Why AI tools fail South African SMEs — and it is not the tools

South African SMEs are buying AI tools at a growing rate. But the returns are not following. Research published in the International Journal of Research in Business and Social Science (2024) found that AI adoption by South African SMEs remains low in effectiveness — not because the tools are bad, but because the framework for adoption is missing. The tools get bought. The integration never gets built.

We see this pattern in almost every diagnostic we run. A business has a CRM. It also has an email marketing platform. And a scheduling tool. And a spreadsheet that one staff member owns. None of them talk to each other. Every lead that comes in gets manually entered in three separate places. Every follow-up gets sent by hand. Every report gets assembled on a Friday afternoon by someone who should be doing something else.

The investment paradox

A widely cited figure in the South African market suggests that roughly 73% of local businesses have invested in some form of AI or automation tooling — but fewer than half report seeing a meaningful return. The gap is not a tool quality problem. It is an integration problem. The tools were purchased as standalone solutions to single problems. No one designed the architecture that would let them work together.

Why it happens

Most AI tool purchases in SMEs are reactive. A staff member sees a demo, or a competitor mentions they use something, or a problem gets bad enough that someone buys a solution. The purchase decision is made in isolation — without mapping how the new tool connects to everything else already in the stack.

Researchers at the University of Johannesburg (Barnes, Sachs & Pelser, 2023) studied digital technology adoption in South African manufacturing SMEs and found the primary constraint was not cost or technical skill — it was the absence of a framework for adoption. Businesses were acquiring tools without a model for how those tools would change the workflow around them.

The three failure modes we see most often

• Siloed implementation — each department buys its own tool, no one owns the connective layer, data never flows between systems.
• No baseline measurement — the business does not know how long the manual process currently takes, so it cannot measure whether the new tool saved any time.
• No handover plan — the tool gets built by an outside consultant or a tech-savvy staff member, nobody else learns how to run it, and when that person leaves the system dies.

The fix is not another tool

The answer to a broken integration is not buying a better CRM or a smarter email platform. The answer is mapping what you already have, identifying where data stops flowing, and building the connective tissue. In most cases, the tools businesses already own are adequate. What is missing is the architecture that makes them work together.

A well-designed integration means your CRM receives new leads automatically from your website form. It means your email platform knows a lead has been marked as a client and stops sending prospect nurture sequences. It means your scheduling tool creates a calendar block and notifies the right team member without anyone touching a keyboard.

"Most businesses don't have an AI problem. They have an integration problem. That's exactly what we fix."

— Jimmy Motsei, Founder, Maru Online

What to do before you buy anything else

Before any new tool purchase, map what you already have. List every software platform your business uses. For each one, ask: where does data enter this system? Where does it leave? Who moves it manually? The answers will show you where the integration gaps are — and which ones cost the most in team time.

That is exactly what a Maru diagnostic does. It maps your existing tools, quantifies the cost of the gaps in hours and rand, and prioritises the integrations that will save the most. Before anything gets built, you have a clear picture of what is actually costing you.