Ethical AI in South Africa: Managing Business Risk

SA businesses are embracing AI, but ethical blind spots are becoming a major operational risk. 

 

South African organisations are accelerating their adoption of Artificial Intelligence, but many are overlooking the single factor that could derail efficiency gains, expose them to regulatory scrutiny, and damage brand trust, says Alkemi Collective CEO Bradly Howland. 

 

He argues that ethical AI is no longer a marketing conversation, but a core business governance issue that belongs on leadership agendas alongside compliance, cybersecurity, and reputational risk. 

 

“AI is now integrated across customer service, fraud detection, recruitment, content creation, and decision support systems. Yet the governance frameworks around how these tools are designed, trained, and deployed remain worryingly thin,” he says. “Companies are moving faster than their risk controls, which is a red flag for any board.” 

 

Recent investigations into bias within fraud detection tools used by medical schemes have highlighted how untested or unbalanced datasets can result in discriminatory outcomes, operational failures, and significant reputational exposure. 

 

“These are not theoretical risks. When an algorithm misclassifies, the consequences are felt by real people and the financial and reputational fallout lands with the organisation, not the software provider,” Howland adds. “You cannot outsource accountability to a machine.” 

 

He believes the challenge for South African businesses is twofold. First, the country’s social and economic complexity makes biased automation particularly dangerous. Second, the rapid global push towards AI regulation means organisations without proper governance will soon find themselves out of step with emerging compliance standards. 

 

“South Africa cannot afford a trust deficit in technology. If consumers or stakeholders believe AI reinforces old inequalities or operates opaquely, the damage will be lasting. Ethical AI is not a moral accessory; it is a business continuity requirement.” 

 

A framework for responsible AI in South African organisations 

Howland argues that businesses should urgently strengthen four areas of governance if they want to protect long-term value and maintain stakeholder trust: 

 

Transparency

AI-generated or AI-assisted outputs should be clearly disclosed to internal and external stakeholders. Transparent communication reduces reputational risk and aligns with emerging global standards. 

 

Data and bias auditing

AI systems must be trained and tested on data that reflects South Africa’s racial, linguistic, and geographic diversity. Regular audits should be mandatory to ensure models do not reinforce historical inequalities or embed unfair decision-making. 

 

Human oversight

Human decision makers must remain ultimately accountable. All AI-supported actions, from content production to risk scoring, should be vetted for accuracy, cultural alignment, and compliance with ethical and legal frameworks. 

 

Skills development

Teams need deeper fluency in both the capabilities and the limitations of AI. Without upskilling, organisations risk misusing tools, misunderstanding outputs, and missing early warning signs of algorithmic failure. 

 

“AI can transform how businesses operate, but only the organisations that prioritise governance, clarity and trust will see sustainable value,” says Howland. “At Alkemi, we work with clients to interpret and oversee their AI use, shape ethical communication frameworks, and guide responsible adoption. Our focus is on helping businesses set the right guardrails and communicate transparently so that stakeholders understand, trust, and support the role AI plays in their operations.” 

 

Why this matters for business media audiences

According to Howland, the organisations that treat AI governance as a strategic business issue now will be the ones that gain a competitive advantage as regulation catches up. 

 

“Boards want clarity, executives want capability, and consumers want trust,” he says. “The communications industry has a critical role to play in helping companies navigate this new frontier with intelligence, responsibility, and transparency.” 

 

He adds that as South Africa’s economy becomes more digitally dependent, ethical AI will influence capital decisions, brand reputation, regulatory compliance, and customer loyalty. 

 

“AI is a business tool with ethical consequences. If we do not address the governance gap, the cost will be measured not only in failed campaigns, but in damaged brands, unnecessary litigation, and eroded public trust.” 

AI in the Workplace: Why Human Judgment Still Matters

By Bradly Howland 

CEO, Alkemi Collective 

 

A few months ago, one of my team leads submitted a first draft of a client strategy document in under two hours. It was polished, well-structured, and covered all the expected ground. It was also, on closer reading, almost entirely generic. The AI had done its job. The thinking had not. 

 

That moment sits with me, because it captures something the industry is not yet being honest enough about. Artificial intelligence has not simply made us more efficient. It has changed what efficiency means… and in doing so, it has created a new and underappreciated risk: that we begin to mistake fluency for insight. 

 

The productivity gains from AI are real

The productivity numbers are real. Research from MIT and Harvard Business School consistently shows that AI-assisted knowledge workers complete tasks 20% to 35% faster, often with measurable quality improvements. Across the communications and marketing industry, output volumes have surged. Agencies that once needed teams to produce a content campaign can now do it with a fraction of the headcount. Clients are asking for more, faster, and at lower cost. The market is obliging. 

 

However, here is the argument I want to make and I accept it may be unpopular in some boardrooms: the productivity gains from AI are real, but they are also creating a form of output inflation that we are not yet equipped to manage. 

 

The Hidden Cost of AI Productivity

Output inflation works like this. As the cost of producing work drops, so does its perceived value. When content is abundant and cheap to generate, differentiation becomes harder. The premium shifts away from production capability and towards something far more difficult to automate: judgment, interpretation, and the ability to navigate genuine complexity. Yet the commercial pressure is moving in exactly the opposite direction. More deliverables. Faster turnaround. Shorter thinking cycles. 

This is not a technology problem. It is a leadership problem. 

 

Organisations embedding AI into their workflows are making implicit choices about what they value. If the metric is speed and volume, then depth and originality will be quietly deprioritised. Nobody announces this. It simply happens, gradually, as the incentive structure shifts. Over time, it hollows out the very capability that justified the premium in the first place. 

 

AI adoption in South Africa

In South Africa, this dynamic carries an additional layer of complexity. AI adoption is uneven - not just across industries, but within them. Some organisations are accelerating rapidly, deploying tools across research, content, and client service. Others are still building foundational digital infrastructure. The result is a dual pressure: to catch up quickly, often without the governance frameworks or strategic clarity needed to do so responsibly. 

 

The risk of moving fast without that clarity is not only reputational, it is also structural. An organisation that uses AI to replace thinking, rather than to enhance it, is not becoming more competitive. It is becoming more fragile, dependent on tools it does not fully understand, producing work that is difficult to defend when it matters. 

 

What AI means for communications

For the communications industry specifically, this should prompt some uncomfortable self-examination. Our value proposition has never been rooted in volume. It lies in the capacity to read a situation accurately, to shape a narrative under pressure, to exercise judgment when the facts are contested and the stakes are high. These are not capabilities that can be automated. But they can be eroded - through neglect, through underinvestment, and through the slow substitution of speed for thought. 

 

The organisations that will succeed in this next phase are not those that use AI most aggressively. They are those that are most deliberate about where it stops. 

 

That means protecting time for thinking, even when technology makes it possible to move faster. It means being explicit about where human judgment is non-negotiable, and building that distinction into how work is structured, priced, and evaluated. It means resisting the temptation to let a client brief become a prompt, and then mistake the output for a strategy. 

 

AI has made work faster. Whether it makes work better depends entirely on the choices that sit around it - the questions we ask before we run the tool, the standards we apply when we review the output, and the intellectual honesty to know the difference between something that looks good and something that is good. 

 

Speed is not the test. Judgment is. 

 

Howland is the CEO of Alkemi Collective and President of the Public Relations Institute of Southern Africa (PRISA).