Skip to content
SLK ProjectsSLK Projects
Automation

How to Choose a Business Automation Partner in South Africa: A 2026 Buyer's Guide

Nine questions to ask before you sign, the red flags that predict a failed automation project, and what the South African market actually looks like in 2026.

Sam Luc Katshinda 6 min read
How to Choose a Business Automation Partner in South Africa: A 2026 Buyer's Guide

Automation projects rarely fail on the technology. They fail because the wrong thing got automated, nobody owned it after launch, or the supplier who built it was the only person who understood it.

This is the guide we would want if we were buying instead of building. Nine questions, the red flags that predict trouble, and an honest read of the South African market as it stands in 2026.

Nine questions to ask before you sign

1. What are you going to automate first, and why that?

A good partner will push back on your list. The right first automation is boring: high volume, clear rules, low judgement, and something that annoys someone every day. If the answer is whatever you asked for without argument, you are buying compliance rather than expertise.

2. What happens when it breaks at 4pm on a Friday?

Not if it breaks. Automations sit between systems you don't control, and those systems change. Ask who gets notified, how fast, and whether the failure is visible to you or only to them. "We monitor it" is not an answer. "You get an alert, and here is what it looks like" is.

3. Who owns the accounts and the code?

Your automations will run on cloud accounts, API keys and repositories. If those live in the supplier's name, you cannot leave. Insist the accounts are yours from day one and they get access, rather than the reverse. This single point causes more expensive divorces than any other.

4. Show me something you built that failed, and what you did about it.

Everyone has one. A supplier who cannot name a project that went wrong is either new or not being straight with you. What you're listening for is whether they diagnosed it or blamed the client.

5. How will I know it's working in three months?

Ask for the number before the build, not after. Hours saved, errors avoided, response time, revenue touched — whatever it is, agree how it will be measured and where you'll see it. Automation that nobody measures becomes automation nobody trusts, and untrusted automation gets worked around.

6. What is the total cost over two years, not two months?

Build cost is the smaller half. Ask for the monthly running cost — API fees, licences, hosting, support — and get it in rands with an explicit note of what's dollar-denominated. In South Africa that distinction matters: a tool priced in USD gets more expensive here without anyone changing anything.

7. Where does my data go, and does that satisfy POPIA?

If personal information leaves the country or passes through an AI model, you need to know which processor, under what terms, and whether you have a lawful basis. A partner who hasn't thought about POPIA has not thought about your risk, they've thought about your feature list.

8. What does handover look like if we part ways?

Ask for the actual artefacts: documentation, credentials, a runbook, an export of the configuration. Ask how long it would take. A confident supplier answers this comfortably, because they expect to be kept for the work rather than the hostage.

9. What would you tell me not to build?

The best answer to this question is specific and costs them money. Everything a supplier is willing to build for you is not the same as everything worth building.

The red flags that predict a failed project

A quote with no range and no assumptions. A single confident number this early means either padding you'll pay for, or a renegotiation in week three.

Discovery that never mentions your existing systems. Automation is mostly integration. If nobody has asked what you already run, the estimate is fiction.

Everything is possible. A supplier who never says "that will be painful" has not built enough to know where the pain is.

The demo is the product. Impressive demos are cheap now. Ask to see something running in production, ideally with a failure and its resolution.

One person knows everything. Charming, competent, and a single point of failure. Ask who the second person is.

No mention of what happens after launch. Projects that end at go-live tend to end shortly after go-live.

Pressure to sign this month. Urgency is a pricing multiplier for you and a closing tactic for them. Real constraints get explained; manufactured ones get repeated.

What the South African market actually looks like in 2026

The talent is genuinely good and the price is genuinely lower. Rand-denominated engineering against dollar-denominated tooling is the local arbitrage, and it is real. It also means the cheapest quote in the room is often cheap for reasons unrelated to skill.

Your running costs are exchange-rate exposed. Most of the platforms an automation depends on bill in dollars. Budget for that as a variable, not a constant, and ask suppliers to flag which parts of your monthly bill are rand-stable.

Resilience is not optional here. Assume interruptions — power, connectivity, a provider having a bad afternoon. Automations that only work when everything is up are automations that will need manual cleanup. Ask how state is preserved and what happens to the queue when something is unreachable.

POPIA is enforced, and AI made it more interesting. The moment a workflow sends customer data to a model hosted elsewhere, you have a processing question to answer. Most suppliers have not caught up. The ones who have will raise it before you do.

The market is crowded with resellers. A lot of "AI automation" is a thin layer over a no-code tool that you could have licensed directly for a fraction of the fee. That is not automatically wrong — paying someone to assemble and maintain it is legitimate — but you should know which you are buying, and what it would cost you to run it yourself.

The uncomfortable question

Before any of the nine, ask yourself one: is this process worth automating, or worth deleting?

A surprising amount of what businesses want automated exists because of a decision nobody has revisited in years. Automating it makes it permanent and faster. The cheapest project we can do for you is the one where we tell you to stop doing the thing entirely — and we would rather have that conversation than invoice you for the alternative.

If you want that conversation, tell us what's actually broken. If you'd rather see a number first, the quote tool gives you a range in about a minute, and we publish the arithmetic behind it.