The draft policy aims to set out very clearly the benefits and risks of AI, and that these must be fairly distributed. It’s built on 6 core pillars (outlined by IT Law Co):

1. Capacity and Talent Development: Building the national AI skills base through education, training, and industry collaboration, supported by improved digital infrastructure and connectivity.

2. AI for inclusive growth and job creation: Positioning AI as a driver of broad-based economic participation, not just efficiency gains for the already advantaged.

3. Responsible governance: Establishing clear accountability frameworks, including safeguards around data misuse, cybersecurity, misinformation and deepfakes.

4. Ethical and inclusive AI: Centring fairness and bias mitigation to ensure AI systems do not entrench discrimination or exclude vulnerable communityies.

5. Cultural preservation and international integration: Protecting South Africa’s linguistic and cultural diversity while engaging constructively with global AI developments.

6. Human-centred deployment: Keeping people - not systems, at the heart of how AI is adopted and used.

All of this speaks to trust.

Developing talent, job creation, responsible governance, ethics, and human-centered deployment - I’m so encouraged to see that local government is taking a measured, sober approach to tech that has been spoken about quite frivolously, as if it’s this black box. AI doesn’t exist in a vacuum, and the policy invites input NOW to help shape what the future looks like. I’m incredibly excited.

The policy has been approved for publication, but has not yet appeared in the Government Gazette. It’s currently open for public comment, which is critical in that it gives the public a chance to define what responsible, inclusive and beneficial AI governance looks like in the years to come. In terms of a final date for the policy to be in place, think 2026/2027 financial year, with sector-specific regulations and guidelines to follow in 2027/2028. But that doesn’t mean you’ve got a lot of time to get your ducks in a row.

What should be top of mind for marketers, brands, and agencies?

1. Once this policy is gazetted and the policy is finalised, the trade-offs and implications of using AI now become tangible. Accountability is the first step to building trust, and an approved AI policy (come 2028) will leave little wiggle-room for ignorance. For brands, this is critical: AI adoption isn’t a flashy thing, but it has a real impact in how businesses will operate moving forward.

2. Governance is critical. Think mandatory disclosure of where AI has been used (internally and externally, stricter compliance loops to hop through, and audit trails for the use of AI (particularly by agencies). What do I mean by disclosure? Think Toyota’s Starlet ad, where they disclose where AI has been used.

   What do I mean by stricter compliance loops? Think additional sign off that factor in ethical AI use. What do I mean by audit trails? Think proof that you used AI ethically, and didn’t copy an artist’s style in generating an image or a song.

3. Risk isn’t just legal anymore - it’s reputational. I’ve seen ads on KykNet where an AI-generated doctor is recommending a supplement, but the fact that it’s AI (think warped logos and packaging) isn’t disclosed. Hell, I’ve even seen a deepfake of South African president Cyril Ramaphosa encouraging South Africans in a “family meeting” (but in a YouTube Pre-roll) invest in a pyramid scheme. These brands will be in for big legal trouble, and they should be held accountable.

4. Trust will be a competitive advantage. Brands who draw a hard line on how they want to use AI will win in the long term. Strategically, it makes sense for a global brand like Aerie to take a stance from never using AI-generated people. They’re quite vocal about it on their Instagram. But that doesn’t mean they can’t use AI features like Adobe’s Generative Fill to remove an uneven shadow on a product photography shot. Those little distinctions are going to be help brands navigate the tradeoffs I wrote about earlier. Deciding what you want to classify as “AI-use” and backing that up in an internal policy is critical.

What do brands/marketers/agencies need to get in place?

This policy isn’t legally enforceable, but it sets an intent for how government would develop standards and regulations around it. We’re a step before enforcement and consequences come. If you’re working in marketing or in agency and haven’t considered this at all, here’s what you need to focus on right now:

An internal AI policy:

A document that is operational, and sets you and your organisations and suppliers up for success. This should cover what AI tools are allowed, what the approval process is, and principles. This enables a consistent understanding across teams, that not only protects you, but sets you up for better adoption, and clarity on consequences. This policy should be informed by the business’ goals - not airy-fairy fuzz.

AI principles:

This would be your first step in transparency, in that it’s a great step in building trust. It should outline how you’re using AI in your function, what you define as “AI” and rules for the application. I’ve written about Hasbro’s AI principles before, and it’s a great example of what to consider.

Ownership:

Most AI platforms/software have very opaque rules around ownership. If you use ChatGPT to generate imagery for example, ownership of what you created depends on the plan you’re on - which introduces serious risk for brands if a third-party supplier or an agency has used a personal/free version of an image generating tool. And these platforms’ terms of services change ALL THE TIME, which is another governance layer to consider. This isn’t exclusive to novelty image generation. Let’s look at CoPilot, a tool that’s been championed as an efficiency hero and costs around US$30 a month. Buried in Microsoft’s terms of use, for example, is a stark contrast to how AI evangelists have been talking about CoPilot’s use:

You need to be very, very, very clear on the tools you use, how you use them, who owns what you’ve created, and what that means for intellectual property. And this needs to be considered from a lens of trust, just like the Draft Policy requires.

