When Microsoft introduced the concept of the Frontier Firm in its Work Trend Index 2025, it described a new type of organization. The vision is built on “intelligence on tap” and hybrid teams of humans and AI agents.
Since then, the vision has become more concrete. Now the talk is of Frontier Transformation, where AI moves from individual tools and efficiency improvements to an integrated part of how the business functions.
– The interesting thing about the Frontier Firm is that the perspective shifts from what AI can do to how the company functions when AI becomes part of the workforce. It is a much larger question than implementing Copilot or a few AI agents, says Michel Hedenstjerna.
So what exactly is a Frontier Firm?
Microsoft describes a Frontier Firm as a human-led organization where AI and agents take an increasing share of the execution. Companies move from layering AI on top of existing ways of working to actually redesigning how work is performed.
– Many companies still use AI primarily to write, summarize, analyze, or find information faster. Frontier Firm is about the next step: What happens when AI doesn’t just help an individual employee, but starts performing parts of the company’s actual processes?
In that situation, the human role gradually moves from performing the work themselves with the support of AI to managing and orchestrating multiple agents that carry out different parts of a workflow.
From a system you work in to a system that works for you
This is where the development becomes particularly interesting from a business systems perspective. The change is clearly visible in how people interact with their systems.
– For decades, we have taught people how to work in business systems. Which menu should I go to? Which transaction should I make? Which fields should be filled in? Now we are approaching a world where the user instead says what they want to achieve and the system helps solve the rest.
The user then needs to know exactly where or how a task should be performed to a lesser and lesser extent.
– This doesn’t mean the business system becomes less important. Quite the opposite. If an AI agent is to be allowed to perform a business-critical task, there must be a stable core with correct data, business logic, permissions, and traceability. ERP does not become less important. What changes is how we interact with it.
Microsoft is Customer Zero
Frontier Transformation is also not just a vision for Microsoft’s customers. The company uses its own organization as “Customer Zero” and describes itself as being on the way to developing into a Frontier Firm.
– Microsoft is trying to make the same journey they are encouraging their customers to make. And an important lesson is that the big gains do not come from just layering AI on top of existing processes. You also have to dare to rethink how work should be performed when humans and agents work together.
Fundamentally, Microsoft’s Frontier vision is about more than the technology itself.
How much autonomy are we ready for?
Here, Michel sees a question that risks being overshadowed in today’s AI discussion.
– Today, many ask: Where can we use AI? Going forward, I believe an equally important question will be: Which parts of our business are we prepared to let AI perform independently?
An agent can be given a mandate to identify a deviation. The next step is to recommend an action. After that, it might be allowed to carry out the action within certain defined limits. In some processes, multiple agents can eventually handle larger parts of the flow, while the human is involved when something deviates or requires assessment.
– We are moving from designing processes solely for humans to designing processes for teams of humans and digital employees. The technology will likely move the boundary of what is possible faster than organizations move the boundary of what they are comfortable with.
Therefore, Michel believes that AI maturity in the future will be less about how many agents a company has and more about what mandates they are prepared and capable of giving them.
More autonomy requires more control
An AI that summarizes a document involves one type of risk. An agent that can access business-critical information and carry out transactions involves a significantly greater challenge.
Microsoft has therefore made security and trust a central part of the Frontier vision, with features to identify agents, limit their permissions, and monitor what they do, among other things.
– It’s really the other side of the autonomy question. The more an agent is allowed to do, the more important it becomes to know what data it can access, what mandates it has, and what happens if something deviates. Security and governance cannot be added as an afterthought. Autonomy and control must be developed in parallel.
And that, Michel argues, can become a better measure of AI maturity than how much AI a company uses.
– Maturity is not just about how much you dare to let AI do, but also about how good you are at setting the right boundaries for what it is allowed to do.
The competitive advantage is not AI
There is another consequence of this development. If companies in general gain access to the same models, platforms, and AI capabilities, the technology itself becomes difficult to use as a long-term competitive advantage. What sets companies apart may instead be what they do with it.
– If two companies use roughly the same AI models, it’s not obvious that the technology gives one a long-term lead. The difference may instead lie in the company’s own data, processes, and knowledge, and how quickly they succeed in translating that into a better way of working.
This is also why Michel argues that Frontier Firm should not be viewed as just another traditional technology project. There is no system to implement and then consider finished. Instead, it is about a gradual change in how the business works.
– You don’t implement a Frontier Firm. Rather, you gradually build the capability to work in a new way. The question is therefore perhaps not how far away this future lies, but what decisions we make today that determine how well-equipped we are for it.
According to Microsoft, a Frontier Firm is also about an organization that learns continuously, where experiences from the work are fed back into the system so that people, processes, and agents can gradually develop.
– It’s about more consciously deciding what humans should be best at, what AI should be best at, and how the two together can create a business that we couldn’t have built before.
