New IT Bullshit for Old

Martyn Rhisiart Jones

Madrid, Wednesday 21 January 2026

In early 2026 the technology industry is once again telling big confident stories about its own future. These stories dominate earnings calls conference stages and investor decks. They sound transformative urgent and inevitable. Yet when examined closely many of them rest on fragile foundations and selective evidence rather than operational reality.

The first and loudest claim is that AI agents are on the verge of automating most white collar work. In practice agentic systems show promise only in narrow controlled tasks. Real world deployments remain brittle. They hallucinate struggle with edge cases require constant human supervision and often deliver marginal returns. The idea of synthetic workforces ignores the complexity of enterprise integration the difficulty of maintaining reliability and the fact that most organizations are still experimenting rather than scaling. What looks like inevitability feels more like another peak of inflated expectations.

Closely related is the assertion that artificial general intelligence is arriving within a few years. This narrative is repeated by lab leaders and amplified by capital markets but it clashes with persistent technical limits. Hallucinations shallow reasoning instability in multi step systems data constraints and rising compute costs remain unresolved. Polished demonstrations obscure how wide the gap still is between impressive prototypes and dependable general intelligence. The promise sustains investment but does not resolve the underlying problems.

A third claim concerns enterprise productivity. AI was meant to deliver rapid and dramatic returns across large organizations. Instead surveys consistently show that most initiatives fail or stall at the pilot stage. Productivity gains are incremental rather than transformative. The vision of ten times engineers and instant efficiency rarely survives contact with legacy systems governance requirements and integration debt. Meanwhile energy consumption compliance costs and organizational complexity continue to grow.

Even the infrastructure story shows signs of strain. Large technology firms have extended depreciation schedules for servers and accelerators making profits appear stronger and smoother. Yet frontier hardware loses relevance far faster than accounting assumptions suggest. In practical terms machines become outdated within a few years. Falling resale prices for used hardware already hint at oversupply and declining economic value beneath the surface.

Edge based AI is often positioned as the next decisive shift promising lower latency better privacy and reduced cloud dependence. In reality running models on devices delivers genuine benefits only in specific contexts. Hardware constraints power consumption battery drain model compression tradeoffs and fragmented ecosystems limit its reach. It complements centralized computing rather than replacing it.

Quantum computing follows its familiar cycle of anticipation. Technical progress continues and laboratories will announce milestones. Yet error rates qubit stability and scaling challenges keep useful commercial systems well out of reach. The language of breakthroughs persists even as practical impact remains limited.

Finally physical AI and humanoid robots have captured public imagination. Demonstrations suggest a future of automated warehouses and security teams. But deployment remains minimal costs are extremely high reliability is poor and regulatory and safety hurdles are substantial. What exists today is experimental hardware not an imminent transformation of blue collar labor.

Together these claims fuel vendor revenues stock valuations and conference enthusiasm. But the mood is shifting. In 2026 the tone is becoming more skeptical and more demanding. The question is no longer what AI might do someday but what it can demonstrably deliver now. The era of AI everywhere rhetoric is giving way to a simpler request… show the numbers.

Many thanks for reading.

_😺  Click for the last 100 Good Strat articles _😺