For years, the battle for the enterprise desktop was relatively straightforward.
Microsoft owned much of it.
Email, documents, spreadsheets, presentations, collaboration — most employees spent their day inside a relatively small number of applications.
AI may change that.
Because if AI agents become the way employees increasingly get work done, an interesting question emerges:
Who owns the agent?
Think about how we may work a few years from now
Imagine starting your day and instead of opening ten different applications, you simply tell your AI agent what you need.
Prepare me for my 9 AM meeting.
Summarize what happened overnight.
Review this contract.
Analyze these numbers.
Follow up with the people who owe me something.
Create the presentation.
The agent goes into the appropriate systems, gathers the information, performs the work and brings the result back to you.
If that becomes reality, something important has happened.
The agent has become the interface between the employee and the enterprise.
And owning that interface could be enormously valuable.
Everyone has a different advantage
Microsoft already owns much of the employee's working environment. Copilot can sit across Office, Teams, Outlook, GitHub and Azure.
Google has Workspace, search, cloud and data.
OpenAI doesn't have that traditional enterprise software footprint, but it has ChatGPT and an enormous direct relationship with users. Its opportunity may be to make ChatGPT itself the place where work gets done.
Anthropic has taken another route, with Claude becoming particularly strong in software development and other knowledge-intensive work.
Then there are companies like Salesforce and ServiceNow.
They may argue that the agent shouldn't live primarily in the productivity suite or the chatbot.
It should live where the business process actually happens.
And they have a pretty good argument.
But enterprises may want something different
Having spent years running technology inside large enterprises, I suspect CIOs will be uncomfortable with the idea that one AI company owns everything.
Enterprises have spent decades trying to avoid excessive dependence on individual technology vendors.
AI isn't likely to change that.
A bank may use one model for coding, another for customer service and another for risk or research. Some models may run in the cloud. Others may run internally.
So perhaps the winning architecture isn't one agent that does everything.
Perhaps it is an enterprise AI layer that can use many models and many agents while the enterprise retains control of its data, identity, permissions and governance.
That would create a very different competitive landscape.
This battle may be bigger than the model battle
We spend enormous amounts of time comparing models.
Which one is smartest?
Which one is fastest?
Which one is cheapest?
Those things obviously matter.
But models will keep changing.
The more strategic question may eventually be:
Who controls the relationship between the employee, the AI and the enterprise?
Because whoever owns that layer doesn't just provide intelligence.
They potentially control how work gets routed, which models get used, which applications get called and where enterprise context lives.
That is an extraordinarily valuable position.
And I suspect Microsoft, Google, OpenAI, Anthropic, Salesforce, ServiceNow and quite a few others understand that very well.
The next great enterprise technology battle may not be about who has the best AI model.
It may be about who owns the agent that decides when and how those models get used.