Share Synthetic Taste

South Africa’s draft policy is looking to shape a future where trust is critical, and transparency and accountability is non-negotiable. I’ve been advocating for a long-term perspective on AI since I have started this substack, and i’m truly hopeful to see that government is doing the same.

It’s something I’ve spoken to so many colleagues about, and the knock-on effect of a commitment like this does wonders for how a brand shows up to consumer.

Hasbro, a global toy company (think Transformers, Monopoly, Play-Doh, Magic the Gathering, etc), recently went public with what they call their “AI principles”. Chris Cocks (Hasbro CEO) said in February 2026 they’re deploying AI across financial planning, forecasting, order management, and other daily activities (Source here). Their brand purpose is to “create joy and community for all people around the world” (Source here). That feels quite simple, and direct.

But what I’ve found incredibly refreshing, is that the CEO says that teams have a choice on how they use technology, including “not to use it at all when it doesn’t fit their objectives”.

I found this approach incredible. It’s transparent, it inspires exploration and it drives adoption. it keeps the focus on what the company’s actually doing, and though it is incorporating AI, the company is not doing so just the sake of incorporating it.

Where a lot of tech companies (and advertising holding companies) are scapegoating AI for mass layoffs, and talking about digital labour, and restructuring work around AI’s promise, Hasbro has taken an approach that doesn’t deny AI’s promised potential. They’re simply hedging their bets on what they can control - and linking that to their business focus. It keeps their focus on what makes their business tick, while at the same time encouraging people to play. This approach is a long-term one, and one I can get behind.

It doesn’t provide shiny novelty stunts and press releases nobody asked for. It doesn’t deviate from what the brand is known for. And it basically says (internally and externally): We’re using AI, but don’t let it distract you from the fact that we’re selling Monopoly.

From an ad agency perspective, having an approach like this in place is gold.

If a brand is clear on how they’re going to use AI, and the focus of the organisation remains on the business (in this case, toys and joy and community) - it not only sets the brand up for success, but the brand’s staff, partners, and customers.

For the brand’s staff: If AI is optional where it makes sense, it’s incredibly liberating. Instead of horseshoeing it into a function it has no place in (i’m looking directly at Meta’s search tabs) adoption becomes a no-brainer. If someone finds success in using a tool and they get to define what that success means within their role - it removes the hype/fear and immediately gets back to basics. They’re able to define where AI is relevant, which means they’re able to measure the actual impact - and not vaguely evaluate buzzwords so often thrown around.

For the brand’s partners: Hasbro says “final work comes from people” - which encourages accountability, but also makes it very clear where AI’s place is in what they do. If i’m their ad agency, i know that efficiency will come over time, and that it’s not an immediate metric put in place with no reconciliation of promise vs reality.

For their consumers: Transparency is key, and a core principle of theirs literally outlines “We will be clear about when AI is used, what it does, and what data it relies on”, which immediately builds trust.

My favourite part of their principles is this: Creators may use tools for reference, but the final work comes from people, not machines. What we’ve said previously on this topic remains unchanged.

Here’s a link to their AI Principles.

I love this for Hasbro. It’s clear, it’s transparent, and it sets a standard for how brands should think about incorporating AI - not for the short-term, but for the long term.

Generative AI is a tool for executing decisions, and most of marketing’s value lies in making those decisions. Confusing execution and decision is what we’re getting wrong.

Studies evaluating AI performance (like this one from OpenAI) assess how models complete specific, bounded assignments.

OpenAI’s paper references “real world tasks” - what they reference specifically is “a legal brief”, an “engineering blueprint”, a “customer support conversation,” or a “nursing care plan” - and the focus is “predominantly knowledge work”. The researchers developed a valuation called GDPval with the intent to track model improvement over time, focusing on how models perform on economically valuable tasks.

These tasks went through multi-step reviews with experts in each of the fields they identified to quantify results for clear evaluation. These are structured tasks with clear prompts, reference material and evaluation criteria that is hard to actually replicate in the real world.

The framing of “real world tasks” in this paper is problematic because it assumes everyone defines “real world tasks” as outputs of jobs, not parts of jobs. That important little subtext from the end of the article (under the heading “Limitations and what’s next” reframes AI’s capability severely. The research in reality measures performance on bounded assignments, but presents it in a way that collapses that assignment into a job function.

AI is strong at:

• Variant generation (if needed, which requires checking)

• Rewriting for channels (which requires brand tone checking anyway)

• Summarising research inputs (which requires checking anyway)

• Production acceleration (first drafts, which aren’t “final” by any means)

These are execution-heavy activities.

