01 / Human
Human organizations
How people divide work, build trust, make decisions, and develop ways of working that stick.
Research project · Early days
What makes an organization work?
Great individuals don’t automatically make a great team. We’re exploring what does — with people, AI agents, and everyone in between.
See what we’re testingThree ways to organize01 / 03
Different kinds of teams. Surprisingly similar puzzles.
Smart individuals don’t automatically make a smart organization. So what does?
What we study
Different teams.
Same puzzles.
Who decides? Who knows what? What happens when things go wrong? With AI agents, we can now test questions like these in ways that weren’t practical before.
02 / Hybrid
Human × agent
What happens when some of your colleagues are agents? Where should judgment, control, and accountability sit?
03 / Agentic
Agentic organizations
Can teams of AI agents find effective ways to work, recover from mistakes, and improve as they grow?
Research approach
Build it. Test it.
Learn something.
We test ideas in controlled settings, watch what happens over time, then ask whether the lessons hold up with real teams. No magic formula — just better evidence.
Lab A
Organizational Wind Tunnel
Change a rule, rerun the task, compare the results. What happens when you tweak hierarchy, autonomy, or access to information?
Lab B
Artificial Enterprise
Give artificial teams budgets, customers, and unexpected problems. See whether their way of working holds up.
Lab C
Living Organizations
Find out what carries over to human and hybrid teams — and what doesn’t. People aren’t just slower agents.
Proposed first study
Does a “way of working” really change anything?
Start with four artificial teams doing the same work. Give each a different guiding principle: move fast, put customers first, check the evidence, or just get the job done. Then compare what happens.
We’re more interested in the trade-offs than a winner’s podium. Does speed create mistakes? Does caution pay for itself? Does anyone actually follow the rules?
Neutral baseline
Speed First
Customer First
Evidence First
A proposed experiment. Results have yet to be collected.
How we work
Good experiments beat good slogans.
We’ll share methods, measure outcomes carefully, and be clear about what we don’t know. An interesting result in a simulation isn’t automatically a lesson for people.
Get in touch
Curious? So are we.
We’re putting AOLab together. If you’re researching multi-agent systems, studying organizations, or experimenting with new ways of working, we’d love to compare notes.
aolab@oxford.ai