Cape Town's AI and tech scene has grown fast. That is mostly good news — but it also means there are now a lot of agencies, consultants, and no-code resellers all calling themselves "AI companies." For a business owner trying to solve a real operational problem, knowing who to trust is genuinely difficult.

This article gives you six questions to ask any AI company before you commit. They are designed to separate vendors who will actually deliver something useful from those who will sell you a shiny presentation and a generic chatbot.

Why it matters more here than you think

The Cape Town tech market has some specific dynamics. A lot of the larger "AI agencies" are staffed primarily by sales and account management people who outsource the actual building offshore. Others are resellers for overseas SaaS platforms — their margin is the gap between what the platform costs and what they charge you. Neither of these is inherently wrong, but you should know what you are buying.

Real AI implementation for a South African business requires understanding local integrations (Everlytic, Yoco, MRI, PayProp, etc.), local regulations (POPIA compliance), and the specific communication channels SA customers actually use (WhatsApp above all else).

1Do you build, or do you resell?

Ask them directly: are you building custom systems, or are you a reseller for an existing platform? Neither answer is automatically wrong — but you need to know. A reseller will often be cheaper upfront but will lock you into a platform you cannot own or modify. A custom builder gives you more flexibility and ownership but will charge more for development time.

Follow-up: If I wanted to move to a different provider in two years, what would that involve?

2What integrations do you support natively?

Ask for a list of the tools they have actually integrated with before — not tools they claim to support. Most legitimate local providers will have experience with common South African business tools. If they have only ever connected to HubSpot and Slack, they are probably serving a different market than yours.

Specifically ask about: your CRM, your email platform, your accounting software, and WhatsApp Business API. If they go quiet on any of these, that is useful information.

3What does a typical implementation timeline look like?

Legitimate AI automation projects for SMEs take four to twelve weeks from scoping to live deployment, depending on complexity. If someone tells you they can fully automate your business in a week, they are either over-simplifying or selling you a template with your logo on it. If they say it will take six months for a straightforward chatbot, they are probably over-resourcing it to inflate the invoice.

4Can I see a live demo — not a video, a live system?

Any company that has actually built AI automation systems will be able to show you a working version of something similar to what you need. A video recording or a slide deck of screenshots is not a demo. Ask to interact with a live system — even a generic one. If they cannot produce this, they have probably not built anything recently.

5What happens after launch? Who owns the system?

This is where a lot of businesses get caught out. Some providers build systems that only work while you pay them a monthly retainer, and they retain control of the credentials and infrastructure. Make sure you understand:

A reputable provider will hand over documentation and credentials. A lock-in provider will avoid answering these questions directly.

6Do you have references from South African businesses specifically?

AI automation results in the UK or US do not necessarily translate directly to South African conditions. Ask for at least one South African business reference — ideally in a similar industry to yours — that you can speak to directly. If they only have international case studies or testimonials on their website that you cannot verify, be cautious.

Red flags to watch for in proposals

What a legitimate engagement looks like

A good AI implementation project follows a clear sequence: a scoping session where they learn your business and identify the specific process to automate → a proposal with defined deliverables and a fixed price → a build phase with regular check-ins → a testing period where your team uses the system before it goes live → a launch → ongoing maintenance with transparent reporting on usage and performance.

You should be able to see measurable results — time saved, leads captured, queries deflected — within 90 days of launch. If a provider cannot commit to any metrics at all, that is a problem.