AI is weak at:

• Deciding what problem to solve

• Picking the direction a campaign should go after

• Understanding the nuances of audience/channels/relevance

• Owning outcomes

These activities remain hard to formalise, because the success of picking the right problems to solve are rarely fixed.

Most AI companies out there promise efficiency - more, cheaper, better, faster. But in a real-world context, AI introduces a lot more friction that is seldom called out. Speed isn’t synonymous with efficiency.

So, as brands, agencies, and advertising professionals, where do we go from here? Our jobs are to decide what problem is actually being solved - whether it’s a product launch, a brand strategy, or a social media campaign that addresses a business problem. Let’s put this into context - we’re an industry full of human problems, business problems, variables, and ambiguity.

Most briefs brands share with agencies are variations of a recurring problem:

• Launch this product/offering

• Reach this audience through this medium

• How can we target audience x/y/z to get them to do a/b/c?

Some of these asks seem repetitive, but the underlying conditions that inform these briefs make it complex. Categories in briefs repeat, but the constraints never do. The commercial context, the risk appetite, the stakeholders involved brand history, and the audience tension all provide complexity - a field in which AI has very little leverage, and it doesn’t resolve these variables.

A more practical way to think about it

Rather than assuming AI can do marketing, it’s more useful to separate work into two categories:

Non-scalable work: Work that depends on interpretation, negotiation and judgment under uncertainty

Scalable work: Work that is repetitive, and applies clear decisions across formats, audiences or channels.

AI doesn’t answer nuanced questions. It cannot weigh trade-offs between long-term brand equity and short-term performance. It cannot judge organisational risk. It cannot tell you whether a message is “good” or strategically the right one. It can only generate expressions once the tough decisions have already been made.

This is why the promised efficiency often feels disconnected from reality. AI accelerates output, sure. The actual constraint we’re facing isn’t producing assets - it’s choosing which ones are worth backing.

The shift we should be after:

Up until now, we didn’t have to think about different workflows, because the tech didn’t require us to. Instead of trying to “AI-power” things, we need to separate the non-scalable work and the scalable work.

I developed a 4-step evaluation framework to help you figure out if AI can help:

STEP 1 // Frame the problem

What are we actually trying to solve? If the inputs are ambiguous or contested, AI has no role to play.

STEP 2 // Set the direction // How can you start?

Which trade-offs are we willing to make across brand, timing, cost and risk? This is about decision, not options.

STEP 3 // Expressing the decision // How can you get to making

Converting strategy into work: If you’ve got clear direction, a clear goal, and the variables are minimal, AI starts to become useful.

STEP 4 // Scaling // How you get to draw the benefit

When no new thinking is required, and you’re looking to format/adapt/iterate, the right automation tools can help.

Let’s apply this framework to a typical agency brief: A brand wants to launch a new soft drink.

AI has made me ask much weirder questions. The problem we’re trying to solve isn’t in producing work, but defining which work deserves to exist. AI increases the volume of what can be produced, which puts pressure on us to define what is worth making. Organisations need to become far more deliberate about what gets produced in the first place, and who should.

This is a dangerous statement - not just because it feels like a prediction, but because it ignores reality. AI does disrupt our traditional views of creative labor, absolutely - but seeing people recycle this click-bait style hot take, screams “give me attention”.

In 2023, Forrester forecast that 7.5% of agency jobs will be automated by 2030. It’s an interesting study that makes valid points. In this study, they predict an inversion of what the agency workforce looks like - calling out “senior managers to high-paid creator skillsets paired with genAI assistants”. The problem? This study is focused on the United States. Not South Africa. Our challenges, operating models and economic challenges are vastly different.

Subscribe now

The reality that agencies face today is simple: Clients want faster turnaround, more work for less, and also effectiveness. Unfortunately, these desires come with tradeoffs.

For faster turnaround, you compromise on complexity.

For more work for less, some craft will slide.

For effectiveness, you need time.

There’s no real way to work around this. But to scratch the surface of this take, and to give the “future agency” speculation the benefit of the doubt, let’s take a real-world scenario:

You’re running an integrated independent Agency in South Africa, delivering B2C Social, Digital and Above-the-line work. You have three creative directors (plus an army of AI tools) handling one blue-chip client and four smaller ones.

Your blue-chip client is a banking brand, with Retail banking, CIB, and Sponsorship business units.

Your smaller clients are a local fast-food chain, a part of a major car brand (let’s say the digital work), a fin-tech start up, and a local fashion brand.

Now, let’s talk scope:

The bank? Multiple marketing heads and managers, messaging frameworks across business unites, strategy, always-on creative, campaigns, sponsorship toolkits, production, compliance revisions and templates.

