AI Killing Season – maybe we’re not all doomed after all

A recent CIO opinion piece argues that 2026 may mark the start of an “AI killing season”, the moment when...

Andy Bristow

9 min read Reading Time

15 January 2026 Date Created

A recent CIO opinion piece argues that 2026 may mark the start of an “AI killing season”, the moment when organisations finally begin shutting down AI initiatives that cannot demonstrate real return on investment rather than continuing to celebrate pilots, proofs of concept, and shiny tools.

In the same breath, however, it repeats a familiar warning: upskill now or risk losing your job to AI.

Both claims are common yet they are increasingly hard to reconcile.

If AI is already delivering the kind of transformative productivity gains that justify widespread layoffs why are boards only now starting to question whether AI programmes are creating measurable value at all? And if most AI initiatives are about to be cut for lack of ROI, who exactly is being displaced?

The reality is that most corporate AI adoption remains shallow. While organisations talk confidently about “AI-powered” transformation, surveys consistently show that only a minority have embedded AI deeply enough to change core workflows, decision-making or operating models in a way that delivers sustained and measurable impact. In practice, AI is still most often bolted onto existing processes, marketing content, customer communications, software development etc while the harder work of redesigning how the organisation actually runs remains untouched.

That gap between AI rhetoric and AI reality explains far more about current layoffs, stalled programmes, and leadership frustration than the technology itself.

What the data actually shows

Surveys now suggest a clear gap between broad adoption and real transformation. Recent summaries indicate that roughly 70–80% of organizations use AI in at least one business function, yet only about one in five to one in four report adoption at an organizational or enterprise‑wide level.

A 2025 enterprise benchmark found that while more than seven in ten organizations have introduced generative AI, only a relatively small share have moved beyond pilots into fully embedded, scaled deployments across multiple functions. Many enterprises report using AI in three or so functional areas often IT, marketing, customer service or analytics rather than redesigning entire value chains or business models around AI capabilities.

Even among adopters, strong, measurable ROI is concentrated in a minority. 2025–2026 overviews highlight that most leaders report “some” benefit from AI, but only a smaller subset can point to clear, quantified gains tied to AI initiatives across the business. Executives still expect AI to transform 30% or more of work, yet detailed analyses show that actual usage is heavily skewed toward discrete knowledge‑worker tasks like drafting emails, generating content, and assisting with code, rather than fully re‑engineered processes.

Why adoption is superficial, not transformative

Several recurring structural barriers explain why so much adoption remains skin‑deep.

Data and infrastructure gaps: Updated surveys continue to show data quality, fragmentation, and governance as top obstacles: leaders report difficulty integrating AI with existing systems and cite poor data readiness as a central reason pilots fail to scale. Without clean, well‑governed, and connected data, AI tends to remain trapped in narrow tools or isolated teams rather than driving end‑to‑end process redesign.

Talent and capability constraints: Enterprise reports identify lack of employee AI skills, limited in‑house expertise, and weak change management capabilities as key barriers; organizations can buy models or SaaS features, but struggle to re‑architect workflows, incentives, and roles around them. Many firms rely on a small central team or enthusiastic individuals, which leads to pockets of innovation without the organizational capabilities required for scaled transformation.

Organisational inertia: A 2025 benchmark study of over 1,600 AI leaders describes a majority of organizations as “Builders” or “Climbers”: they experiment with advanced use cases, often running multiple AI apps or pilots, but fall short on foundational capabilities and integration into core operations. This pattern, numerous proofs‑of‑concept with little operationalisation, signals that experimentation is outrunning real operating‑model change, leaving AI layered on top of legacy processes instead of reshaping them.

What real transformation would look like

Compared with superficial adoption, genuinely transformative AI looks very different in practice. Leading surveys and predictions highlight that front‑runner organizations are redesigning full value streams such as the entire customer journey, order‑to‑cash cycle, or R&D‑to‑commercialization pipeline rather than just inserting generative AI into isolated steps like email drafting.

These transformations are associated with material, auditable improvements in metrics such as cycle times, error rates, revenue per employee and customer satisfaction, with clear links from AI‑enabled workflows to financial and operational outcomes.

Crucially, the investment pattern is different. Analysis of high‑performing adopters show that they devote a large share of their AI budgets to people, processes, data foundations and governance, not just to algorithms or infrastructure often treating change management and capability building as the primary work.

Prediction reports for 2026 suggest that as hype cools, more organizations will follow this model, shifting away from scattered pilots and toward top‑down, enterprise‑wide programmes with clear ownership, guardrails and measurable impact.

In other words, the gap between AI marketing and AI reality remains wide. The companies that close it are the ones willing to treat AI not as a veneer on existing practices but as a catalyst for rethinking how the business actually works. For jobs this suggests the demand is building up which could yet explode.

At Bristow Holland, we increasingly see that the real differentiator isn’t access to AI talent in the abstract but the ability to help organisations define what kind of person they actually need at this point in their AI journey and how that role should change as the work stabilises.

As the hype around AI gives way to harder questions about delivery, the technology matters less than the operating model and the people inside it. Get those right, and AI stops being a marketing story and starts becoming a source of real, measurable performance.