Every enterprise we work with believes they're further along the AI maturity curve than they actually are. The CEO says "we're doing AI." The CTO says "we have models in production." The reality: a handful of pilots, no shared infrastructure, and no compound value. Here's a maturity model that tells the truth.
What You Need to Know
- Most enterprises are at Level 1 or 2 - experimenting or standardising. Very few have reached Level 3 (scaling), and almost none in the NZ/AU market are at Level 4 or 5.
- The gap between levels isn't technology - it's organisational capability. Moving from Level 2 to Level 3 requires shared infrastructure, governance, and cross-functional collaboration that most organisations haven't built.
- Self-assessment is unreliable. Enterprises consistently rate themselves 1-2 levels higher than external assessment supports. Having AI in production doesn't mean you're mature - it means you've started.
- Each level has specific prerequisites that must be met before progressing. Skipping levels doesn't work. The organisational capability debt catches up within 6-12 months.
- The model isn't aspirational. Not every enterprise needs to reach Level 5. Level 3 (Scaling) is the target for most organisations and delivers the majority of enterprise AI value.
78%
of enterprises self-assess their AI maturity at Level 3 or above; independent assessment places 82% at Level 1-2
Source: MIT Sloan Management Review, Enterprise AI Maturity Study, 2025
The Five Levels
Level 1: Experimenting
What it looks like: Individual teams or enthusiastic people are exploring AI tools. ChatGPT subscriptions. A Copilot licence. Maybe a proof-of-concept or two. No organisational coordination.
Defining characteristics:
- AI initiatives are bottom-up, driven by individual curiosity
- No shared infrastructure or standards
- No AI governance or usage policy
- Success is measured by "does the demo look impressive?"
- AI budget comes from existing team budgets, not strategic allocation
Enterprise AI Maturity: Self-Assessment vs Reality
Source: MIT Sloan Management Review, Enterprise AI Maturity Study, 2025
The test: Can you name three AI capabilities in production that deliver measurable business value? If not, you're at Level 1.
Where most NZ/AU enterprises are: Here. And there's no shame in it - this is where every organisation starts. The problem is staying here too long.
The Experimentation Trap
Level 1 feels productive because people are learning and pilots are succeeding. But pilots that don't scale are sunk costs, not investments. The average enterprise spends 12-18 months at Level 1 before realising that more experiments won't produce operational AI.
Level 2: Standardising
What it looks like: The organisation has decided to take AI seriously. There's an AI strategy (or at least a directive from leadership). Standards are emerging: approved tools, basic governance, a small central team or designated AI lead.
Defining characteristics:
- AI strategy exists and has executive sponsorship
- Basic AI governance: usage policy, data classification, approved tools
- 1-3 AI capabilities in production, delivering measurable value
- Some shared infrastructure beginning to form (but mostly project-specific)
- AI budget is a defined line item
- Readiness assessment completed
The test: Do you have a shared AI platform that multiple capabilities build on? If each AI capability was built independently, you're at Level 2, regardless of how many you have in production.
The transition to Level 3: This is the hardest jump. It requires committing to a platform approach: shared infrastructure, cross-functional teams, and the discipline to build foundations before features.
18
months: average time enterprises spend at Level 2 before either progressing to Level 3 or stalling permanently
Source: BCG, AI Scaling Survey, 2025
Level 3: Scaling
What it looks like: AI capabilities are built on shared infrastructure. New capabilities deploy faster and cheaper than previous ones. The compound effect is measurable. Governance is operational, not just policy.
Defining characteristics:
- Shared AI foundation in production: data pipelines, model orchestration, governance framework
- 4+ AI capabilities in production, with measurable compound acceleration
- Each new capability is demonstrably faster and cheaper than the last
- Cross-functional AI team with both technical and domain expertise
- AI governance is operational: monitoring, audit trails, incident response
- ROI is measured across all three layers: direct, compound, and strategic
- Organisation can deploy a new AI capability in weeks, not months
The test: Is your fourth AI capability significantly faster and cheaper to deliver than your first? If the compound curve is visible, you're at Level 3.
This is the target for most enterprises. Level 3 delivers the majority of AI value: compound economics, operational governance, and the ability to respond quickly to new opportunities. Levels 4 and 5 matter for organisations where AI is the core of the business model.
Level 3 is where AI stops being a technology initiative and starts being an organisational capability. Everything after Level 3 is acceleration.
Isaac Rolfe
Managing Director
Level 4: Optimising
What it looks like: AI is deeply integrated into core business processes. The organisation doesn't just use AI. It designs processes around AI capabilities. Data flows are optimised for AI consumption. The AI platform is a strategic asset.
Defining characteristics:
- AI is embedded in core business processes, not bolted on after the fact
- Processes are redesigned around AI capabilities, not just automated
- Sophisticated model management: multi-model orchestration, automated retraining, performance monitoring
- Internal AI team can build and deploy capabilities independently
- AI governance is embedded in engineering practice: automated compliance, continuous monitoring
- Organisation-wide AI literacy, where domain experts can articulate AI opportunities and constraints
- Measurable competitive advantage from AI capabilities
The test: Are your business processes designed for AI, or is AI retrofitted onto existing processes? Level 4 organisations design new processes with AI as a first-class participant.
Level 5: AI-Native
What it looks like: AI is fundamental to how the organisation operates and competes. The business model depends on AI capabilities. Decision-making at all levels is AI-augmented. The organisation generates and uses data as a core competency.
