AI (concepts of scaling laws and synthetic data)
Introduction
Key AI Concepts for Change Management
-
AI Scaling Laws
What are scaling laws?
Scaling laws describe how the performance of AI models improves as we scale up:
- More data
- More model parameters (size)
- More computer power
As these grow, AI becomes:
- More accurate
- Better at reasoning
- Better at understanding nuance and context
Think of it as: the more a model sees and learns, the better it predicts and supports decisions.
Application to change management
AI scaling laws enable:
- More reliable predictive analytics, eg, who will resist change and why
- Better sentiment analysis, eg, detecting morale shifts early
- Improved personalised communication, eg adapting messages to different employee groups
- Smarter simulations, eg modelling the impact of change decisions before rollout
Summary
Bigger models = deeper insight into organisational behaviour.
-
Synthetic Data
What is synthetic data?
Synthetic data is artificially generated data that mimics real data patterns without exposing private information.
Sources include:
- Simulations
- Generative AI (eg., GPT models)
- Behavioural modelling
It is:
- Privacy-safe
- Cheaper when data is limited
- Useful for training AI where real data is sensitive or unavailable
Example:
Simulating responses of employees to:
- A new HR system rollout
- A hybrid work policy
- Workflow redesign
You can create thousands of realistic behavioural profiles without surveying everyone.
How These Concepts Improve Change Management
| AI Capability | What It Does in Change Management | Benefit |
|---|---|---|
| 1. Scaling laws | Improve predictive models and behavioural analytics as more change data is collected | Better forecasting of resistance and adoption |
| 2. Synthetic data | Simulate employee reactions, stress points, network structures | Test strategies before real-world rollout |
| 3. Combined | Enable large-scale virtual testing of change plans | Reduce risk and refine interventions |
Practical Use Cases
| Phase of Change | Application |
|---|---|
| 1. Readiness | Generate synthetic data to model different leadership styles and communication strategies |
| 2. Engagement | Scale sentiment models to detect resistance pockets in real time |
| 3. Training | Create synthetic user workflows to train AI support tools |
| 4. Adoption | Stress-test new systems before launch to avoid disruption |
| 5. Sustainment | Simulate long-term effects of change on retention, productivity, culture |
Strategic Insight
AI's growing power (scaling laws) with simulated human behaviour (synthetic data) has the ability to predict and prevent failure in change initiatives.
Instead of reacting to resistance, leaders can design for acceptance from day one.
Summary
| Concept | Definition | Value for Change Leaders |
|---|---|---|
| Scaling Laws | Bigger AI results in better decision intelligence | Clear foresight into human and organisational dynamics |
| Synthetic Data | Artificial data that reflects real patterns | Safe, scalable experimentation on change strategy |
(main source: Ajay Agrawal et al, 2018)