In the rush to deploy AI, most organisations have been asking the wrong question.
By Martin Pienaar, chief operations officer at Mindworx Academy
The debate has centred on which platforms to use and how quickly engineering teams can ship. The harder question, and the one that will separate the winners from the cautionary tales, is who in your business is actually equipped to think clearly about what AI should do, verify that it has done it correctly and take responsibility when it hasn’t?
The answer may already be on your payroll. And it is not a Data Scientist.
The Cost of Getting AI Wrong
The cost of deploying AI without adequate human oversight is becoming impossible to ignore. In 2025, Deloitte was forced to partially refund a $290,000 Australian government contract after a single academic researcher discovered that their report contained fabricated court quotes and references to non-existent academic papers.
In South Africa, the Minister of Communications was forced to withdraw the country’s draft AI policy recently because a journalist identified that it contained fictitious citations. The culprit was unverified generative AI. It took diligent outsiders to catch what both a Big Four firm and the government communication department’s internal processes had missed entirely.
These are not isolated incidents. Estimates suggest that global business losses attributed to AI hallucinations, instances where AI confidently invents facts, reached as much as $67.4 billion in 2024. Microsoft’s 2025 Work Trend Index found that knowledge workers now spend an average of 4.3 hours per week verifying AI-generated work, roughly $14,200 per employee per year in verification overhead alone.
The lesson is not that AI is unreliable. It is that AI is an amplifier and is only as good as the thinking you put into directing it and the oversight of its results.
The Missing Skill Set Is Already in Your Organisation
Good AI output is not the product of clever conversational prompts. It is the product of precise, structured instructions that define context, constraints, output format and success criteria. That is a requirements specification by another name, and writing those is what Business Analysts have been doing for decades.
The difference is stark. Asking an AI agent to “summarise our customer feedback” produces generic results. Instructing it to “analyse Q1 2026 customer feedback, identify the three most frequently cited service failures and present findings as a prioritised list for the executive team, excluding pricing-related responses” produces actionable insight. One is a wish. The other is a brief.
Business Analysts are trained to do exactly this kind of work: translate ambiguous business intent into structured, testable, unambiguous instructions. Swap “developer” for “AI agent” and the craft is almost identical.
They are also the natural candidates for the “Human-in-the-Loop” governance roles that mature AI deployments require. They design feedback loops, stress-test outputs against business requirements and intervene where bias, error or risk has crept in. Before a line of code is written, they define the workflows these systems will execute. Once deployed, they provide the human check that keeps decisions accountable.
Judgement Becomes the Scarce Resource
As AI absorbs more routine task execution, the value of the modern professional shifts decisively to judgement. AI can process data at scale. It cannot interpret context, build alignment across stakeholders, challenge a flawed assumption in a steering committee or weigh the real-world consequences of a decision on customers, staff or regulators.
Those are not technical skills. They are business analysis capabilities, and in an AI-first operating model they stop being support functions and become strategic ones. The organisations getting this right are not replacing their Business Analysts with Prompt Engineers. They are retraining them to be Prompt Engineers, AI product owners, model validators and governance leads, because the underlying mindset already fits.
The Advantage Is in Recognising What You Have
The temptation, when AI enters the conversation, is to believe the answer lies in hiring entirely new teams with exotic titles. For most employers, it does not. The advantage lies in recognising and developing the expertise that already exists: professionals trained to translate complexity into structured decisions, challenge assumptions and ensure that outputs stand up to practical scrutiny.
AI does not reduce the need for that kind of thinking. It raises the price of not having it.
The companies that move now, by investing in their Business Analysts as the human layer of their AI stack, will build an advantage that their competitors, still shopping for tools, will struggle to match.