Defining characteristics:
- Business model is AI-dependent, where competitive position relies on AI capabilities
- AI-augmented decision-making across all business functions
- Continuous AI capability development, with new capabilities shipping weekly
- Advanced AI operations: self-healing systems, automated model lifecycle, real-time performance optimisation
- The organisation is a talent magnet for AI professionals
- AI ethics and governance are cultural, not just procedural
The test: Could your business operate competitively without its AI capabilities? If not, you're at Level 5.
Who reaches Level 5: Tech companies, AI-first startups, and the rare traditional enterprise that has made AI foundational to its operating model. This isn't the target for most organisations, and that's fine.
The Honest Assessment
Here's how to assess your organisation against the maturity model. Be honest. Overestimating helps no one.
Assessment Matrix
| Dimension | Level 1 | Level 2 | Level 3 | Level 4 | Level 5 |
|---|---|---|---|---|---|
| Strategy | No AI strategy | Strategy exists, exec sponsor | Strategy + platform commitment | AI integrated into business strategy | Business model is AI-dependent |
| Infrastructure | No shared infrastructure | Some project-specific | Shared AI platform | Optimised platform, multi-model | Self-evolving platform |
| Governance | None | Basic policy | Operational governance | Embedded in engineering | Cultural + automated |
| Team | Individual explorers | Small AI team | Cross-functional AI capability | Independent AI delivery | Organisation-wide AI fluency |
| Value | Demo-stage | 1-3 production capabilities | 4+ with compound acceleration | Process redesign around AI | AI-dependent business model |
| Measurement | Anecdotal | Project-level ROI | Three-layer ROI framework | Real-time value tracking | AI value is business value |
Scoring: Your maturity level is your lowest dimension, not your highest. An organisation with Level 3 infrastructure but Level 1 governance is effectively at Level 1. The governance gap will block scaling.
The Lowest-Dimension Rule
Maturity is constrained by your weakest dimension. Investing more in your strongest area while ignoring your weakest is the most common maturity mistake. Identify the constraint and address it directly.
Progressing Through the Levels
Level 1 → Level 2 (3-6 months)
Focus: Strategy, governance foundations, first production capability.
- Secure executive sponsorship: one senior leader who commits resources and attention
- Establish basic AI governance: usage policy, data classification, approved tools
- Complete an AI readiness assessment across the five dimensions
- Run a discovery sprint to identify the highest-value first capability
- Deliver one AI capability to production with measurable value
Level 2 → Level 3 (6-12 months)
Focus: Platform commitment, shared infrastructure, compound delivery.
- Make the platform decision. Commit to shared infrastructure over standalone projects
- Build or extend your AI foundation with the second and third capabilities
- Establish cross-functional AI team with both technical and domain expertise
- Operationalise governance: monitoring, audit trails, incident response, not just policy
- Implement three-layer ROI measurement to track the compound curve
- Deliver 3+ additional capabilities on the shared platform, each faster than the last
Level 3 → Level 4 (12-24 months)
Focus: Process redesign, deep integration, organisational capability.
- Identify core processes that should be redesigned around AI, not just automated
- Build multi-model orchestration capabilities
- Develop internal team to lead AI delivery independently
- Embed governance into engineering practice with automated compliance checks
- Invest in organisation-wide AI literacy so domain experts understand AI capabilities and constraints
Level 4 → Level 5 (Ongoing)
Focus: AI as business model, cultural transformation, continuous innovation.
This transition is strategic and takes years. It requires AI to become fundamental to how the organisation competes. It's not a target for most enterprises, but the natural evolution for organisations where AI is the core differentiator.
Common Mistakes
Skipping levels. An organisation at Level 1 that attempts a Level 3 initiative (shared platform, multi-model, operational governance) will fail. The organisational capability isn't there. Progress sequentially.
Measuring the wrong level. Having AI in production doesn't make you Level 3. Having a strategy doesn't make you Level 2 either. Maturity is about organisational capability, not technology deployment.
Investing in strengths, not weaknesses. If your technology is Level 3 but your governance is Level 1, investing more in technology won't progress your maturity. Address the constraint.
Confusing activity with maturity. Running more pilots doesn't move you from Level 1 to Level 2. Deploying more standalone capabilities doesn't move you from Level 2 to Level 3. Maturity advances through organisational capability, not activity volume.
The most dangerous position is Level 1 with Level 3 ambitions and no plan for Levels 1.5 and 2. The maturity model isn't a ladder you can skip rungs on.
Dr Tania Wolfgramm
Chief Research Officer
- Where are most NZ/AU enterprises on this model?
- Honestly, Level 1-2. The NZ/AU market is in a "second wave," past initial experimentation and beginning to standardise. A handful of enterprises (primarily in financial services and government) are approaching Level 3. Level 4+ is rare in this market.
- How long does it take to reach Level 3?
- Typically 12-18 months from Level 1, assuming executive commitment and appropriate investment. The Level 2 to Level 3 transition is the longest because it requires the platform decision and shared infrastructure build. Organisations that partner with experienced AI delivery firms can compress this to 9-12 months.
- Is Level 5 realistic for a traditional enterprise?
- For most traditional enterprises, Level 3-4 is the appropriate target. Level 5 implies AI is fundamental to the business model. That's relevant for tech companies and AI-first organisations, not for an insurer or manufacturer where AI augments but doesn't define the business. Aiming for Level 5 when Level 3 is the right target wastes resources and creates unrealistic expectations.
- Should we use this model with our board?
- Yes, with honesty. The maturity model is most valuable when it tells the board where the organisation actually is, not where they want it to be. Present the honest assessment, the target level (usually Level 3), and the specific steps to get there. Boards respond well to structured progression, and it's far more credible than a vague "we need to do more AI."