Which AI tools should a South African SME use?
Most South African SMEs should start with the AI assistant built into the productivity suite they already pay for: Microsoft 365 Copilot on Microsoft 365, or Gemini on Google Workspace. Those assistants work inside the email, documents and meetings your staff already use, they respect the permissions you have already set, and their business terms commit not to train on your data.
A general-purpose assistant such as ChatGPT or Claude earns a place next to them when a team does a lot of drafting, research, analysis or coding outside Office or Workspace. Many businesses end up running two: the suite assistant for everyone, and a general assistant for the few people who use it hardest.
| Tool | Where it works | How it is priced | Best for |
|---|---|---|---|
| Microsoft 365 Copilot Chat | Browser, Teams and the Microsoft 365 app | Included with Microsoft 365 business plans | Safe everyday chat for every staff member, with enterprise data protection |
| Microsoft 365 Copilot Business | Outlook, Teams, Word, Excel and PowerPoint, using your own mail and files | About R400 per user per month, VAT included | Microsoft 365 businesses of up to 300 users who live in Outlook and Teams |
| Gemini in Google Workspace | Gmail, Docs, Sheets, Meet and the Gemini app | Included in Google Workspace business plans, with more features on higher tiers | Teams that already run on Google Workspace |
| ChatGPT Business | Browser, desktop and mobile apps | Per user per month, billed in US dollars | Drafting, research and data analysis outside Office or Workspace |
| Claude Team | Browser, desktop and mobile apps | Per user per month, billed in US dollars | Long documents, careful analysis, writing and code |
Rand prices for the Microsoft licences, including Copilot, are on the Microsoft 365 price list, which is re-checked against Microsoft's list prices and the exchange rate. The AI adoption page compares the assistants feature by feature.
Is it safe to use AI tools with company data?
It is safe on business tiers and not on free or personal accounts. Business and enterprise tiers of Copilot, Gemini, ChatGPT and Claude commit in writing not to train on your data and give administrators control over who uses what.
Consumer tiers generally reserve the right to keep and learn from what is typed into them, and they sit outside your admin console, so you cannot see or stop what leaves the business. Under POPIA, pasting a client's personal information into a personal AI account is a disclosure the business is responsible for. The guide to using AI without leaking business data covers what staff should and should not paste.
What changes when AI assistants become agents?
The change is from advice to action: an assistant answers a question, while an agent is given a goal, an account and permission, and then does the job. That one shift rewrites most of the risk model a business already has for AI.
- Data exposure becomes structural. A chatbot only sees what someone pastes into it. An agent connected to your mailbox, file shares and CRM sees everything those permissions reach. The question stops being "what did staff paste" and becomes "what can the agent read", which is a question about identity and access control, not AI.
- Mistakes execute. A wrong answer in a chat is caught when a person reads it. An agent that misreads an instruction can send the email, change the invoice or delete the folder before anyone looks. The controls that matter are the ones you would put on a new junior employee: limited permissions, approval for anything irreversible, and logs.
- Vendor terms decide ownership. As agents gain access to more of the business, the no-training clause and the admin controls in a business-tier contract are worth more than any feature.
What does AI mean for a South African SME specifically?
For a company of ten to three hundred people, AI is mostly good news with three local complications. The good news is that productivity gains which once needed an enterprise budget now come bundled into Microsoft 365 and Google Workspace licences.
- POPIA still applies. The responsibility for personal information stays with the business whatever tool touched it, so an assistant reading customer records needs the same lawful basis and safeguards as a staff member would.
- Data residency is a real choice. Microsoft, Amazon and Google all run cloud regions in Johannesburg or Cape Town. For regulated sectors it is worth choosing services that keep processing in the country where that option exists.
- The rand moves the price. Most AI subscriptions are priced in US dollars, so an add-on that looks cheap in an American blog post costs a South African company noticeably more, and should be justified by a measured saving.
There is a security point too. The same models that help your staff also help attackers write convincing phishing emails and probe systems for known flaws, which makes multi-factor authentication, prompt patching and tested backups non-negotiable. That is the same cybersecurity groundwork a well-run firm should already have.
How much does AI cost a South African SME?
Budget for AI per user per month, and buy it only for the people with a measured use case rather than for the whole company on day one. Copilot Chat costs nothing extra on Microsoft 365 business plans, so it is the sensible starting point for everyone.
A paid assistant has to earn its licence. At about R400 per user per month, Copilot Business for ten people costs about R48,000 a year. For someone whose time costs the business R250 an hour, that licence pays for itself if it saves them about 1.6 hours a month. Measure that in a pilot before you widen it.
How should an SME roll out AI in the first 90 days?
Spend the first 90 days on foundations rather than on picking a model, because every model will change twice before the year is out and the foundations will not. In order:
- Fix identity first. Enforce multi-factor authentication for every account and turn on conditional access. Agents act with the permissions of the account they run under, so an unprotected account becomes an unprotected robot.
- Classify and label data. Decide what is public, internal, confidential and personal, and apply labels in Microsoft Purview or Google's data protection tools. Copilot and Gemini respect those labels; without them, an assistant treats the board pack and the canteen menu the same way.
- Write a one-page AI policy. Which tools are approved, on which tier, what may never be entered, who approves connecting an agent to a system, and how incidents are reported. A policy that fits on one page gets read. AMEA writes these as part of IT policy and documentation.
- Move staff to business tiers. Retire personal accounts used for work. The no-training commitment and the admin controls are the whole point.
- Pilot one tool on one measurable job. Pick a task with a clear before and after, such as first-draft replies to a shared inbox or summarising weekly reports, give the tool read-only access at first, and measure hours saved and errors made for a month before widening it.
- Log and review. Keep audit logs on, and have someone read a sample of what the assistant did each week. This is how you catch the confident mistake before a customer does.
- Train people on the new failure modes. Staff already know not to click phishing links. They also need to know that an assistant can be fed a malicious instruction hidden in a document or web page, and that "the AI said so" is not a verification step.
What should a business not do with AI?
Three mistakes come up again and again, and all three are avoidable. Do not ban AI outright, because staff will use it on their phones with none of your controls. Do not connect an agent to production systems with a global admin account "just for the pilot", because pilots have a habit of becoming permanent. And do not buy a licence for everyone on day one; buy for the team with the clearest use case, prove the saving in rand, then expand.
The preparation is unglamorous: identity, data hygiene, a policy, a measured pilot and people who know what the tools can and cannot be trusted with. Those are the same foundations that make a business secure and compliant, which is why AMEA treats AI adoption as an extension of security and policy work rather than a separate project.