FROM AI WORKFLOWS TO AI-NATIVE TEAMS

If you have AI working in production, how might your enterprise move closer to becoming AI-native?

For businesses that have moved beyond prototypes, digital intelligence is already helping with individual tasks and, increasingly, complete workflows.

Is the next logical focus on the team, building on what is already working?

A team could coordinate through agents, provided people and agents share an understanding of objectives, decisions and work in progress. The models would also need to be sufficiently capable for those agents to take on defined responsibilities, with clear authority and routes back to people.

In this way of working, people and agents work through problems together, with agents bringing relevant knowledge, exploring alternatives and carrying work forward, while people contribute experience, judgement and direction.

That collaboration could make more intelligence available across the team, expanding both its capacity to deliver and the range of ideas it has time to investigate.

When the team deliberately captures its decisions, corrections and working knowledge, agents could carry that context between tasks, making lessons from one piece of work available to others.

Colleagues across the business could then build on a broader pool of experience, while the original experts continue to challenge the agents’ suggestions and decide when an unfamiliar situation calls for a different approach.

Early examples strongly suggest arrangements like this are possible. What we don’t yet know is how widely they can be replicated – or what returns to expect. Still, they give us something concrete to investigate.

The test would be whether the whole team completes more work at the required quality, with full costs counted: compute, supervision and correction. Faster tasks could simply create longer queues elsewhere. We would need evidence that the team’s overall contribution had improved.

If the results held up, they could inform changes to neighbouring teams and eventually the department.

Yet those arrangements would remain provisional. Just as we begin to understand AI transformation, improving capabilities could change what an AI-native organisation might look like again.

Perhaps a useful ambition for 2027 is to choose one team, build on the digital intelligence already in its workflows and discover what organising around that capacity could make possible.

For more information on the expanding boundaries of AI – and how to govern them – take a look at our applied research website.

Our applied research explores how people and agents can work together, make expert knowledge more widely available and keep decisions accountable. You’ll find the research and open-source tools here: https://brightbeam.works/