Getting comfortable with AI at work, without the hype
A digestible orientation for any South African business owner who has heard about AI for three years and not yet built a single workflow around it.
Ty Panaino is the founder of C-Suite Holdings (Pty) Ltd. Since 2017 he has built paid-acquisition, lifecycle, and AI-engineering systems for South African and offshore clients, and now runs C-Suite, two managed tax engines for South African accounting practices.
If you have read three explainer pieces about AI in the last six months and are still unsure where to start, you are in good company. Most of what is written about AI for business is either tech-bro fantasy ("agents will replace your workforce") or a recycled ChatGPT tutorial that does not survive contact with how a South African business works.
This guide is the orientation C-Suite gives a partner before any workflow gets built. Five minutes here saves you four weekends of reading the wrong things.
The mental model that matters
The thing you are actually buying is a fast, capable, slightly forgetful junior who works for a modest monthly subscription and never sleeps.
That junior is excellent at three things:
- Reading and summarising: long documents, email threads, contracts, SARS letters, transcripts.
- Drafting and rewriting: replies, proposals, briefs, social posts, the first version of anything.
- Pattern matching: spotting outliers, categorising, comparing, asking the questions a senior would ask.
That junior is terrible at three things:
- Doing maths reliably: they will confidently get the VAT calculation wrong.
- Knowing what they do not know: they will invent a SARS form number that does not exist.
- Remembering anything from yesterday: every conversation starts from zero unless you tell them otherwise.
The game is learning which tasks fit the strengths and steering clear of the weaknesses.
Where to start (in this order)
1. Pick a paid plan
If you are evaluating AI on the free tiers of ChatGPT, Claude, or Gemini, you are evaluating last year's models. The paid plans (a modest monthly subscription each) get you the current frontier models, the longer memory, and the document upload features that make real work possible.
Pick one. ChatGPT Plus is the safe default for most businesses, Claude Pro suits work that is writing-led, and Gemini Advanced fits a firm that lives inside Google Workspace and wants AI sitting inside Docs and Sheets. There is no wrong answer, so commit to one and use it for thirty days before you compare.
2. Pick one task you do every week
Write down the five tasks you do every single week that you wish you did not have to do. The list usually includes:
- Drafting the same kind of email reply for the fifth time this month.
- Summarising a long document for someone who will not read it.
- Writing the first draft of a proposal that is 80% boilerplate.
- Categorising a stack of incoming things (CVs, invoices, emails, leads, or the unmatched lines in a Xero or Sage bank export).
- Pulling key facts out of a contract or a SARS letter.
Pick the one you resent most; that becomes your first AI workflow. Risk decides the order in which these first tasks are worth trying, and every row below keeps a person on the part that carries consequence:
| Task type | Risk level | What stays with a person |
|---|---|---|
| Reading a SARS letter and extracting the deadline and required action | Low | The response to SARS and the decision on what happens next |
| Redrafting a routine email reply | Low | The final read before it reaches the client |
| Coding the unmatched lines in a Xero or Sage bank export | Low to medium | Accepting, correcting, or querying each suggested line |
| Writing the first draft of a proposal that is 80% boilerplate | Medium | The scope, the fee, and the final edit |
| Reviewing a VAT201 pack before sign-off | High | The calculation and the submission, since AI gets an end-to-end return calculation wrong about 5-10% of the time |
3. Brief the model like a junior
The biggest mistake new AI users make is treating the model like a search engine. You do not type "summarise email". You brief the model the way you would brief a junior.
A good brief has four parts:
- Who is asking ("A partner at a Sandton-based commercial law practice.")
- What you want done ("Draft a reply to the attached client email.")
- Constraints ("Keep it under 150 words, in South African English, and avoid promising a timeline.")
- The raw material (paste the email or upload the document.)
4. Repeat until it is boring
The compounding effect of AI at work shows up when one task becomes ten becomes a hundred. A one-off use changes nothing. A workflow you run every Friday for three months changes the week.
After thirty days of daily use on one task, add a second. After sixty days, add a third. By day ninety you will know more about what AI does inside your business than any consultant pitching you on it.
- 1Pick one paid planChatGPT Plus, Claude Pro, or Gemini Advanced, a modest monthly subscription. Use it for 30 days before comparing.
- 2Pick one weekly taskThe task you do every week that you wish you did not have to do. One task only, to start.
- 3Brief the model like a juniorFour parts: who you are, what you want done, constraints, and the raw material (pasted or uploaded).
- 4Repeat until it is boringAdd a second task at day 30, a third at day 60. By day 90 you understand AI inside your business.
A four-step linear diagram showing: pick a paid plan on a modest monthly subscription, pick one weekly task you wish you did not have to do, brief the model like a junior with a four-part brief, then repeat daily until a second and third task can be added.
The POPIA obligations you carry
This is the part everyone glosses over, and the part that matters most in South Africa.
POPIA (the Protection of Personal Information Act, 4 of 2013) governs how you process personal information about identifiable people. Pasting a client's name, contact details, ID number, or sensitive personal information into ChatGPT or Claude is processing, and the same rules apply that would apply to any other vendor.
What this means in practice:
- Anonymise where you can. Replace real names and identifiers with placeholders before pasting.
- Use the paid plans that offer data processing terms. OpenAI, Anthropic, and Google all publish Data Processing Agreements for their paid business plans. Their free consumer tiers do not.
- Disable training-on-your-data where the option exists. Both ChatGPT and Claude offer this toggle inside settings.
- Treat AI outputs the same way you treat a junior's draft. Reviewed by a human before it goes to a client.
The hard part nobody tells you
The hard part is not the technology. The hard part arrives three weeks in, when you realise the workflow you wanted AI to fix is a process problem wearing a technology problem's clothes. Your proposals take four hours because nobody has documented what a good proposal looks like, and the drafting itself was always the quick part.
That is fine, and it is the most useful thing AI does. It surfaces, at speed, the parts of your business that nobody documented in the first place.
Most of the work C-Suite runs is this. C-Suite watches the function for a couple of weeks, then writes the brief that should have existed all along, with a human reviewer in the loop. After that, half the work runs itself as a managed AI workflow.
Where to go next
- If you run an accounting partnership, the AI for month-end close guide is the next natural read, and AI for accounting firms in South Africa gathers the wider set of firm workflows in one place.
- If you would rather skip the reading and talk through your specific way of working, the discovery call ends with a one-page brief on which function to start with.