My top 3 expectations for AI in 2026
A forward-looking reflection on what we’ll likely notice when the noise fades
Let’s start with a thought experiment.
Imagine it is December 2026: artificial intelligence is no longer the headline topic in every executive meeting. The excitement has not disappeared, but it has matured. AI is present in daily work now, embedded in systems and routines, no longer constantly discussed, yet quietly influential.
Looking back at the year, we probably would not describe 2026 as a breakthrough year. Instead, it would stand out as the year when patterns became visible. Not all predictions would have played exactly as expected, but a few shifts would clearly define how AI reshaped work, decision-making, and responsibility.
This article is written from today, but with the benefit of imagined hindsight. Not to predict the future, but to sharpen what leaders should already be paying attention to.
Expectation #1
We would say AI didn’t take jobs, it redefined who carried judgment
If we were reflecting calmly at the end of 2026, mass unemployment would likely not dominate the narrative. What would stand out instead is a quieter, more structural change: the redistribution of judgment inside organisations.
Many organisations would look flatter, with junior layers thinner and middle layers more selective. Not because companies suddenly became ruthless, but because AI absorbed large parts of work that once justified those layers: drafting, coordination, first-level analysis.
Early signs already emerged in 2025. Elite MBA programs reported unprecedented placement challenges, not because employers stopped hiring, but because traditional entry points disappeared.(1) What recruiters once called the "career ladder" began to look more like a "career wall," where the bottom rungs simply vanished. Top business schools found their graduates hitting what some called a "silicon floor," a barrier where AI had already claimed the analytical work that once trained future leaders. Investment banks and consulting firms still hired, but they hired differently. The work that justified twenty analysts now required five, and those five needed different capabilities than their predecessors.
What remained scarce was not effort or speed, but judgment under uncertainty.
Looking back, we would likely recognise that AI did not reward people who simply “used the tools,” but those who could integrate AI into decisions responsibly. Three human capabilities would consistently separate those who gained influence from those who struggled to stay relevant:
Framing the problem before turning to AI, rather than outsourcing thinking itself
Interpreting and challenging AI output, instead of accepting it at face value
Owning decisions and consequences, especially when outcomes were ambiguous or risky
In practice, this reshaped how work was learned and evaluated. Entry-level roles shifted from manual execution to review, correction, and sense-making. Performance discussions moved away from visible activity toward decision quality and downstream impact. Fewer people “did the work,” but more people were expected to understand why it was done a certain way.
That shift raises an uncomfortable question: if AI removes much of the work that once trained people, where will your organisation now develop its next generation of judgment, and who is accountable for that design?
In hindsight, we probably would not say that AI eliminated work.
We would say it eliminated work without judgment.
This redistribution of judgment created conditions for the second major shift.
Expectation #2
We would realise the AI arms race peaked, and consolidation defined the year
From today’s perspective, the AI landscape feels noisy and overcrowded. New tools appear weekly, each claiming to transform how work gets done, but markets rarely stay this fragmented once scale, integration, and reliability begin to matter more than novelty.
Looking back from December 2026, the picture would likely feel calmer. Not because innovation stopped, but because the market settled into fewer, deeper layers.
Many AI startups would not have failed outright. Instead, they would have been absorbed, embedded, or reduced to features inside larger platforms. Competitive advantage would have concentrated around three forces:
Control of workflows, not just model capability
Access to proprietary data, not public benchmarks
Distribution at scale, embedded where work already happens
Microsoft's moves in late 2025 illustrated the pattern precisely. The company embedded Copilot more deeply into Word, Excel, Outlook, and PowerPoint, making AI assistance feel native rather than optional.(2) Then, having established dependency, Microsoft announced Office 365 business price increases effective July 2026.(3) Users gained capability, but lost negotiating position. The tool that once sat beside your work became the platform where work happened. Price increases followed naturally once switching costs rose high enough.
For people inside organisations, AI would feel less like a tool and more like infrastructure, with less experimentation and more dependence, and with fewer choices but higher switching costs.
In hindsight, we would likely point to:
AI agents embedded directly into enterprise systems
Fewer pilots, more long-term commitments
Procurement decisions treated like infrastructure bets, not innovation experiments
The lesson would be clear:
The AI race was never about who built the smartest model.
It was about who controlled how work actually flows.
Platform consolidation, in turn, accelerated the third transformation.