The smaller clients’ scope would include a mix of promotional campaigns, social content, launches, digital always-on work, paid media assets, website content, performance creatives, investor decks, seasonal campaigns, lookbooks and influencer toolkits.

I’m going to ignore the obvious:

• Revisions caused by indecision.

• Internal alignment time.

• “Just one more tweak”.

• Stakeholder politics.

Realistically, three creative directors with AI can probably conceptualise, write, design, and present to clients.

They won’t be able to manage the volume of adaptations a blue-chip requires, or keep up with revisions. The context switching required to do the above is a fast-track to burning out, and you’re asking three creative people to move between the regulated banking sector, promotional fast food content, digitally focused car-content, abstract fin-tech growth goals and exclusivity in their fashion client.

Where the hot take really falls apart.

Creative delivery requires taste.

Junior resources don’t just execute. They bring their lived experience to the table: they absorb cultural nuance, understand and anticipate clients (and politics), and learn how to read a room. They’re also critical in sensing when a brand will overstep culturally - which really reduces risk when you service a brand as an agency.

Client Relationships are labour intensive.

Much of agency value lives in translation, reassurance, accountability; iteration and expectation management. That work doesn’t disappear.

AI speeds up parts of production - not decision cycles.

If you remember one thing from this piece, let it be the above.

What’s probably more likely:

Hybrid juniors who are creative thinkers with strategic vision and AI literacy,

CD’s acting as system designers: Setting prompts, creative guardrails, brand memory and evaluation criteria. I think that changes fundamentally what creative directors used to do, and the title probably needs rethinking.

What I think the agency of the future looks like:

Agencies should become cultural partners: Human heavy, slower, higher margin, fewer clients - long live the indies!

AI supporting execution once the above is done: Automating repetition, where clients understand the real effort required.

I do not think that the future of agencies are senior creatives with generative AI. That’s incredibly oversimplified. It’s what it’s always been, just slightly different: A cultural partner that makes brands relevant in the lives of consumers.

If we look at how the leaders in AI are talking about, we see vague futurism. Words matter. Broad, hyperbolic language in public forums get clipped for social media, and the more tempered views are shared in expert analysis forums. The gap between these two modes is where credibility erodes.

Let’s take Elon Musk, who said AI will “make work optional” in “maybe… 10-20 years, something like that.” This guy. Looking at how the job market is evolving with AI, this honestly feels a tad exaggerated.

Sam Altman, from OpenAI, said in his newsletter from 2024 ASI (Artificial Super Intelligence) is “thousands of days away”. When asked about this statement later, he admits it’s a “bold statement”. This is the same person who initiated 2024 said “Ads plus AI is sort of uniquely unsettling to me”, saying “I kind of think of ads as a last resort for a business model.” In 2026, OpenAI. Announced they are testing ads in ChatGPT.

Is this the “last resort”?

In 2024, Mark Zuckerberg told Dwarfish Patel that AI would write Meta’s Llama project’s code by 2026 - and that AI can “write code like a mid-level engineer”. These predictions haven’t happened - software engineering remains human-led in most contexts, and AI simply augments.

The way we talk about AI makes honest evaluation harder.

If I told my boss an incredible project I’m working on is “thousands of days away”, and say “maybe, something like that, potentially” the way these fellas do, my credibility would be shot. But the pattern we keep seeing, is confident claims, quietly recalibrated. These words don’t describe uncertainty. They defer accountability.

I’ve adopted a simple rule that’s helped me a ton. If you can’t explain the value of whatever AI tool/case study/thing you’re pitching to me like I’m 5 years old, the promise is probably inflated. Complexity is real, but value should still be legible.

This becomes obvious when you look at how products are marketed. I keep seeing ads for an AI meeting notetaker, and bear with me - I’m going to break it down.

This AI-powered tool promises focus, clarity, action, alignment, memory, and momentum. But if you, break it down to a five year old: You’re paying for a tool that records meetings, summarises them, creates tasks, and makes notes searchable. Which sounds useful, sure. But they’re not revolutionary. It promises “efficiency” - but what it actually means, is I’m reading (another) email, I’m now worried about data (privacy & usage policies are generally quite murky), and in order for me to get the most out of the tool, I need to use the features adjacent to the meeting notetaker, and pay more to access these features, and end up getting more emails.

That already makes one less efficient. It doesn’t make anything disappear from the user’s day. It just adds more fluff. Productivity is being rearranged - and that’s the nature of where we’re at at the moment. After three years of AI everywhere, most things feel the same as they did before - if we’re being honest. We’re still making ads. We’re still trying to come up with ways to connect people with brands. There’s just a lot more thinking & explaining that needs to happen.

This is one example of how the promised hype doesn’t quite meet reality. Three years on, the value is incremental - not earth shattering. Let’s not do what Sam or Elon or Zuck’s doing. Let’s do talk about AI in a way that we’re able to explain it’s impact to a five year old - and in a way that it still makes sense.

