Deloitte’s recent State of AI in the Enterprise 2026 study clearly shows this shift. Employee access to AI increased by 50% in a single year, and organizations expect to move significantly more AI projects into production in the coming period. At the same time, only 34% of organizations are using AI to fundamentally transform their organization.
That creates an interesting contrast.
AI is becoming more widely available, but that does not automatically mean organizations are fundamentally changing the way they work.
From experimentation to integration
The first phase of generative AI was relatively straightforward. Employees gained access to tools that could generate content, analyze information, write code, or speed up repetitive tasks.
This can quickly deliver productivity gains.
According to Deloitte, 66% of surveyed organizations now report improvements in productivity and efficiency. 53% see better insights and decision-making, while 40% report cost savings.
But once AI needs to go beyond supporting individual employees, the challenge changes.
AI then needs to become part of processes, systems, and decision-making.
An employee using AI to create a document faster is very different from a process in which information is automatically collected, assessed, processed, and passed on to other systems. And a standalone AI assistant is very different from software where AI is actually embedded into the workflow.
That is where the next phase begins.
Making an existing process faster is not always the best solution
Many organizations understandably start with the way they already work.
They look at existing processes and ask where AI can save time.
That can be valuable, but the risk is that you simply make an existing process faster.
The more interesting question is: How would we design this process if AI, automation, and modern software had been available from the start?
That can lead to a very different solution.
A process involving several manual checks, for example, might be largely automated. Information that employees currently collect from multiple systems could be made available centrally. A manual analysis could become continuous. And software could do more than simply present information to employees: it could also recommend the next action.
AI then shifts from being a tool to becoming part of the process itself.
Deloitte is already seeing this shift. 30% of surveyed organizations are redesigning key processes around AI, while 34% are going further and using AI for deeper transformation, for example by developing new products, services, processes, or business models.
At the same time, this means that a significant share of organizations are still applying AI without fundamentally changing the underlying processes.
AI places greater demands on the software around it
As AI becomes more deeply embedded in processes, the quality of the underlying software becomes increasingly important.
After all, AI needs something to work with.
Data needs to be available and reliable. Systems need to communicate with each other. Roles and responsibilities need to be clear. Integrations need to work reliably. And organizations need to determine which decisions AI can make independently and where human oversight remains necessary.
An organization with many disconnected systems, manual handovers, and difficult-to-access data can use excellent AI models and still struggle to create sustainable value from them.
That is why AI transformation is also a software and architecture challenge.
It is not just about selecting a model or AI platform, but about the environment in which AI needs to operate.
The more important AI becomes within a business process, the more important the software, data, and architecture around it become.
Agentic AI makes this dependency even greater
This becomes even more relevant with the development of AI agents.
Where generative AI primarily generates information or supports users with individual tasks, agents can independently execute multiple steps and initiate actions within systems.
Deloitte expects the use of AI agents to grow significantly in the coming years. At the same time, only one in five surveyed organizations currently has a mature governance model for autonomous AI agents.
That gap between technological capabilities and organizational readiness matters.
An agent that only analyzes information has a very different risk profile from one that modifies customer data, processes an order, starts a workflow, or independently controls other systems.
The more autonomy software is given, the more important architecture, security, logging, authorization, data quality, and human oversight become.
The challenge is shifting from AI to the organization around it
One of the most striking conclusions from Deloitte’s research may be that organizations increasingly consider themselves strategically prepared for AI, while operational readiness is lagging behind.
42% now consider their AI strategy to be well prepared. At the same time, organizations are less confident about their readiness in areas such as infrastructure, data, risk, and talent.
That is understandable.
Coming up with AI use cases is becoming easier. The technology is accessible, and new possibilities emerge almost every day.
The complexity begins when such an idea needs to become part of an existing organization.
Which systems are involved? What data is required? How do we integrate the solution? Who remains responsible for decisions? How do we secure the process? And how do we make sure the solution is still manageable two or three years from now?
Ultimately, these are the same questions that have always mattered in good software development. AI simply makes them more urgent.
From AI project to software transformation
The next phase of AI adoption will therefore probably be less about adding yet another AI tool.
The bigger opportunities lie in rethinking processes and determining where AI, automation, and software together can enable a fundamentally better way of working.
Sometimes that means extending an existing system. Sometimes it requires new integrations or a different architecture. And sometimes you may conclude that a process is so specific to the organization that custom software provides the most logical foundation.
AI then stops being a separate layer on top of the organization. It becomes part of how the organization operates.
That also requires a different approach to software development. As AI accelerates processes, enables more automation, and becomes increasingly capable of performing tasks independently, analysis, architecture, security, development, and testing need to evolve with it.
At Infodation, we therefore look not only at what we can build with AI, but also at how AI changes the way we build software. Within our New Way of Working, we combine AI, automation, and modern software development to move faster from a business challenge to a working solution, without losing control over architecture, security, and quality.
Technology is changing rapidly. But ultimately, the most important question is not which AI tool you use.
The more interesting question is how you would design your processes and software today if you could start again with the capabilities AI now gives you.
Source: This article is partly based on insights from Deloitte, The State of AI in the Enterprise 2026: The Untapped Edge, a global study of 3,235 senior business and IT leaders across 24 countries.