Expectation #3
We would say regulation didn’t slow AI, data sovereignty decided who could compete
Today, regulation is often framed as friction. From the perspective of late 2026, it would likely look more like a market-shaping force.
AI adoption did not slow, but access became uneven.
The decisive shift was not just rules about models, but rules about data: where it lives, who controls it, and under which legal regimes it can be used. Governments and large institutions increasingly treated data as strategic infrastructure, not a neutral input.
Airbus's 2025 decision to prepare a tender for European sovereign cloud providers marked a turning point.(4) The aerospace giant needed to move critical AI infrastructure to providers that could guarantee data never left European jurisdiction. What executives initially dismissed as regulatory theater became strategic necessity. When aerospace giants repositioned their data geography, the message traveled fast: proximity to power mattered less than control of data flows.
Defense establishments reinforced this shift. British defense chiefs issued alerts about foreign intelligence services potentially accessing sensitive organizational data through AI systems.(5) What began as theoretical risk became active threat assessment. Concerns about Chinese or Russian access to training data, model behavior, or usage patterns moved from speculation to procurement requirements. Security clearances started applying not just to people, but to the infrastructure where AI processed sensitive information.
Data sovereignty reshaped competition in subtle but powerful ways:
Companies operating across borders faced fragmented AI capabilities
Domestic players gained advantages through privileged data access
Global AI strategies became constrained by local regulatory realities
From a human perspective, expectations changed too. Customers demanded transparency when AI influenced decisions about them, employees expected clarity on how AI affected evaluation and opportunity, and boards discovered that accountability could not be delegated to algorithms or vendors.
Looking back, we would likely cite:
Mandatory disclosure of AI use in sensitive processes
Aggressive efforts to eliminate shadow AI
Boards held explicitly accountable for AI-driven outcomes
“The system decided” stopped being an acceptable explanation.
In hindsight, regulation did not kill innovation.
It removed irresponsible advantage and re-drew the boundaries of competition.
Looking beyond 2026: when AI stops being a phase
If this thought experiment holds, December 2026 would not feel like an endpoint, it would feel like a transition.
By then, most foundational AI companies would already exist. Breakthroughs would continue, but they would be incremental rather than explosive, and the era of constant surprise would give way to an era of integration, reliability, and constraint.
Organisations would gradually shift focus:
From speed to trust
From experimentation to governance
From raw capability to legitimacy
Regulation would no longer feel external or temporary, it would feel like infrastructure, something leaders plan around, not react to.
In that phase, advantage would no longer come from adopting AI faster than others.
It would come from being trusted, by customers, employees, regulators, and partners, to use it well.
Why this thought experiment matters now
The value of imagining December 2026 is not prediction, it is decision quality. If this is roughly the direction we are heading, the most important leadership questions today are no longer:
Should we use AI?
Which tool should we try next?
The harder, more valuable questions are:
Who becomes more valuable when AI is everywhere?
Where does accountability truly sit?
What kind of organisation are we designing for a world where AI is normal, not novel?
Because when we eventually look back on 2026, the difference will not be who adopted AI first.
It will be who learned how to live with it: deliberately, responsibly, and with judgment.
References
AI Journal (2025). “Stuck on the Silicone Floor: The AI Barrier Keeping Elite MBA Graduates Unemployed.” Available at: https://aijourn.com/stuck-on-the-silicone-floor-the-ai-barrier-keeping-elite-mba-graduates-unemployed/
The Verge (2025). “Microsoft Copilot chat comes to Outlook, Word, Excel, and PowerPoint.” Available at: https://www.theverge.com/news/822789/microsoft-copilot-chat-outlook-word-excel-powerpoint
Times of India (2026). “Microsoft to hike Office 365 business prices from July 2026: What users should know.” Available at: https://timesofindia.indiatimes.com/technology/tech-news/microsoft-to-hike-office-365-business-prices-from-july-2026-what-users-should-know/articleshow/125781890.cms
Cloud Computing News (2025). “Airbus prepares tender for European sovereign cloud.” Available at: https://www.cloudcomputing-news.net/news/airbus-prepares-tender-for-european-sovereign-cloud/
The Sun (2025). “Defence chiefs issue spy alerts over foreign intelligence concerns.” Available at: https://www.thesun.co.uk/news/37352851/defence-chiefs-spy-alerts/