Op-ed: Is it too early to appreciate AI?

Artificial Intelligence Appreciation Day, marked on 16 July, arrives at an awkward moment. AI is everywhere, yet it is also unfinished. It is astonishingly capable in some settings and strangely brittle in others. It writes, summarizes, codes, searches, predicts, detects, recommends and automates; but it also fabricates, misclassifies, reinforces bias, leaks sensitive information, and creates fresh governance problems for organisations that have barely finished adopting cloud computing. The question, then, is not whether AI deserves attention. It clearly does. The question is whether it deserves appreciation — or whether a more suitable theme would be AI accountability.

The observance itself is commonly presented as a day to recognise the role of artificial intelligence in daily life and to promote discussion about ethics, regulation and responsible use. Several public calendars list AI Appreciation Day, with recent descriptions emphasising both celebration and governance rather than uncritical enthusiasm. That balance matters. AI has moved beyond novelty. It is now embedded in search engines, logistics systems, fraud detection, medical imaging, customer service, translation tools, software development, marketing platforms and workplace productivity suites. Yet the extent of that embedding often exceeds the maturity of organisational controls.

This is the concern raised by Poonacha Kongetira, co-founder and CEO of Classie, who argues that AI appreciation should also be a reminder to use the technology responsibly. His warning, sent to Digital Journal, about “shadow AI” is particularly apt. Many employees are already using generative AI tools informally, outside approved systems, policies or audit trails. The problem is not only that staff may paste confidential information into public tools. It is also that disconnected AI use fragments knowledge, creates inconsistent outputs, and makes it difficult for organisations to know which information, prompts, assumptions or models contributed to a decision.

This is why the phrase “trusted knowledge ecosystem” is more than a piece of corporate language. AI is only as useful as the information environment in which it operates. If company data are duplicated, outdated, poorly governed or trapped in departmental silos, AI will not magically produce clarity. It will accelerate confusion. This point is echoed by Ravi Achanta, founder and CEO of RSA America, who notes that smaller retailers can use AI to gain the sort of business visibility previously available mainly to larger chains, but only if customer, promotion, e-commerce and loyalty data are connected. In a fragmented environment, AI does not solve the silo problem; it amplifies it.

Does it pay to be cautious?

The workplace evidence supports a cautious view. McKinsey’s 2025 analysis found that nearly all companies were investing in AI, and 92 percent planned to increase investment over the following three years; yet only 1 percent of leaders described their organisations as mature in AI deployment, meaning AI was fully integrated into workflows and producing substantial business outcomes. This is the uncomfortable gap in today’s AI market: widespread adoption is not the same as business transformation. It is relatively easy to buy tools, run pilots, and encourage experimentation. It is much harder to redesign workflows, validate outputs, train staff, govern use cases, and measure whether AI is creating value rather than merely activity.

There is also evidence that AI is reshaping work, although not always in the straightforward “robots replace people” sense. Stanford’s 2026 AI Index describes AI capability as accelerating and reaching more people, with organisational adoption reaching 88 percent and four in five university students using generative AI. It also highlights the “jagged frontier” of AI capability: advanced systems can perform extremely well on complex scientific, mathematical or coding benchmarks while still failing on apparently simple tasks, such as reliably reading analogue clocks. For employers, this means AI is neither a toy nor an all-purpose substitute for human judgement. It is a powerful but uneven tool that changes the distribution of work tasks.

The effect on employment is therefore likely to be uneven. AI can reduce repetitive administrative effort, accelerate software development, improve document analysis, support customer service, assist scientific research and enhance decision-making. But it can also deskill roles if employees become passive reviewers of machine-generated outputs, or if organisations remove human expertise before they understand the limits of the technology. The main risk is not simply job loss; it is a degradation of accountability. If no one understands how a recommendation was produced, who owns the decision?

This concern is apparent in the comments from Binny Gill, founder and CEO of Kognitos, who compares AI Appreciation Day to “throwing a party for fire.” Fire is useful, but it is not something to applaud without controls. It requires respect, containment and a smoke detector. The analogy is persuasive because previous general-purpose technologies — electricity, aviation, automobiles, pharmaceuticals — became socially useful through standards, testing, regulation and professional practice. AI should be no different. Indeed, Gill’s suggestion of “AI Accountability Day” may be closer to what organisations need: appreciation for systems that can be audited, that ask before they act, and that keep humans in the loop.

Balancing the outcomes: Performing risk analysis

That framing aligns with the direction of formal risk management. The U.S. National Institute of Standards and Technology’s AI Risk Management Framework is designed to help organisations manage risks to individuals, organisations and society, and to incorporate trustworthiness into AI design, development, use and evaluation. NIST’s framework emphasises characteristics such as reliability, safety, security, accountability, transparency, explainability, privacy and fairness. These are not optional extras. They are the conditions under which AI becomes acceptable in business-critical contexts.

Industry-specific AI may also prove more valuable than general-purpose systems. Robin Gilthorpe, CEO of Earnix, makes this point in relation to insurance. A general-purpose AI system may generate fluent answers, but underwriting, pricing and customer decisions must reflect regulation, portfolio performance, business strategy and customer context. In other words, language competence is not the same as domain competence. This is a crucial distinction. For regulated sectors such as insurance, banking, healthcare and pharmaceuticals, AI needs process knowledge, auditability, validation and context-specific controls. The future is unlikely to be dominated by one generic chatbot sitting across the enterprise. It is more likely to involve specialised AI embedded into defined workflows.

Cybersecurity reveals the dual nature of AI most sharply. Chance Caldwell of Cofense notes that AI has become a force multiplier for defenders and attackers alike. Security teams can use AI to analyse patterns, automate workflows and scale detection. At the same time, attackers can use AI to produce more polished and personalised phishing campaigns. The old signs of malicious emails — poor grammar, generic wording and obvious errors — are becoming less reliable as AI-generated messages become more convincing. This creates a more dangerous threat landscape, where speed, scale and plausibility are all improved by the same tools defenders are trying to deploy.

This does not mean humans become irrelevant. On the contrary, in cybersecurity the employee may become more important as a sensor. AI can detect anomalies, but humans understand context: whether a request is unusual for that supplier, whether a payment instruction feels irregular, whether a QR code appears in an unexpected workflow. The best future for cyber defence is not AI replacing people; it is AI strengthened by trained people. This has broader relevance across all sectors. The organisations most likely to benefit from AI will not be those that simply automate aggressively. They will be those that combine machine speed with human judgement.

Time for AI appreciation?

Yet, there are reasons to appreciate AI. It can expand access to expertise, remove drudgery, improve accessibility, support medical and scientific discovery, help small businesses compete, and produce faster insights from complex data. It can be a productivity tool, a creativity tool, a diagnostic aid and a decision-support system. But appreciation should not mean naivety. AI is not good enough to be trusted blindly. It is good enough to be taken seriously.

So, is it too early to appreciate AI? No, provided appreciation is mature. We should appreciate the engineers, researchers, data scientists, ethicists, security teams and frontline workers who are making AI useful. We should appreciate systems that are transparent, validated, monitored and governed. We should appreciate AI when it improves human work rather than obscuring responsibility.

If AI Appreciation Day becomes is simply a marketing event, it misses the point. The correct posture is neither worship nor panic. It is informed respect. AI deserves appreciation when it earns trust and it earns trust through evidence, accountability and responsible use. July 16, 2026 should therefore be less a celebration of artificial intelligence as a product, and more a checkpoint for asking whether our organisations are intelligent enough to use it well.