Machines That Sit Beside You

For a long time, software sat where we left it. It waited for instructions, completed a task, and disappeared back into the background. It was useful, but it was inert. What makes this new generation of AI tools feel different is not just that they are faster or more capable, but that they increasingly behave less like instruments and more like participants in work itself. 

That distinction matters a lot because people do not experience work as a sequence of isolated commands. Work is full of ambiguity, interruptions, context-switching, partial information, and half-finished thoughts. Microsoft’s 2025 Work Trend Index describes a workplace strained by what it calls a “Capacity Gap,” noting that 53% of leaders say productivity must increase while 80% of workers report lacking the time or energy to do their jobs, and employees are interrupted by a meeting, email, or ping every two minutes (Microsoft, 2025).  In that kind of environment, the most valuable AI tool is not the one that merely executes a prompt well. It is the one that can stay close to the flow of work and remain useful across moments, tasks, and decisions. 

That is why the best AI tools are beginning to feel like colleagues.

Not because they are human, and not because the metaphor is perfect, but because they are starting to occupy a role that software rarely did before. They summarize what matters, hold context across documents and meetings, surface relevant information at the right time, and increasingly help people move from intention to action instead of stopping at suggestion.  Microsoft frames this as the rise of “human-agent teams,” where intelligence is available on demand and work is increasingly organized around combinations of people and agents rather than fixed org charts alone (Microsoft, 2025). 

A good colleague does not simply answer when spoken to. They help you think. They reduce friction. They notice what is falling through the cracks. Increasingly, that is the direction AI tools are moving in. McKinsey argues that today’s models are becoming “human-like thought partner[s]” rather than mere engines for retrieval and synthesis, especially as reasoning and agentic capabilities improve (McKinsey & Company, 2025).  Microsoft Research has also explored AI systems that improve meetings not by replacing participants, but by helping them reflect on whether the conversation is tracking the group’s goals in real time, which is a surprisingly colleague-like function: less assistant, more thoughtful presence inside the work itself (Microsoft Research, 2025).

This also explains why people are becoming attached to certain tools while ignoring others. The tools that endure are not necessarily the flashiest ones. They are the ones that develop a kind of practical intimacy with work. They sit inside writing, planning, search, note-taking, coordination, and follow-up. They remember enough context to be helpful and become reliable enough to delegate to, at least in narrow ways.  The relationship is still instrumental, of course, but it begins to feel collaborative because the tool is no longer encountered as a one-off utility. It becomes part of how a person organizes thought and attention across the day. 

There is also a deeper reason this metaphor has taken hold. Work has changed. McKinsey found that employees are already more ready for AI than leaders tend to assume, with 94% of employees reporting at least some familiarity with generative AI tools, and employees self-reporting far greater day-to-day use than leaders estimate (McKinsey & Company, 2025).  The same report found that managers are regularly fielding questions about AI from their teams, and that 68% of managers said they had recommended a generative AI tool to solve a team member’s challenge in the past month, with most saying it worked (McKinsey & Company, 2025).  Once a tool starts functioning as shared workplace infrastructure rather than private experimentation, it stops feeling like optional software and starts resembling a co-worker everyone is learning how to work with. 

Still, calling AI a colleague should not flatter the technology too much. The metaphor is useful because it reveals how work is changing, but it can also obscure what remains stubbornly human. Colleagues carry judgment, accountability, social intuition, and consequences. AI does not. Microsoft itself makes this point indirectly when it argues that organizations will need to think carefully about the right “human-agent ratio,” especially in moments where customers prefer a human touch or where society expects a person to remain responsible for the outcome (Microsoft, 2025).  McKinsey, similarly, notes that trust, safety, accuracy, cybersecurity, and governance remain central barriers to mature deployment, even as adoption accelerates (McKinsey & Company, 2025). 

So perhaps the better way to put it is that the best AI tools feel like colleagues because work is no longer asking software merely to perform tasks. It is asking software to participate in cognition, coordination, and momentum. That does not make AI human. But it does make the old idea of software feel strangely inadequate. 

And that may be the development worth paying attention to.

Not that machines are becoming people.

But that work is being reorganized around systems that increasingly sit beside us, respond to us, and shape the pace and texture of thinking itself.