This week, a consulting firm made headlines after submitting a report to a government client. They allegedly used AI to complete the report, and after several errors were found in this report, the firm had to provide a partial refund for the US$440000 report. That’s the first of many lessons I hope we all can learn from.

In advertising, I think we’ve yet to talk about the trade-offs. I know some folks feel we can have both, but the idealist and realist in me struggles to reconcile the two. That old project management pyramid - time, cost, and quality are involved: Pick two - still rings true.

I reimagined it for AI, based on the promises involved, and it looks like this:

What happens if you pick Safety + Scale at the expense of Speed?

This is the cautious path, veering away from innovation, and honestly, the status quo. It’s what everybody working in advertising has always been doing. You’re not chasing “more, faster” here. You’re chasing controlled creative, brand safety, and not tapping into AI’s promise of speed. You risk slower rollout, but you don’t have to satisfy the need to do more faster - because time isn’t the priority here.

What happens when you pick Speed and Scale, and trade off Safety?

The benefits of this approach? Rapid rollout of assets, many variants of creative, real-time optimisation.

What you trade off: You sacrifice credibility, and you open yourself up to reputational risk.

An example of this would be the classic Coca-Cola ad reimagined for the age of AI.

They scaled out visual work at speed, but risked losing trust in how authentic the brand actually is.

Another example would be the viral AI companion start up Friend’s ads (read more here: https://www.businessinsider.com/ai-companion-startup-friend-calls-defaced-subway-ads-entertaining-2025-10?utm_source=chatgpt.com).

They ran the largest subway ad blitz in history (literally 11000+ ads).

The public defaced them with graffiti messaging, resulting in severe reputational backlash.

This is a great example of Speed and Scale triggering the erosion of trust.

What happens if you pick Speed and Safety at the expense of Scale?

This is a focused route, and where you hedge your bets. The benefit of this approach is a controlled output, which strongly aligns with brand, but you’re still experimenting. You end up trading off experiments and potentially higher cost.

An example of this would be the Nike x Serena Williams ad, from when she announced her retirement.

They didn’t milk this campaign - they did one singular bold art piece to demonstrate their idea, preserving their brand image while not lagging behind.

It’s a wild time we’re living in. Like I mentioned in my previous mailer, the public’s still not entirely keen on AI-assisted advertising - no matter how good the craft is. But using AI carelessly can cause reputational damage that erodes brand trust, brand love, and overall how brands’ customers view them.

However, nobody truly wants to fall behind. Refusing or under-investing in AI also carries a cost, as the general public will become used to seeing AI-assisted work, and it will undoubtedly make certain parts of production (ie resizing imagery/editing) easier. A big mistake (like Coca-Cola’s AI generated ad) hurts massively in the short-term, but innovation (like Fizzion, read about it here) will benefit them in the long-term.

Subscribe now

Generative AI doesn’t give you everything. It really gives you choices, and each one of them has a cost. The gains come with trade-offs, and pretending otherwise might just bring on a pretty expensive lesson, like that consulting firm learned.

The threats/parallels between the arrival of digital and the arrival of GenAI are quite similar, and press from the year 2000 (read the full article here - what a nice look back in time).

The cycle’s the same: Consumer habits are what they are (whether it’s 2000 or 2025), new tech/change comes, and people get mad, it becomes normal, and people move on.

Why I like this analogy, is that it puts the chaos/hot takes/nonsense into context. When agencies started producing ads using digital, some of the drawbacks on film became more apparent:

• Clients/directors/agency could see immediate playback on set, versus having to wait for dailies to be processed.

• Film reels could be lost, scratched or damaged, with digital formats magically storing hundreds of gigs of footage on harddrives/servers/etc.

• Every take costs money - and film stock, processing, reshoots etc cost bags, and bags, and bags of money. Digital made “one more take” pretty much free.

• Roles also had to change: film labs closed, negative cutters were no longer needed, couriers moving reels between production and agency lost business - but new jobs emerged, allowing editors, color graders, and later even compositors to work.

The whole way of work - including roles, risks, possibilities and creativity - shifted. The consumer’s reception to the work is something entirely different - but the shift today is pretty much the same.

We’re really super early in this journey of generative AI. Nobody’s an expert at this, everybody’s guessing, and I think few people are hedging their bets in the right way.

The proof will ultimately be in the pudding, and AI-assisted ads are becoming more and more visible on our phones, our televisions, and even cinema.

Kalshi’s viral NBA Finals ad was the first piece of slop I couldn’t really fathom. It was made by a creator (who calls himself an “AI filmmaker”) producing the ad in 2-3 days, with production costing around US$2000. The creator generated 300-400 different clips, from which 15 were selected, and stitched together.

