Platforms that cling to drag-and-drop metaphors will struggle to compete against systems where AI writes human-readable code that runs in isolated containers, with clear access maps, and complete logging. The post–low-code era will not be defined by more AI magic in the UI, but by more disciplined software practices underneath: deterministic AI workflows, explicit contracts, and simple text files that humans and agents can both read. Low-code had its moment. The next one belongs to AI that makes code, not clicks, the accessible path for everyone.
