Generative AI: A Powerful Tool, Not a Panacea

Key Takeaways

• Treat GenAI as a useful, but limited, tool not a business transformation panacea – success lies in understanding its constraints and applying it strategically rather than broadly.
• The true costs of implementing GenAI are often underestimated, extending beyond the per-query expense to include the need for specialised human oversight, data management, and continuous retraining.
• The full benefits of AI will only come with the expansion of AI capabilities beyond Generative AI.

AI Isn’t the Business Magic Bullet Everyone Thinks It Is

Artificial intelligence, or more specifically Generative AI (GenAI), has become business leaders’ newest obsession. Company boardrooms buzz with talk about ChatGPT, DeepSeek and similar tools. CEOs fear that AI-first competitors will outpace them. Investors flood any AI-mentioning startup with money. The technology generates human-like text and creates photorealistic images, leading many to believe it will transform everything from marketing to healthcare.

But this excitement is misplaced. GenAI writing tools, despite their impressive tricks, aren’t some sort of silver bullet that will, on its own, transform business the way supporters claim. Like the internet boom and blockchain, GenAI promises more than it delivers. Companies buying into the marketing hype and rushing to use them risk disappointment, wasted money, and sometimes real damage.

AI isn’t reliable enough

The biggest problem is that GenAI isn’t consistent. Well-designed traditional programs typically deliver consistent results: identical inputs usually produce identical outputs. Modern business relies on this predictability. GenAI writing tools don’t work like this; they make educated guesses based on patterns they learned from data. The same question can generate two completely different answers: one great and one terrible.

Imagine a bank using GenAI to decide who gets a loan. The system might approve someone one day and reject the same person the next day, even with identical information. This inconsistency makes GenAI unsuitable for critical business decisions.

Even worse, these systems confidently make things up; inventing facts, citing studies that don’t exist, or giving advice that could get users in legal trouble. We are used to normal software problems that usually cause things to break, but GenAI mistakes can look like expert advice. This unreliability is just one reason why GenAI’s true costs extend far beyond the advertised per-query pricing.

The real costs are much higher

Simple GenAI tools might cost pennies per question, but real business uses need much more. Companies need data experts to gather and clean information, AI specialists to adjust the systems, and people skilled in ‘prompt engineering’ (basically, knowing how to ask AI the right questions). These scarce experts command high salaries in a competitive job market.

Upkeep costs add up fast too. Rapid business changes quickly outdate GenAI systems, forcing companies to invest in expensive retraining that costs large enterprises millions annually. Unlike traditional software that becomes more cost-effective with scale, GenAI expenses increase with usage. Each question consumes computing power; each user escalates costs.

But the hidden costs go beyond money. Instead of eliminating work, GenAI often creates new types of time-consuming supervision. Staff must check every GenAI-written document. Developers must review every line of GenAI code. Experts must approve every GenAI suggestion. Companies hire expensive specialists not for creative work, but to supervise GenAI outputs.

GenAI has built-in limitations

Business leaders fundamentally misunderstand what GenAI can do. The technology is good at recognising patterns and copying styles but struggles with reasoning, understanding context, and real expertise. GenAI might generate functional code that harbours hidden flaws, triggering crashes months later. It might recommend plausible medical treatments while overlooking critical patient details. It might make investment advice based on old information while missing obvious market changes.

Fundamental flaws cause these problems. GenAI systems absorb biases, mistakes, and false assumptions from their training data. Biased training data can cause GenAI systems to perpetuate discrimination in hiring decisions. When training data contains outdated medical knowledge, systems provide dangerous health advice.

GenAI also can’t keep up with current events. While human experts can use this morning’s news in their analysis, GenAI is stuck with whatever it learned during training (often months or years old). In fast-changing areas like technology, finance, and regulation, this delay can be devastating.

New rules are coming

Governments are tightening oversight. The European Union introduced its AI Act in 2024, mandating transparency and human supervision for high-risk uses. Regulators may require companies to explain their AI decision-making processes, conduct regular audits, and have ways to override problematic systems. Governments worldwide are drafting similar legislation. Compliance costs will erode GenAI’s promised savings.

Financial regulators are paying attention too. The Bank of England requires AI banking systems to meet traditional software reliability standards. The Federal Reserve has started examining how AI failures could trigger widespread problems. Tighter oversight will replace early GenAI experimentation with careful, heavily regulated implementation.

A sensible approach

GenAI writing tools aren’t useless, however. The technology can genuinely help with specific jobs: writing first drafts of routine documents, brainstorming ideas, or summarising research.

But these uses are narrow and specific, not the broad transformation that enthusiasts promise. The smartest companies will treat GenAI like any business tool: with clear goals, realistic expectations, and careful measurement of results. They’ll start small, focusing on tasks where GenAI’s limitations matter least and benefits are clearest. They’ll invest in human oversight and keep the ability to go back to manual processes when the computer systems fail.

Thoughtful GenAI implementation, not maximum usage, determines business success. Smart companies will combine artificial intelligence with human judgement, using each for what it does best.

This careful approach lacks the excitement of AI-powered promises but delivers greater value: lasting competitive advantage built on solid foundations rather than wishful thinking.

The full AI revolution might be real, but it’s not as immediate or complete as supporters suggest, and GenAI is just an initial step. Other forms of AI, such as agentic AI and causal AI (more about these in subsequent posts), will address some of the stated concerns and add to the full AI tool kit.

Meanwhile, the old-fashioned strengths of human expertise, careful analysis, and sound judgement remain as valuable as ever. Businesses should remember this before betting their futures on systems that, despite their sophistication, still can’t tell truth from fiction.