The ad feels kinda appropriate for the sector (where regulatory risk is quite murky) and it’s flighted on a prime slot. It’s a good example of what you get when brand interests align with opportunity, and a great example of what you can get in 2-3 days’ work.

The NFL launched their season teaser ad (in partnership with 72andSunny) and showed how AI blends/compliments the process of ad making. It took 165 people, and the total unique job titles were 113. This must’ve taken ages to make - but the craft is so good! Yes, the synthetic baby is terrifying, but the float itself, the ice statue of Joe Brrr, the self-deprecating Druski cameo that lets the audience into the joke - it’s just lovely. But it also shows what’s possible when you expand your skillset. Hell yes for adapting with the times!

However, public sentiment is still a bit iffy. The YouTube comments on both of these ads are an absolute cesspool. A lot of negativity, a lot of “put it in the AI slop pile”, a lot of “this is not impressive”.

“Collapse of civilisation” is a bit excessive. But unfortunately, this isn’t going away - and it’s ultimately going to be up to brands to really interrogate why they want to use AI in the first place. Yes, it’ll cut costs. It’s going to take time for AI-generated work to become palatable for the public. It’s short-sighted to not see the bigger picture and think about what’s happening: It’s new tech is here, people are mad, and soon it’s going to be normal.

Like with the coming of digital when film dominated, it’s just another technological advance people will inherently become used to. In 5 years’ time, nobody’s going to care that you used AI.

But personally, I believe people are going to care why and how you did. And whether you’re happy with the creative product you’re putting out there is the main thing: no matter the medium.

Here’s what the Collins dictionary says.

Efficiency means achieving the maximum possible output with the minimum amount of wasted resources. For example, a machine can be efficient if it produces a lot of work using little energy. A process can be efficient if it gets results quickly without unnecessary steps. I don’t count “time” as a resource here.

I rarely hear how people contextualise “efficiency”, but it’s thrown around so much - especially when AI is the topic of conversation. Now granted, that’s always been how people sell tech (as an efficiency driver), but again - this only sounds appealing when said out of context.

I asked ChatGPT to define it for me in an advertising context, and it’s a great place to start.

Operational efficiency (work process) sounds like a great way to talk about what you’re actually after. And though time is a factor here, it’s a factor in the context of process. Let’s say you’re rolling out a bunch of banners - “required work”, that is templatised. This kind of work isn’t the heart-strings, brand-love, TVC or creative idea type work. It’s the kind of work that can be automated.

That’s a major callout I rarely hear on panels, webinars or in articles. Just a bit of nuance to the hype would do wonders for managing expectations, being clear on what you’re working towards, and actually being able to back up what you say.

I’m a big fan of research, and an even bigger fan of putting the research into real-world context. This is from IBM’s paper, “How does AI improve efficiency”:

In theory, this sounds incredible. In reality, this is speculative. I think efficiency should remove friction. The advancements in AI however, creates friction. Release notes for tools get updated every week (doesn’t matter the platform, doesn’t matter the tech-giant), and adopting new features when the principles aren’t even agreed on is a major challenge. If we’re being brutally honest, AI has only really been spoken about over the last three years - and marketing was the first thing people say will suffer. As an industry, we’re still adopting the tech - and being at the forefront of this for the better part of the last three years has really taught me that you need to be clear on what you’re after.

What I do know to be true:

• Tools need people to produce. Whoever’s operating the tool (a copywriter, a student, or a LinkedIn grifter), is what influences the output. And the output will always feel robotic.

• Ai is a multiplier - it can produce a ton of stuff. But sometimes, client’s don’t need a lot.

• AI isn’t automation, and my word do people confuse the two.

• Effectiveness is a more measurable focus than efficiency. Doing stuff faster or cheaper is cool, but that doesn’t mean the work actually worked.

• Differentiation matters more than ever. If everybody’s producing “efficient, AI-generated” content, what’s ultimately going to stand out is authenticity, the brand’s voice, and a unique point of view.

Efficiency isn’t special. Everyone’s racing to make ads fast, with AI, and cheaper. But as an industry, I genuinely we should remember that we’re competing in how creative, original and effective we are. Not how fast we can generate ads, or make commercials, or hyper-personalise banner ad messaging.

We’ve all heard that AI makes the possibilities of creative endless. Sure! I agree. But what does that actually mean? Generally, when people have elaborated on what “creative possibilities” AI actually opens up, it just results in more noise - and more confusion.

We’re stuck in a loop where the top dogs launch updates/new features, and the start ups launch new tools, promising the world, and caveatting the hell out of it.

I wanted to explore some claims I saw on LinkedIn again. Bear with me.

“You can draft creative faster, because AI can write ad copy in seconds/generate imagery in seconds/generate video in minutes!”

