Chapter 2: The Autonomy Spectrum — A Maturity Model
Understanding where your organisation sits on the autonomy spectrum is the first step toward transformation. The Pragmatix Agentic AI Maturity Model provides a framework for assessing your current state and planning your journey toward autonomous operations.
Level 1 — Reactive (Chatbots & Basic Assistants)
AI responds to direct questions with pre-defined answers or simple generation. No persistent goals, no tool use, no autonomy. The user drives every interaction.
Characteristics:
- FAQ bots, simple search interfaces
- No integration with business systems
- Every interaction is standalone
Example: A website chatbot that answers common customer questions.
Level 2 — Assistive (Copilots)
AI works alongside humans, offering suggestions and drafting outputs. The human remains in control, reviewing and approving everything. AI may have limited tool access for information retrieval.
Characteristics:
- AI drafts, humans approve
- Some integration with business data (read-only)
- Productivity gains but no autonomy
Example: An AI that drafts emails or summarises documents for human review.
Level 3 — Active (Agents)
AI can pursue multi-step goals, use tools, and make decisions within defined boundaries. Operates with bounded autonomy — acting independently within guardrails, but humans set goals and constraints.
Characteristics:
- AI executes tasks end-to-end for defined scenarios
- Read/write access to business systems
- Human oversight for exceptions and approvals
- Clear escalation paths
Example: An AI that monitors your inbox, triages messages, drafts responses, and flags items needing attention.
Level 4 — Autonomous (Autonomous Systems)
AI operates with minimal human oversight for specific business functions. Sets sub-goals, coordinates with other agents, handles exceptions independently. Humans define objectives and intervene only for escalations.
Characteristics:
- End-to-end process ownership
- Multi-agent coordination
- Self-monitoring and error recovery
- Human governance rather than human operation
Example: An AI system managing customer onboarding from first contact through account setup, escalating only edge cases.
Level 5 — Adaptive (Self-Improving Systems)
AI continuously learns and optimises its own performance. Identifies process improvements, adapts to changing conditions, and evolves its capabilities over time — all within governance frameworks.
Characteristics:
- Continuous learning from outcomes
- Proactive optimisation suggestions
- Adapts to new scenarios without reprogramming
- Strong governance and audit trails
Example: An AI system that not only runs customer onboarding but identifies bottlenecks, tests improvements, and refines its own processes.
Connecting to the Pragmatix AI Capability Model
The five maturity levels map to the Acting domain of the pX-AICM — they represent increasing autonomy in how AI takes action in the world.
Levels 1–4 progress from reactive responses (Generative AI) through to goal-directed autonomous behaviour (Agentic AI). The progression is about increasing agency: from answering questions, to assisting with tasks, to pursuing goals independently.
Level 5 (Adaptive) is different. Self-improving systems don't just act — they learn. This crosses into the Thinking domain, drawing on machine learning capabilities like reinforcement learning to evolve behaviour based on outcomes. It's not just more autonomy; it's a qualitatively different capability.
This is why Level 5 is rare in practice. It requires not only mature agentic capabilities, but also the data pipelines, feedback loops, and governance frameworks to enable continuous learning safely.
→ See Addendum: Pragmatix AI Capability Model (pX-AICM)
Assessing Your Current State
Most organisations today operate at Level 1-2 across their business functions. Some may have pockets of Level 3 capability in specific areas. Very few have achieved Level 4-5 at scale.
A Pragmatix assessment examines your organisation across key dimensions:
- Process readiness: Are your processes documented, consistent, and digitised?
- Data availability: Do your agents have access to the information they need?
- Integration capability: Can AI systems connect to your tools and platforms?
- Governance maturity: Do you have frameworks for AI oversight and accountability?
- Risk tolerance: What level of autonomy is appropriate for your context?
- Skills and culture: Is your team ready to work alongside autonomous systems?
The output is a maturity scorecard and a prioritised roadmap identifying which business functions are ready for increased autonomy — and what foundations need to be laid first.
The Transformation Journey
Moving up the maturity levels isn't about racing to Level 5. It's about finding the right level of autonomy for each business function, based on risk, complexity, volume, and value.
The key question isn't "how autonomous can we make this?" but rather "what's the right level of autonomy for this task, given our context?"
Pragmatix helps organisations answer this question — and then execute the transformation pragmatically, one step at a time.