Forget low-code?
Here is what South African businesses actually need to know.
BusinessTech ran a sponsored post on September 11 saying AI has rendered low-code development "virtually obsolete." The sponsor was a development company selling custom software. Here is the independent version — without a sales pitch — covering when low-code still makes sense, when AI-augmented custom development is the right call, what each option actually costs, and what nobody is telling South African businesses about either.
When a development company pays to publish an article saying "forget the cheaper alternative, hire us instead," that is not analysis — it is advertising. BusinessTech's partner content label is there, and to their credit it is visible. But the argument deserves an honest response because the underlying question — what should South African businesses actually use to build software in 2026 — is genuinely important and genuinely complicated.
The sponsored post got one thing right: AI has fundamentally changed the cost and speed of software development. A competent developer using Claude or Cursor today ships code significantly faster than they did two years ago. The gap between what a small development team can produce with AI assistance and what a large team produced without it has collapsed in ways that are still being absorbed by the industry.
What it got wrong — or rather, left out: Low-code is not a single thing competing against custom development. It is a spectrum of tools that solve different problems at different cost and complexity points. Collapsing all of them into "obsolete" to make the case for hiring a custom development firm is like saying "forget buses, hire a driver" — true in some situations, wildly inappropriate in others.
What low-code actually is — and what it is not
// Chapter 01 — Defining the landscape honestlyLow-code platforms like Microsoft Power Apps, Retool, Bubble, and Zoho Creator let you build applications visually — dragging, dropping, connecting to data sources — with limited manual coding. They are fast to start, accessible to non-developers, and good for internal tools, simple workflows, and prototypes.
No-code platforms like Webflow, Zapier, and Glide let you build entirely without writing code. They are excellent for websites, automation workflows, and lightweight apps where the platform's constraints match your use case.
Vibe coding — directing an AI like Claude to write code for you — sits between low-code and custom development. You are producing real code, you own it, and you can do anything code can do. But the quality, security, and maintainability of that code depends entirely on how well you can direct the AI and review what it produces.
Custom AI-augmented development — what Codehesion is selling — is a professional development team using AI tools to work faster. The output is production-grade custom software. The cost is professional development rates, plus the AI tooling. This is appropriate for complex, high-stakes systems.
Gartner projects that 75% of new enterprise applications will be built using low-code technologies by 2026. That is not a prediction that low-code is obsolete — it is a prediction that it is dominant. The claim that AI has rendered it obsolete appeared in an article published on the same platform three days after Gartner's figure was widely reported.
When low-code is genuinely the right answer
// Chapter 02 — The use cases that belong thereLow-code wins when your primary constraint is time or internal capacity, not customisation. The best-fit approach is to use no-code for front-end workflows and internal tools, and low-code for backend logic, scalability, and complex integrations. Specifically:
| Use case | Low-code / No-code | Vibe coding | Custom development |
|---|---|---|---|
| Internal HR or ops form | Best fit — Power Apps or Zoho Creator. Fast, connected to existing Microsoft stack. | Possible — worth it if you want to own the result. | Overkill — cost exceeds value. |
| Marketing website | Best fit — Webflow or a well-built static site. Fast, maintainable by non-developers. | Good fit — full ownership, no monthly platform fee. | Overkill unless brand requirements demand it. |
| Client portal with document access | Possible — but security rules are limited, vendor lock-in is real, per-user costs escalate. | Good fit — if you understand what you are directing the AI to build. | Best fit — for complex access rules and sensitive data. |
| SaaS product you will sell | Risk — vendor lock-in, limited white-labelling, platform dependency is a business risk. | Possible for MVP — validate before investing in full custom build. | Best fit — once validated, you need full ownership. |
| Workflow automation | Best fit — Zapier, Make, or n8n. Purpose-built for this. Faster than any custom solution. | Possible — but why? The tools are better. | Unnecessary — unless deeply integrated with custom systems. |
