01 / Learn
Capture learning
Turn each deployment's successes and failures into knowledge the next team can actually use.
AOLab research / XC · Working title
Can organizations get better at getting better?
What if every successful transformation made the next one easier, faster and more valuable? XC is a research programme investigating whether we can deliberately improve the returns from organizational change.
XC-01 research protocol v0.1 drafted · Field experiment not yet underway
Most transformation programmes measure whether one change worked. We want to know whether it makes the next one work better.
The question
Move the exponent.
Some organizations experience diminishing returns as they add tools, teams and complexity. Others seem to learn, connect capabilities and improve their ability to change. Can we identify the mechanisms — and influence them?
We use a simple power-law model as a hypothesis to test, not a universal law.
The difference an exponent makes
Value = k × ScaleC
Here C describes how value changes as the effective scale of adoption increases. Move the slider to see how the shape changes.
0.51.01.5
Increasing returnsValue grows more than proportionally with scale.
Proportional (C = 1)Selected exponent
Illustrative curves using a shared coefficient, shown on fixed axes. This is not observed research data. A power law with C > 1 is superlinear scaling, not necessarily exponential growth over time.
Four candidate mechanisms
What could make returns improve?
If every implementation starts from scratch, transformation remains expensive. XC investigates whether four deliberate practices can change the economics of what comes next.
02 / Reuse
Build on what works
Reuse validated agents, workflows, integrations and adoption practices instead of reinventing them.
03 / Connect
Combine capabilities
Test when new capabilities increase the value of those already in place.
04 / Improve
Reduce friction
Remove recurring obstacles without weakening the controls that make transformation safe.
From adoption to increasing returns
Prove one change. Improve the next.
An adopted workflow and a transformation system that compounds value are not the same thing. XC connects them, while measuring each separately.
Level 1 / Prove · 90 days
Make a new way of working real
Establish routine, unprompted use; displace the old way; verify the agreed business outcome. One successful implementation is evidence of adoption, not of superlinear scaling.
CurateCommitGo-liveUseDisplaceResult
Level 2 / Compound · Across deployments
Make the next implementation better
Carry learning and reusable capabilities forward. Connect adjacent workflows. Measure whether time, cost and value improve across comparable implementations.
LearnReuseConnectImprove
First proposed study
Protocol drafted · Seeking collaboratorsXC-01: The Compounding Adoption Experiment
Our proposed pilot would compare teams using the same 90-day adoption discipline, with one group additionally receiving structured learning, reuse, complementarity and friction-reduction interventions.
The first goal is to test feasibility and mechanisms, not to declare that we've proved C > 1.
What we'd measure
- Time to sustained, meaningful adoption.
- Fully loaded implementation cost.
- Verified operating outcomes and business value.
- Reuse, reliability and quality of integrations.
- Whether gains persist beyond the initial team.
The XC-01 research protocol is at version 0.1. The field experiment has not yet started; no results are claimed.
Request the research protocolBeyond human organizations
What happens when the teams are agents?
What changes when organizational knowledge can be shared between agents, workflows reproduced cheaply and coordination increasingly automated? Which new bottlenecks appear?
XC offers a way to investigate these questions across human, human × agent and predominantly agentic organizations.
The applied methodology is being developed by Implementing Intelligence. AOLab investigates the mechanisms, alternative explanations and evidence — including findings that challenge the hypothesis.
You don't have to believe the hypothesis. A willingness to test it is enough.
Get involved
Help us test the idea.
We're looking for researchers and organizations with repeatable AI-enabled workflows, comparable teams and credible performance measures. An interesting objection is as welcome as an interesting opportunity.
aolab@oxford.ai ↗