I have just wrapped up my tenure as the AI Hyperautomation Lead at KPMG to start my private venture. Having worked in consulting for more than seven years and in the corporate world for over 15 years, I’ve had the opportunity to see firsthand how technology is reshaping the way we work. Just a month ago, I architected and implemented a generative AI solution at KPMG for augmenting internal audit planning. As the project concluded, I found myself reflecting on how this shiny new technology is changing the way we think, work, and make decisions.
In writing this article, I want to be fully transparent—I leveraged ChatGPT Deep Search to build upon my initial premise: that over time, our cognitive ability for independent decision-making might suffer from overreliance on generative AI. I wasn’t sure if this was correct or not, so I engaged in a discovery exercise with ChatGPT, using it to explore broader thinking on the subject from people more skilled and experienced in cognitive science, automation, and AI ethics. The findings were illuminating and have shaped the discussion in this article.
This topic is deeply relevant to me on a personal level. As a non-native English speaker, I have found myself relying more and more on ChatGPT to refine my correspondence. While this has been helpful, I have also started to notice a growing feeling of inadequacy regarding my written English and eloquence. It made me wonder—could generative AI be subtly eroding our confidence in areas where we previously had competence? If this is happening at the individual level, what might the long-term impact be on organisations and industries that are integrating generative AI into their decision-making processes?
The Allure of Efficiency: Generative AI’s Early Wins in Consulting
It’s no secret why consultants have been quick to embrace generative AI tools. These systems (like GPT-4 and other large language models) can draft reports, analyse datasets, write code, and even brainstorm strategy options in a fraction of the time it once took a team of analysts. Initial productivity studies are striking – one experiment with over 700 consultants at Boston Consulting Group found that when professionals used GPT-4 for tasks within its capabilities, their performance improved by nearly 40% compared to peers without AI (Boston Consulting Group, 2023). In my consulting days, producing a client-ready presentation or market analysis often meant late nights cleaning data or fine-tuning text. Now, AI can generate a solid first draft of a slide deck or an industry overview in minutes, allowing consultants to focus on refining insights rather than starting from scratch.
Broader industry data echo these gains. McKinsey’s analysis of generative AI’s economic potential estimates an addition of $2.6 to $4.4 trillion in annual value through productivity improvements across sectors (McKinsey & Company, 2023). From banking to marketing, companies report that generative AI dramatically speeds up knowledge work. Early adopters in customer service saw junior agents handle inquiries faster and more effectively with an AI copilot, closing knowledge gaps and improving quality (McKinsey, 2023). In consulting, junior staff can use AI to quickly gather market research or generate a list of strategic options, tasks that might have taken them much longer on their own.
Such efficiency gains are a game-changer for businesses. Faster turnaround and higher throughput mean more value delivered to clients and potentially increased revenue. It’s no wonder consulting firms are weaving AI into their workflows – the promise of doing more in less time is simply too attractive to ignore. However, these early wins come with an important caveat: they assume AI is used appropriately, within its areas of strength. The same BCG study that lauded 40% performance boosts also issued a warning. When consultants pushed AI beyond its reliable frontier – essentially asking GPT-4 to tackle problems it wasn’t equipped to handle – performance dropped by 19 percentage points (Boston Consulting Group, 2023). Why? Participants tended to “switch off their brains and follow what AI recommends,” even when the recommendations were wrong. In other words, the efficiency allure can lull professionals into a false sense of security, where speed overtakes sense-checking. This finding is a microcosm of a broader issue that is only beginning to surface.
The Confidence Conundrum: Are We Losing Our Independent Judgement?
As reliance on AI grows, an unintended side effect is emerging: professionals may become less confident in taking independent action or making decisions without algorithmic support. In consulting – a field that prizes confident problem-solving – this is the confidence conundrum. Imagine a young consultant who has used AI to generate every client deliverable so far. Will they feel equipped to handle a high-stakes meeting or an ambiguous problem on the fly, without turning to an AI for guidance? Many leaders worry that the answer might be no.
Psychologists have a term for our tendency to trust automated solutions too readily: automation bias, meaning we give undue weight to AI outputs and overlook contradictory information or our own judgement (Parasuraman & Manzey, 2010). Essentially, the more we lean on AI, the more we may discount our inner voice. A recent Microsoft research survey of knowledge workers highlights this tension. The study found that while using generative AI does reduce the mental effort workers feel they need to expend, it also “encourag[es] over-reliance on AI, with confidence in the tool often diminishing independent problem-solving.” (Microsoft Work Trend Index, 2023). In other words, people become so confident in the AI that they lose confidence in themselves. This manifests as a reluctance to make decisions without first consulting the AI, or an impulse to always double-check with the AI even after one has a perfectly good answer from experience or intuition.
This reminds me of a recent post I wrote about using Microsoft Teams’ Speech Coach—while incredibly useful, I noticed how easy it became to rely on it instead of actively engaging in note-taking or refining my speaking skills organically. The tool provided real-time feedback, which helped improve clarity and pacing, but at times, I found myself more focused on the AI-generated alerts than on the conversation itself. This reinforced the same pattern of cognitive offloading that AI-driven tools can subtly encourage, making me more aware of how technology can shift the way we engage with our own skill development.
Embracing Imperfection: The Key to Maintaining Confidence
The long-term impact of generative AI on business will not be determined solely by the technology’s capabilities, but by how wisely we integrate it into our work. This calls for leadership and mindfulness. Executives and managers should recognise that initial productivity gains, while exciting, are not the whole story – they must also invest in their people’s cognitive development and create guardrails against overdependence. Professionals, for their part, should view AI as a powerful assistant, not an infallible oracle.
Reflecting on my own journey, I can see how generative AI has both empowered and subtly challenged my confidence. I initially turned to ChatGPT to refine my correspondence as I tend to be a perfectionist—often, to my own detriment. When I realised the side effects, I took a more intentional approach to embrace imperfection in my writing style, recognising that it was a small price to pay to preserve authenticity and mitigate the impact of overreliance on AI. This mirrors the broader organisational challenge: how do we use AI to enhance human capabilities without diminishing confidence?
As I step into my new venture – attaché x, I am committed to leveraging AI as a tool for amplification, not replacement. The future belongs to those who can harness AI’s strengths while keeping their human genius intact. The ultimate consulting insight may be knowing when to rely on the AI – and when to trust yourself.
