Change Fitness: Preparing Your Organisation for Continuous AI Evolution

Agentic Business Design

23 March 2026 | By Ashley Marshall

Harvard Business School’s 2026 outlook introduced a concept that every business leader should internalise: change fitness. It is the organisational capacity to absorb, adapt to, and benefit from continuous change, without burning out or losing coherence.

Why Traditional Change Management Fails for AI

Traditional change management assumes a stable end state. You plan the change, execute it, embed the new way of working, and move on. The Kotter model, ADKAR, Lewin’s unfreeze-change-refreeze: they all assume you eventually refreeze.

AI does not refreeze.

Consider what has happened in the past 12 months alone:

• Reasoning models went from experimental to production-ready
• Multimodal capabilities expanded from text and images to video, audio, and real-time interaction
• Agent frameworks matured from demos to deployable systems
• Token costs dropped by 80 to 90 percent for comparable capability
• Open-weight models closed the gap with proprietary frontier models

Each of these shifts invalidated assumptions that organisations made just months earlier. The enterprise that deployed an expensive proprietary model pipeline in January found cheaper, better alternatives available by June. The team that ruled out agents in Q1 missed viable agent patterns by Q3.

What Change Fitness Looks Like in Practice

Change-fit organisations share five characteristics:

1. Modular Architecture

Instead of building monolithic AI systems, they build modular ones. Each component (model selection, prompt engineering, retrieval, output formatting, human review) is a distinct layer that can be updated independently.

When a better model arrives, they swap it in without rebuilding the entire pipeline. When a new retrieval technique emerges, they test it in isolation. This modularity is not just good engineering. It is strategic resilience.

2. Continuous Evaluation

They do not just measure AI performance at deployment and move on. They measure continuously:

• Accuracy metrics tracked weekly, not quarterly
• Cost per task monitored in real time
• User satisfaction measured through feedback loops, not annual surveys
• Comparison against new models and approaches run monthly

This continuous evaluation turns AI management from a project into a practice.

3. Learning Culture

The hardest part is not technical. It is cultural. Change-fit organisations:

• Expect and reward experimentation, even when experiments fail
• Share learnings across teams openly and quickly
• Invest in ongoing skill development, not one-off training programmes
• Treat AI literacy as a core competence for all roles, not just technical staff

4. Flexible Governance

Rigid governance frameworks that take months to approve a new model or tool create a bottleneck that guarantees you fall behind. Change-fit governance is:

• Tiered by risk. Low-risk changes (swapping a model for a cheaper equivalent on internal tasks) get fast-track approval. High-risk changes (deploying AI in customer-facing or regulated contexts) get thorough review.
• Principle-based, not prescription-based. Instead of listing approved technologies, define the principles (data privacy, accuracy thresholds, human oversight requirements) and let teams select technologies that meet them.
• Reviewed regularly. Governance frameworks themselves need a review cadence. What was appropriate six months ago may be unnecessarily restrictive today.

5. Strategic Patience with Tactical Speed

Change-fit organisations move fast on tactical decisions (which model to use, how to structure a prompt, when to update a workflow) while maintaining patience on strategic ones (which business processes to transform, where to invest deeply, what competitive position to target).

This combination prevents both analysis paralysis and reckless deployment.

Building Change Fitness: A Practical Framework

Phase 1: Assess Your Current State (Weeks 1 to 4)

• How long does it take to deploy a new AI model or tool from decision to production?
• How often do you evaluate whether current AI implementations are still optimal?
• How do teams share learnings about AI tools and techniques?
• What is the process for updating or replacing an AI component?

If the answers involve “months” and “we don’t,” you know where to start.

Phase 2: Build the Infrastructure (Months 2 to 3)

• Implement modular AI architecture (or refactor toward it)
• Establish continuous evaluation dashboards
• Create a lightweight governance framework with tiered approval
• Set up a regular “AI review” cadence (monthly for most organisations)

Phase 3: Develop the Culture (Months 3 to 6)

• Launch ongoing AI literacy programmes (not one-off workshops)
• Create safe spaces for experimentation (sandboxed environments, innovation time)
• Establish cross-team learning channels (internal demos, shared case studies)
• Recognise and reward adaptation, not just delivery

Phase 4: Sustain and Iterate (Ongoing)

• Review and update your governance framework quarterly
• Benchmark against industry peers and emerging best practices
• Adjust your AI strategy based on continuous evaluation data
• Invest in the skills and tools that support ongoing adaptation

The Cost of Low Change Fitness

Organisations with low change fitness pay a compounding tax:

• They overpay for AI because they do not update to cheaper, better alternatives
• They miss opportunities because approval processes are too slow
• They suffer from change fatigue because every update feels like a major project
• They lose talent because skilled people leave for more adaptive organisations

This tax is invisible in any single quarter but devastating over time.

The Competitive Advantage

The organisations that build change fitness now will compound their advantage over the coming years. Each adaptation is faster, cheaper, and less disruptive than the last. They will adopt new AI capabilities while competitors are still evaluating them.

This is not about being reckless or chasing every new trend. It is about building the capacity to respond when trends become reality, without starting from scratch each time.

Precise Impact works with organisations to build change fitness for AI: from modular architecture design to governance frameworks to cultural change programmes. Contact us to discuss your organisation’s readiness for continuous AI evolution.

Frequently Asked Questions

Why does traditional change management fail when applied to AI?

Traditional change management assumes a stable end state where changes can be ‘refrozen’. However, AI is constantly evolving, with new models and capabilities emerging frequently, making a stable end state impossible to achieve.

What are the key characteristics of a change-fit organisation?

Change-fit organisations typically exhibit five key characteristics: modular architecture, continuous evaluation, a learning culture, flexible governance, and a strategic approach to resilience. These allow them to adapt quickly to new AI developments.

How does flexible governance support change fitness in AI adoption?

Flexible governance avoids rigid frameworks that can slow down AI adoption. It tiers approvals based on risk, allowing low-risk changes to be implemented quickly while ensuring thorough review for high-risk deployments.