| Payment processing system | Avoid — security constraints, limited PCI compliance control. | Possible — with careful security review of every line. | Best fit — financial liability requires professional oversight. |
| Compliance document system | Risky — POPIA obligations require audit trails and deletion capabilities that most platforms handle poorly. | Possible — if the developer understands POPIA requirements. | Best fit — compliance by architecture, not workaround. |
The total cost of ownership — what the sponsored post did not mention
// Chapter 03 — The numbers that matterThe argument against low-code always sounds compelling until you price out the alternative. Here is an honest cost comparison for a mid-size South African business building a client portal for 200 clients with document management and automated workflows:
- R1,200 per user per month
- 8 staff users = R9,600/month
- R115,200/year ongoing
- No setup cost but no ownership
- Vendor price increases hit immediately
- Platform shuts down — you start over
- Limited security rule customisation
- Fast to start — days not months
- R45,000–R65,000 build cost
- R800–R2,500/month GCP hosting
- Pays back vs Retool in 6 months
- Full ownership — no vendor risk
- Security rules you control entirely
- POPIA compliance by architecture
- Unlimited users — no per-seat cost
- 6–10 week build timeline
- R150,000–R400,000+ build cost
- R5,000–R15,000/month maintenance
- Full ownership
- Production-grade security
- Long build timelines — 3 to 6 months
- Expensive for most SA businesses
- Right for large enterprise or regulated industry
- AI-augmented teams now build faster
The custom build pays back against Retool in about six months and then runs at R2,000 per month versus R9,600 per month forever. Over three years, the custom build on GCP costs approximately R127,000. The same three years on Retool costs R345,600 — and you still do not own anything.
This is not an argument against Retool. For a team that needs something running this week, Retool is the right answer. The argument is against the framing that one approach is universally correct — which is what a sponsored post from a custom development company has every incentive to claim.
What AI-augmented development actually means in practice
// Chapter 04 — The honest versionBy June 2026, the convergence of artificial intelligence with low-code platforms has fundamentally transformed how enterprises build, deploy, and maintain software applications. Forrester's Q2 2026 landscape report identifies AI-native development as the defining characteristic separating market leaders from legacy vendors.
What this means in practice: a developer using Claude or Cursor in 2026 produces roughly the same output in one day that took a week two years ago. That acceleration applies whether they are building on a low-code platform, writing custom code, or directing AI to write it. The tools made everyone faster. They did not make one category of tool obsolete.
Every employee becomes a builder — AI will let business users create bespoke applications without writing code, while professional developers guide standards, security, and integration. The line between "developer" and "user" will blur as agentic systems become part of daily work. That is not a prediction that low-code is dead. It is a prediction that the distinction between low-code and custom development will blur — because AI is democratising custom development the same way low-code did, but further.
The South African market sits in a particularly interesting position here. South African organisations are under intense pressure to digitise faster, cut costs, and improve customer experience — all while dealing with skills shortages and tight IT budgets. Low-code automation is making it easier than ever to turn automation ideas into production-ready systems. Telling those organisations to "forget low-code" is advice that serves agencies, not businesses.
The Forge Vertical position — because transparency matters
// Chapter 05 — Where we sit in thisForge Vertical builds custom production infrastructure on GCP. That means we have a commercial interest in you choosing custom development over low-code for certain use cases. You should factor that in.
We also use Zapier for client workflow automation when it is the right tool. We recommend Webflow to clients who need a marketing site and have no technical team. We tell people to use Make rather than building a custom integration when Make does the job. The goal is that you build the right thing, not that you hire us specifically.
What we build well: client portals, document management systems, admin dashboards, automated business workflows, payment integrations, and anything that needs real security rules and POPIA compliance by architecture rather than workaround. If what you need is a form connected to a spreadsheet, use Google Forms. If what you need is a client-facing portal that handles sensitive documents, tracks job status, sends automated notifications, and gives your team a full audit trail — that is what we build.