AI has changed marketing work more in two years than the decade before it. Used well, a tool like ChatGPT, Claude or Gemini is a tireless junior teammate: it drafts, brainstorms, summarises, and researches in seconds. Used badly, it floods the world with generic, sometimes-wrong content that all sounds the same. The skill is not 'using AI' — it's directing it and editing it like the senior person in the room.
What AI is genuinely good at in marketing
- Beating the blank page — first drafts of captions, emails, ad variations, outlines you then sharpen.
- Volume and variation — ten headline options, five audience angles, a month of post ideas to choose from.
- Research and summarising — condensing a long report, explaining a competitor's angle, gathering questions customers ask.
- Repurposing — turning one blog into a thread, a caption, an email, a script.
- Images — generating and editing visuals (with LeadAfrik's Ad Studio /adstudio for ad copy and Background Remover /removebg for clean product shots).
The prompt formula
Weak: 'Write an Instagram caption for my shop.' Strong: 'You are a witty copywriter for a Nairobi thrift fashion brand. Write 5 Instagram captions for a new arrivals drop, aimed at style-conscious 20-somethings, each under 25 words, playful but not cringe, ending in a WhatsApp CTA.' Role + context + audience + format + constraints = usable output.
Where AI hurts you if you're not careful
- Confident wrong answers — AI 'hallucinates' facts, figures and sources. For anything factual (a tax rate, a statistic, a claim), verify independently. On Kenyan numbers, check a real source — this is exactly what LeadAfrik's grounded tools and Mjuzi (/mjuzi) exist for.
- Sameness — everyone using the same tool with lazy prompts produces the same bland voice. Your brand voice (module 4) is what AI can't invent; you must impose it.
- Losing your judgement — AI can draft, but it can't decide what's true to your customer or your positioning. That stays your job.
Human-in-the-loop, always
Treat every AI output as a draft from a fast, clever intern who sometimes makes things up. You are the editor. Verify facts, cut the fluff, rewrite in your voice, and take responsibility for what you publish. 'The AI said so' is never a defence to a customer.
A practical AI workflow
- Brief it well — paste your positioning, ICP and voice words so it writes as your brand, not a generic one.
- Generate options — ask for several, not one; you'll pick and combine.
- Edit hard — cut, sharpen, add a real detail only you know, fix the voice.
- Fact-check anything concrete — numbers, claims, names — against a real source.
- Keep a swipe file — save prompts and outputs that worked so you compound over time.
Check your understanding
An AI tool drafts a marketing post stating 'Kenya's VAT is 14%'. What's the correct response?
Check your understanding
Two businesses use the same AI tool. One gets bland, generic captions; the other gets sharp, on-brand ones. What's the likely difference?
Exercise · try it first
Put AI to work, responsibly. (1) Take one real marketing task (a week of captions, or an ad) and write a STRONG prompt using the role + context + audience + format + constraints formula, including your brand voice words from module 4. (2) Generate options, then edit the best one hard — note what you cut and what you added. (3) Identify every concrete claim or number in it and say how you'd verify each. (4) Reflect: which parts of this task should stay human forever, and why?