The approval form went live in an afternoon, built in a no-code tool by the person who runs the process. Eighteen months later, it needs a second approver for orders over a threshold, read access for finance only, and a record of who approved what. The tool has no setting for any of the three.
That is the decision underneath "low-code or no-code". Both get a first version out quickly. What separates them is the cost of the second and third year. By then the requirements have grown, and the person who built the tool may have moved on.
The two options, defined
No-code is building software entirely through visual editors and configuration panels, with no code at any stage. Zapier, Airtable, and Bubble are common examples. Each covers different ground: automation, structured data, and web apps.
Low-code is building software with visual tools that let you write code wherever the tools stop. Platforms such as Retool connect to databases and APIs, and let developers add queries, logic, and custom components. What is low-code covers the category in more depth.
The practical difference is what happens at the edge. When a no-code platform does not support something, there is no code layer to add it. On a low-code platform, a developer writes the missing piece.
How no-code looks to the buyer
No-code has the lowest cost to start. An operations manager or analyst can build a form, a status tracker, or a notification without filing an engineering ticket. Licensing is usually simple, and the first version can be live in hours.
That speed is real, and for some work it's the right answer. For a team still running a process in spreadsheets and email, a no-code form is a meaningful step toward structure.
The limits show up as the work gets more complex, regulated, or high-stakes. No-code platforms typically can't handle:
A custom data model beyond the platform's own structure.
Logic with more than a few conditional branches.
Fine-grained permissions by role.
Integrations outside the platform's connector library.
Audit logging to the standard financial services, healthcare, or legal work expects.
When a workflow outgrows the platform, the usual path is a rebuild somewhere else. The rebuild often costs more than the original, and it lands when people already depend on the tool.
The other cost is accumulation. One no-code tool solves one problem cleanly. Twelve no-code tools, adopted by twelve teams, leave data in many places and permissions that differ by tool. Nobody is sure what connects to what. That is how tool sprawl starts. Tools adopted without IT's knowledge also become shadow IT.
How low-code looks to the buyer
Low-code costs more to start. There are platform licenses, and some work needs a developer or a technical builder. The first version often takes days, rather than hours.
In return, the tool can grow without a rebuild. A developer can add logic the visual tools don't cover, connect any API or database, and set permissions per role. Low-code also keeps internal tools out of the engineering team's main queue for most changes, while letting developers step in when a change needs them.
Retool's new app builder takes this further. Its apps are React code, written in the builder's own chat and code views or by a coding agent such as Claude Code. They deploy to Retool. Retool supplies the hosting, version history, source control, permissions on each data connection, SSO, and audit logs. Those controls are what an IT or security review asks about, and they're already in place.
What the two years actually cost
The upfront comparison favors no-code. The two-year comparison depends on three questions:
Will the requirements grow? If the tool will gain users, roles, approval steps, or integrations, plan for the platform that can absorb them without a rebuild.
Will it face a review? The tool may have to pass a SOC 2 review, an internal audit, or a regulator's question about who changed what. If so, it needs role-based permissions and audit logs from the start.
Who will maintain it? If the person who built it leaves, someone else has to understand and change it. Code in source control, with version history, is easier to hand over than a chain of configuration screens.
The ROI of low-code covers the numbers to gather before committing.
Which fits which
No-code fits a simple workflow that is unlikely to change and has light security needs. It also fits a prototype that tests an idea before anyone commits engineering time.
Low-code fits a tool that will spread across teams, connect to the systems where the data lives, or face a security review. The same goes for any process expected to change shape within two years.
Our view
We recommend low-code for any internal tool a team will still be running in two years. The reason: requirements tend to grow, and a rebuild usually costs more than the platform. No-code stays the better choice for prototypes and simple, stable workflows, where it's faster and cheaper, and nothing depends on it.
When a no-code tool has outgrown its platform, the person who built it is usually the best product owner for the low-code version. They know the process and the edge cases. We bring the architecture, the integrations, and the running of it, and they keep the final say. Low-code or traditional development covers the next decision up, if the choice is between a platform and a custom build.
