Why companies are increasingly comparing the cost of labour with the cost of computation
In 2026, companies began comparing not only employees’ salaries with one another, but also the cost of human labour with the cost of computation.
Not long ago, this idea would have seemed radical. Today, it increasingly shapes investment decisions, organisational design and management practices.
Over the past two weeks, several developments have highlighted how profound this shift has become. OpenAI created a new structure to deploy AI in businesses, SAP restricted the use of external AI agents, major companies openly linked workforce reductions to intelligent systems for the first time, and leading financial institutions handed over more responsibilities to AI.
At first glance, these stories appear unrelated. Yet together, they point to the same underlying transformation.
Artificial intelligence is no longer a standalone tool. It is becoming part of the operational infrastructure of organisations. The question is no longer how well AI performs individual tasks. The real question is how it changes the way businesses are governed.
1. From Models to Implementation
One of the most revealing developments of recent weeks was OpenAI’s launch of The Deployment Company (DeployCo), a venture valued at around USD 10 billion. Its purpose is not to sell licences, but to integrate AI into the operational processes of thousands of businesses.
For the past two years, competition focused on model quality. Today, it is becoming clear that the key question has shifted. The challenge is no longer the models themselves. The challenge is implementation.
How do organisations embed intelligence into real processes? How do they adapt internal controls? How do they measure economic value?
Technology companies are increasingly competing not only with one another, but also with traditional consultants by offering practical pathways for business transformation rather than technology alone.
The winners will not be those with access to the best models. They will be those who can embed intelligence into their operations while maintaining control over the transformation that follows.
2. From Record-Keeping to Access Control
Almost unnoticed outside specialist circles, SAP restricted the use of external autonomous AI agents within enterprise systems. At the same time, Anthropic expanded Project Glasswing while keeping access to its most sensitive capabilities under controlled conditions.
Together, these developments point to a broader transition. Enterprise systems that once focused on recording and processing information are gradually becoming systems of access control.
As AI capabilities expand, questions of governance become increasingly important. Who has access to data? Who can initiate actions? Who is accountable for the outcomes?
The discussion is slowly shifting from choosing the right model to a more fundamental question:
How can organisations maintain control over data, processes and business-critical decisions?
This is where cybersecurity and AI governance become part of mainstream business risk management.
3. From Automation to a New Organisation of Work
According to Challenger, Gray & Christmas, companies have started openly linking workforce reductions to AI adoption for the first time. At the same time, JPMorgan is preparing to deploy autonomous AI agents in private banking and analytics, while Vanguard already uses machine learning in funds managing approximately USD 13 billion in assets.
These developments are often interpreted as evidence that AI is replacing people. Meanwhile, headlines about astronomical spending on AI have become almost routine: new data centres, massive investments in chips and billions committed to computing infrastructure.
However, the more important conclusion may be different.
In 2026, companies began comparing not only employees’ salaries with one another, but also the cost of human labour with the cost of computation.
This represents more than the automation of individual tasks. It signals a fundamental redesign of organisational models.
For decades, companies compared labour costs across departments, countries and markets. Today, another variable has entered the equation: the cost of computation.
This does not mean the end of human work. But it may signal the end of traditional organisational models.
The question is no longer simply which activities can be automated. It is about how work should be redistributed between people and intelligent systems, which capabilities become most valuable, and how accountability should be organised.
AI increasingly takes over analysis, monitoring and the preparation of recommendations, while people focus on relationships, complex judgement and responsibility for outcomes.
The future will not belong to organisations that reduce headcount the fastest. It will belong to those that learn how to combine human capabilities with intelligent systems while maintaining control over their organisation.
What connects these developments?
Taken together, these signals reveal a broader shift.
If the defining question of 2024 and 2025 was, “What can AI do?”, businesses in 2026 are beginning to ask different questions.
How should AI be integrated into existing processes? How can organisations retain control over access to data? How should roles evolve? Which accountability mechanisms are needed? How can businesses remain governable in this new environment?
This is why the most valuable conversations about AI today are no longer about technology.
They are about management.
What does this mean for business?
Organisations should start not by choosing a model, but by identifying the processes that genuinely require change. Rules governing AI access to corporate data should be established before large-scale deployment begins, and human oversight should remain in place for critical decisions.
At the same time, companies need to rethink employee roles, shifting the focus from routine execution towards control, analysis and judgement. Audit trails and accountability mechanisms should be built before autonomous agents become part of everyday operations.
Conclusion
Perhaps this is what 2026 will ultimately be remembered for. Not another model release. Not another investment record. But the moment when companies began comparing the cost of human labour with the cost of computation — and were forced to rethink what it means to govern an organisation.
For the past thirty years, businesses have built systems of control around employees. In the decades ahead, they will likely need to build them around employees and intelligent agents at the same time.
This is where the line between adopting a technology and truly transforming a business lies.
And this is where the conversation about governance, risk and the new architecture of business begins.
