In theory, everything looked beautiful. Generative artificial intelligence was supposed to automate entire business processes, agent-based AI was supposed to conduct conversations with other agents without human intervention, and quantum computers were about to revolutionise computing. However, practice has shown otherwise – andCIOs are increasingly bold to admit: many of these technologies have not grown into their own legend.
A pattern is emerging from annual lists of the most overhyped technologies. Novelties that were still central to digital transformation strategies a year or two ago are now more likely to end up in the ‘high-risk experiments’ section. The reason? Implementation reality has not kept pace with the marketing narrative.
Generative AI: too much hope, too few implementations
For the third consecutive year, generative AI has topped the list of disappointments. Its potential – from process automation to creative content generation – is not in doubt. The problem is that the results of pilots rarely translate into production deployments. According to IDC data, up to 90% of such projects end prematurely. CIOs are therefore learning from their mistakes and increasingly focusing on niche, measurable applications instead of wide-ranging transformations.
In practice, this means a change in narrative – from ‘AI will revolutionise my business’ to ‘where AI can actually give us a return on investment’. There is also a growing awareness of the costs: both financial and time. Companies need to invest in surveillance, verification of results and integration with systems already in place.
Agent-based AI and ‘digital workers’: the illusion of autonomy
A new entrant to the roster are so-called AI agents – that is, software designed to autonomously make decisions, collaborate with other agents and independently achieve goals. While this sounds like a dream come true for self-service processes, the reality is much more mundane.
Gartner estimates that up to 40% of Agent AI projects will be abandoned by 2027 due to unclear business value and rising costs. The industry also suffers from a lack of clear definitions – many vendors simply ‘repaint’ existing chatbots and RPA tools with a trendy label.
In a similar vein, experts commented on the so-called digital workers. Their job was supposed to be to take over tedious administrative tasks and even manage processes. In reality, ‘digital workers’ are mostly well-dressed chatbots, performing very limited operations. General-purpose agents – able to analyse, plan and act independently – can be forgotten for the time being.
AIOps and observability: more data, less benefit
Artificial intelligence in IT operations (AIOps) was supposed to solve one of the biggest problems facing large organisations – an overabundance of operational data and a lack of context. Meanwhile, many companies are finding that the complexity of their deployments and the inadequacy of their telemetry means that instead of responding faster, they simply receive more alerts with no real meaning.
As a result, investment in AIOps often leads to… further investment – in teams to analyse the results of AI work. Paradoxically, it is the increased information noise from these systems that is today becoming a barrier rather than an aid to incident response.
Classic AI: expectations versus reality
Against the backdrop of these detailed analyses, another conclusion emerges: artificial intelligence itself as a technology category has also come under fire. CIOs note that a misconception of AI’s capabilities – as a ‘do-it-all’ solution – is leading to costly mistakes. It is not uncommon for companies that have attempted to reduce staff counting on AI support to now have to rebuild the team and trust.
Disappointments arise not from a lack of potential, but from over-simplified communication. Successful AI requires data, integration, adaptation and… patience. Contrary to popular belief, implementing AI is not simple, quick or cheap.
Quantum computers and the metaverse: a distant horizon
In the ‘future yet to come’ category, quantum computing is once again in the spotlight. Although companies such as IBM, IonQ and Google are developing the technology intensively, its real business applications remain a song of the future. For CIOs, this means one thing – planning for possible impacts (e.g. post-quantum encryption), but without a concrete implementation roadmap.
The metaverse and XR technologies (AR/VR/MR) have suffered a similar fate. Despite growing investments (Meta, Apple, Microsoft), CIOs still treat them as curiosities with potential but no clear business cases. The problem remains ergonomics, infrastructure cost and incompatibility of solutions.
Multicloud: a strategy without a strategy
Also on the list of over-hyped concepts is the multicloud strategy. Although many companies claim to use more than one cloud, real interoperability and flexibility are rarely achieved. Workloads are still tied to a single provider and attempts to move them often fail due to differences in architecture, security models and data transfer costs.
Multicloud in practice has become a collection of haphazard decisions rather than a well-thought-out strategy for independence from vendors.
Green energy and electric vehicles: disappointing the end user
Interestingly, some CIOs have decided to extend the list of disappointments to consumer technologies too – such as electric cars and green energy. Both the interfaces in EVs and the problems with the real-world range and variability of solar panel performance have caused some IT leaders to begin to approach the vision of a rapid, widespread energy transition with more distance.
The conclusions are analogous to those for strictly IT solutions: even the most laudable ideas have to face the reality of the end user – and this can be surprisingly difficult.
Maturity versus excitement
CIOs today are learning to recognise the difference between potential and market readiness. Technologies that are at the so-called ‘peak of inflated expectations’ (according to Gartner’s hype cycle) may be fascinating – but not necessarily useful in the here and now.