• That’s really cool, but a 10-second draft is useless if you still have to spend time refining it into something that is ready for a client. The benefit of whatever you’ve generated in minutes kinda doesn’t matter if the feedback requires you to make the logo bigger.

“AI can instantly turn one ad into hundreds of assets!”

• Totally get this one. But the only way this actually works is if the work is templatised, and then it’s technically automation. Approving hundreds of different assets all saying the same thing sounds like a nightmare. Also: What’s the point of turning one ad into hundreds of other ads - even if they’re personalised - if you don’t have the media budget to test?

“AI minimises production lag!”

• That sounds incredible on face value, but what really causes production/roll-out lag (unclear feedback, indecision, revisions) isn’t something AI should technically solve. It’s a distraction, because AI cannot solve human ambiguity. “Make the logo bigger” is one thing. “We’re changing the brief slightly” is a whole different thing.

Just because something promises to be shiny, doesn’t mean it is.

When I train teams on AI, I always include an AI generated image of a guy fixing a flat tire with a banana. Sure, you can forcefit AI into solve problems, and shout about how incredible it is. But sometimes, it isn’t the right tool for the job.

It’s a now debunked internet myth that you can fix a tyre with a banana - but the principle remains: You can use superglue to repair cracked nails. You can use dental floss to cut soft foods. If you’re Lady Gaga, you can use meat as a dress. Should you?

This past weekend, I went to a conference. This particular conference was about AI in marketing, and the attendees were marketing big dogs, advertising leaders, innovation folks, trend analysts, and a few start up bros. I loved Gianni Giacomelli’s talk (Great guy to follow, by the way!). He said a few things I haven’t heard before, and brought a fresh perspective to a shallow landscape.

Not once in his talk did he mention “efficiency”. I found it curious, because it’s the first thing we see on LinkedIn when people (who don’t even work in marketing) talk about advertising. The word means something different to everybody, and because nobody’s measuring it in the same way, it sets creative agencies up for failure and it gives clients an inflated expectation of what’s really possible. Instead, his talk offered this pearler:

We should move from asking “how can we use AI” to “Why should we use AI”.

I recently saw a fully AI-generated TVC on KykNet. It was advertising a brand called Sinupret SA, and when I tried to find it, I saw it’s been removed from YouTube. (You can have a look at their Instagram here, where you will see some synthetic humans enjoying relief from colds and flu.) The TVC features an AI-generated voice over, a distorted packshot, a grimacing synthetic human, an AI-generated doctor recommending Sinupret as a solution to colds and an end frame saying: “Doctors trust Sinupret. Why wouldn’t you?” Ethically, a bit weird (but that’s a topic for another day). Seeing AI-generated slop sandwiched between a high-production TV ads for Nedbank and MTN felt a bit dystopian.

I understand it’s a trade off. Sometimes, brands don’t have budget to shoot. Sometimes, the medium isn’t the message. But reconciling what AI is good at, with what you actually want to do, is where you should start.

If you want to tone down the noise, it really is much more useful to ask “why” instead of “how”.

The rules out around AI right now are fluffy - whether you’re asking questions as an agency, a brand, or just a human on this ball of rock and water, the answers are largely neither right nor wrong.

Who owns an AI generated image?

What about style?

Where do you draw the line on ownership?

If you use generative fill does it really count as “AI” - the same way we comp imagery?

Can you ethically use a still image you shot, run it through a video generating tool, and make it seem like the person is doing something they didn’t actually do?

There are no right or wrong answers here. We’re all considering costs (because budgets), we’re all considering the endless creativity available now (because novelty), and we’re all trying to figure out how this actually works.

I’ve got a few thoughts if you’re struggling to answer these questions.

Your first starting point should be your stance.

Your second should be ethics.

Your third should be your vision, answering the question: “What do we want to be known for?”

Your stance on the use of AI should be informed by your values. If your brand is looking to promote trust, is it ok to use synthetic humans in ads - moving or still? What about voices? What do my consumers think? (There’s a whole lack of research around this, so to opt for data with vague sources isn’t the answer.) There’s the old Levi’s situation from 2023, where they used synthetic models to “increase diversity”, which brought a whole lot of backlash particularly from consumers. Brave/Innovative/Disruptive move in the moment - but 3 years on? Not so much.

Your second point of call would be your ethics, which I think should also informed by your values. If your brand is big on integrity, do you want to disclose where you’ve used AI? Does it matter? Currently, nobody is legally required to disclose the use of AI in the way Financial Services Providers are required to disclose that they’re legit. Would you still consider it? And what if your organisation is big on people? Are you comfortable letting people go for the sake of Tech and innovation?

Klarna slashed their headcount by 24%, backed by their CEO (Sebastian Siemiatkowski) saying in 2024: “AI can already do all of the jobs that we as humans can do… we stopped hiring about a year ago.” In this article by Lance Huan, the story’s told brilliantly. Siematkowski has now backtracked, saying their AI-focused strategy wasn’t the right path. Siemiatkowski told Bloomberg in 2025 “Really, investing in the quality of human support is the way of the future for us.” (Source here.)

