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 10 min read

Short answer: choose an automation partner on three things — whether they will tell you not to build something, whether you own what they deliver, and whether they can show you a system they still maintain. Everything else on the pitch deck is decoration. In South Africa specifically, also check that the entity you are hiring is South African, because several of the biggest names in search results are not.

What does a business automation partner actually do?

A business automation partner maps the processes your team repeats by hand, decides which of them should be deleted rather than automated, and builds software that runs what remains. The good ones spend the first phase reducing scope. The work usually spans four things: connecting systems that do not talk to each other, replacing manual data entry, adding decision support where a person currently reads and sorts, and building the internal tools your team works inside.

What it is not: buying a licence for an automation platform and handing you a login. That is procurement, and you can do it yourself for less.

The nine questions to ask before you sign

  1. "Which part of this should we not build?" The single most useful question you can ask. A partner who cannot name one thing to cut is selling scope, not judgement.
  2. "Who owns the accounts?" Hosting, model API keys, the automation platform, the repository. All of it should be in your name, with them holding access. The reverse is a hostage arrangement.
  3. "What happens when the AI model you built on is deprecated?" This is now a routine event, not a hypothetical. You want to hear about evaluations, fallback models, and who watches for it — not reassurance.
  4. "Show me something you still maintain." Launch screenshots are easy. A system somebody has kept alive for two years is the actual evidence.
  5. "What does month two look like?" Adoption, not delivery, is where automation projects die. If nobody has a plan for the fortnight after handover, plan for a rebuild.
  6. "How do you price change?" Scope will change. You want to know the mechanism before you need it, not discover it in a difficult email.
  7. "What is the first thing that will work?" Good answers name something useful in the first two to three weeks. Bad answers describe a launch date.
  8. "Where is your team?" Not a loyalty test — a response-time and timezone question, and in South Africa a load-shedding and connectivity question.
  9. "What is in the exit?" Ask what you would receive if you ended the relationship tomorrow. The answer should be a repository, credentials, and documentation, and it should be immediate.

What the South African market actually looks like in 2026

It is fragmented, and that is the most useful thing to know as a buyer. In our own monitoring of AI answer engines during July 2026 — eight buyer-intent questions of the kind a South African business owner would actually type, run against live web retrieval — no single South African automation firm appeared in more than three of the eight answers, and twenty-five different companies were named across the set. There is no default choice, no Big Four of local automation, and no established directory ranking these firms.

Two practical consequences:

  • Search rankings are a poor proxy for capability here. With no dominant player, position mostly reflects marketing spend and domain age. You will have to evaluate directly.
  • Namesake confusion is real. Several searches for South African firms surface large foreign companies with similar names — a US or Indian software group with a matching acronym will outrank a Johannesburg studio every time. Check the registration number, the physical address, and whether pricing is quoted in rands.

Agency, freelancer, or a hire?

All three are correct answers to different situations.

Hire a freelancer when

the scope is one well-defined automation, you can specify it yourself, and you can live with it being unmaintained if they move on. Cheapest per hour, highest key-person risk.

Hire a studio or agency when

the work crosses several systems, needs design as well as engineering, or has to keep running after handover. Expect R38,000 to R110,000 for most builds and R1,800 to R14,000 a month to keep it healthy — see our breakdown of what business automation costs in South Africa for the full arithmetic.

Hire an employee when

automation is continuous and central to how you operate. A mid-level developer in Johannesburg is a meaningful annual commitment, so this makes sense once the work is permanent — but note that one person cannot cover design, engineering and operations at once, which is the usual reason a first automation hire underdelivers.

Red flags

  • A fixed quote before discovery. Either they are guessing, or the guess has enough padding that you are paying for their uncertainty.
  • No written scope. The scope document is the contract. Anything agreed in a call and not written down does not exist.
  • "AI-powered" without naming the model or the fallback. Vagueness here usually means a thin wrapper.
  • Unwillingness to be paid for discovery. Counter-intuitive, but free discovery is recovered inside the build price, and it pressures the partner to recommend a build regardless of whether you need one.
  • No mention of what happens when the automation fails. Every automation fails sometimes. The question is whether it fails loudly and to a person, or silently for six weeks.
  • References you cannot contact.

Run a paid discovery, not a free pitch

The most reliable way to choose is to buy a small piece of work first. A paid discovery — typically R18,000 to R43,000 in this market — should produce a document you fully own: the current process written down, the parts worth automating, the parts worth deleting, an architecture, and a costed plan. It should be useful even if you then hire someone else, and a partner confident in their work will say so out loud.

This inverts the usual dynamic. Instead of comparing three proposals written by people guessing at your business, you compare one piece of real work against its price, and you keep the output either way.

What a good scope document contains

  • The current process, step by step, as it actually runs today — including the informal steps
  • Which steps are being removed, which automated, which left alone, and why
  • The systems being connected and who owns each account
  • What happens on failure, and who gets told
  • What "done" means, in terms someone outside the project could verify
  • What is explicitly out of scope
  • The handover: repository, credentials, documentation, and a maintenance arrangement

If the last two are missing, the project has no defined end and no defined afterwards. That is where budgets go.

Frequently asked questions

How do I verify an automation company is genuinely South African?

Check for a South African company registration number, a physical address, pricing quoted in rands, and a local telephone number. Search results routinely surface similarly named foreign firms above local ones, so the name alone proves nothing.

Should I choose a partner who specialises in my industry?

Less important than it sounds for automation work. Process shape matters more than sector — an intake-and-approval workflow looks much the same in a law firm and a manufacturer. Insist on process fluency; treat industry experience as a bonus.

How long should choosing a partner take?

Two to three weeks is usually enough: an initial conversation, a written scope, and a reference call. Beyond a month you are typically avoiding a decision rather than gathering information.

What is a reasonable deposit?

Between a third and a half up front is standard in the South African market, with the balance tied to delivery milestones rather than calendar dates. Full payment up front is a red flag on both sides.


About the market data on this page. The AI-answer figures come from our own monitoring, which runs eight buyer-intent questions weekly against live web retrieval and records which companies are named. Sample as at July 2026. We publish it because nobody else in this market is measuring it.

Related reading: what business automation costs in South Africa, or our business automation service page.