Offloading work tasks to AI comes with a cost—to our brains

When a team employs generative AI to rush through a high-pressure policy paper, the results may appear polished, but they often contain hidden flaws. For instance, South Africa's Draft National AI Policy had to be withdrawn due to fictitious citations, while Deloitte faced a $440,000 refund for providing a document with fabricated sources. These incidents illustrate the hidden costs of offloading work tasks to AI.

The modern workplace culture typically prioritizes the final deliverable, such as a report or presentation, over the actual process of producing it. This outcome-driven approach can lead to a loss of understanding and critical evaluation skills when people rely too heavily on AI-generated outputs. Recent research indicates that over two-thirds of people use AI regularly, but more than half believe it improves their performance.

However, 61% of respondents have received no AI training, and many use it inappropriately, showing a lack of transparency and appropriate use.

Even the most advanced AI models struggle to produce work of professional quality, with around 85% of projects failing to meet acceptable standards. This reliance on AI for final deliverables strips away the quality control that comes from human involvement. Instead of banning or limiting AI usage in the workplace, organizations should adopt a more human-centric approach.

This includes setting realistic time frames for deliverables, disconnecting from AI tools during brainstorming and structuring phases, and cultivating a critical culture that questions the logic, structure, references, and assumptions behind generated work. Encouraging more human-led ideation, verification, and constructive critique will ensure that the true value of knowledge work—the clarity of thought and the iterative process—is preserved.