Answering the question “what do we want to be known for” is a HARD one. It’s part what you’re known for already, and part what you want to be known for. From consumers, partners, and staff - all of it’s important. In the age of AI, think about your brand 5 years from now. When all the noise is gone, when everyone’s using the same tech, when tech’s been regulated - your brand will be remembered for what it has prioritised.

I will always champion taking a long-term view. Having a vision for your brand, sticking to it, and not confusing attention/clicks/“claimed” efficiency for what your customers think is key.

We’re ultimately creating the world we (and our children) will live in, and keeping that at the back of my head (even as a lowly creative) always helps.

Vania Heymann’s work on ASAP Rocky’s “Tailor Swif” is a great example of this craziness, and the comments on the video are filled with gems of discussion. It got a Grammy Nomination, and a few D&AD’s, because undoubtedly it’s craft at it’s finest - tapping into an avant-garde aesthetic that resonates with a digitally savvy audience.

The latest Sabrina Carpenter video is another reflection of the director’s zany brain, with missable, random details packed into every frame that begs you to watch it again because you can’t believe your eyes.

Nabil Elderkin is another director with a completely different aesthetic. His videos are thoughtful with a dark undertone, and the songs reflect the style. Take his work on Foals’ “Late Night” for example:

AI has come around, and people have dabbled in making music videos using AI. Most prominently, unfortunately, Kanye West. In a video for his song “530”, we see a range of AI-generated clips stitched together - grossly highlighting puppets resembling women undergoing surgery.

Subjectively, it’s not for me. It’s unsettling, gross, and completely devoid of a narrative. I can’t see anything about the artist, or his style. It kind of feels like “ai for the sake of ai”. I thought well, let’s look at other AI videos - and guess what, Kanye came up again. This time for Vultures in what is called “Havoc version”, they tapped Jon Rafman (“internet artist” and creator of “the first AI Pop Star “Cl0udyH3art”). The result? Visuals that don’t really articulate anything that kinda feels like an experiment.

I can appreciate experimentation, and all the noisy, unconstructive disclaimers around “the tech will get better”, but I’m not willing to applaud an “AI video” because it was made with AI.

It’s like applauding an ice cream that is served in a croissant instead of a cone. It’s cool, I guess - but not for everybody.

The same thing is happening in advertising: Novelty for the sake of novelty. I’m a champion of work that makes people feel something, but I hope we can really soon start to talk about ads that are good, instead of “AI ads that are good”.

The medium used to be the message. I think today, it’s more about creating stuff worth engaging with.

There’s a buzzword that has completely usurped people’s ability to think for themselves. It’s GenAI. Whether it’s posting about AI, making something novel with AI, or posting a hot take - everybody’s talking, but nobody’s really saying anything.

Every time I see a LinkedIn post starting with “(insert AI tool here) has just replaced the creative agency/art director/film maker”, I roll my eyes. Every time I see a panel discussing the what ifs, the future, and the implications of AI, I leave disappointed. Every new piece of tech (which, 8 out of 10 times, is just ChatGPT and a few image gen models wrapped into one) claiming that it’s game changing, I groan.

“I made an ad in seconds.”

“Midjourney will replace art directors.”

“(Insert tool here) just changed everything.”

If you were to ask the poster to explain what they’re saying to a 5-year-old, it’d be gibberish. Even worse, if you ask them to contextualise it - the claim will lose it’s relevancy, punchiness, and credibility.

I’m not a hater of AI. I think it’s useful. What I don’t like is that people are blowing it completely out of proportion for the sake of clicks. I want the dialogue to change.

In the world of advertising (whether you’re client or agency side), it’s even worse. At a conference here in Johannesburg, I heard someone say:

I damn near choked on my disbelief. In the moment, it was a hot take, and the audience nodded and felt some sort of unity in the uncertain times this person predicted.

But that was it. No context. No nuance. No follow ups. Just a hot take.

The only way I could look at that statement and make sense of it, was to assume that this person’s desire to say things first has become more important than logic, context and well - accountability.

I started this Substack because I’m hoping to debunk some of the nonsense with some critical thinking. Hiding behind the mystique of buzzwords, word salad, and quotes from tech companies about what’s coming isn’t constructive to any sort of conversation. The “what if”, “not yet”, “potentially” caveats to every chat about AI doesn’t help anyone.

My hope is that in reading this, you learn something, and that we can raise the dialogue around AI in advertising a bit. Less sweeping statements. Less noise. More accountability and more creativity.

I would love some comments/questions too - because we’re all struggling with the same noise. Let’s bring some sense into the nonsense.