AI Maturity: A Practical Framework

Where does your company actually sit on the AI spectrum — and does it matter?

Every company in 2026 describes itself as “AI-native” or “AI-powered.” The terms have become table stakes in pitch decks and landing pages. But there’s a meaningful difference between a company that uses ChatGPT in Slack and one that has AI woven into its product architecture. The gap between aspiration and execution is where most companies actually live.

After spending time assessing companies — both as a potential contributor and as someone building AI-first tooling myself — I’ve found existing maturity models don’t quite fit. Gartner’s five-level AI Maturity Model (Awareness → Active → Operational → Systemic → Transformational) is designed for enterprises with hundreds of employees and dedicated AI budgets. Microsoft’s framework assumes you’re already deep in Azure. McKinsey talks about organisational readiness across business units that smaller teams don’t have yet.

What’s missing is something practical for a small company trying to figure out when and how to introduce AI — not as a checkbox, but as a genuine capability.

The AI Maturity Spectrum

I use six levels. They’re not strictly sequential — a company might be Level 3 on data readiness but Level 0 on AI engineering capability. The value is in honest assessment, not in claiming a higher number.

Level 0 — AI-Unaware

No AI thinking anywhere. Not in the product, not in engineering workflows, not in strategy. The team may be aware that AI exists but hasn’t considered how it relates to their problem space. This is increasingly rare in 2026 but still exists in deeply vertical industries.

Level 1 — AI-Aware

The team uses AI tools for productivity — Copilot, Claude, Cursor — but as individual workflow accelerators, not as a coordinated engineering practice. The product has no AI features. There’s no strategic document about where AI fits. The leadership might say “we’ll add AI later” without specifics.

Level 2 — AI-Aspirational

This is where it gets interesting, and where most companies sit. The company has a written or articulated vision for AI in the product. The CEO might have written a blog post about it. The pitch deck might have an “AI roadmap” slide. They might describe themselves as “AI-born” or “building the foundation for AI.” But there are zero AI features in production, no ML infrastructure, and no one on the team with AI engineering as their primary skill. The gap between the thesis and the codebase is wide.

This level is actually a strong signal — it means the “why” is done. What’s missing is the “how.”

Level 3 — AI-Augmented

First AI features ship to production. These are typically bolt-on capabilities: smart defaults, auto-suggestions, anomaly detection, LLM-powered summaries. The core product is still deterministic and rule-based, but AI handles edge cases or reduces manual configuration. The team might be using a managed API (OpenAI, Anthropic, Cohere) rather than training their own models — and that’s perfectly appropriate at this stage.

Level 4 — AI-Integrated

AI is a core product differentiator, not an add-on. Features exist that literally couldn’t work without AI — predictive systems, adaptive behaviour, intelligent routing based on learned patterns. The company likely has proprietary data that makes their AI capabilities defensible. There’s dedicated AI engineering capacity, even if it’s one person. The architecture accommodates ML pipelines as a first-class concern.

Level 5 — AI-Native

The product is designed around AI from the ground up. Self-healing systems, agents making decisions within defined boundaries, models trained on proprietary operational data. AI isn’t a feature — it’s the architecture. The product doesn’t make sense without it, the way Cursor doesn’t make sense without code completion or Harvey doesn’t make sense without legal reasoning.

The Dimensions That Matter

A single number is reductive. What’s more useful is assessing across dimensions, similar to how Gartner evaluates seven pillars (strategy, product, governance, engineering, data, operating models, culture) but adapted for smaller teams:

Strategic clarity — Does the team know where AI fits in their product, or is it a vague “we’ll add AI” handwave? Greg Rudakov at Support Fusion, for example, has written specifically about contract ingestion driving automatic integration configuration. That’s Level 2 strategic clarity with a Level 4 vision.

Data readiness — Does the product generate data that would actually make AI features valuable? This is the most underrated dimension. A company syncing thousands of tickets across organisational boundaries is sitting on a goldmine of operational patterns, even if they haven’t touched an embedding yet.

Engineering capability — Does anyone on the team have AI/ML engineering as a primary skill? Can they build an evaluation pipeline? Do they understand the difference between prompting and fine-tuning? At the early stage, this often comes down to: could we hire or engage one person and actually build something in 4-6 weeks?

Infrastructure readiness — Can the existing architecture accommodate async ML workloads, model versioning, or even just storing and retrieving embeddings? Or would introducing AI require a significant infrastructure rebuild?

Product integration depth — Is AI a separate feature (“click here for AI summary”) or is it woven into existing workflows so naturally that users don’t think about it?

Why This Matters

Most companies I encounter are at Level 2 — AI-Aspirational. They’ve done the strategic thinking, often quite well. But they haven’t crossed the execution threshold into Level 3.

The transition from 2 to 3 is the most important jump because it’s where the thesis gets tested. You discover whether your data is actually as valuable as you thought. You learn whether your users want AI-driven suggestions or find them annoying. You build the muscle of shipping AI features that need to be evaluated differently than deterministic code.

Gartner’s research supports this pattern at the enterprise scale too: their 2025 survey found that most organisations cluster at Level 1-2 (what they call Awareness and Active), and only 6% reach what they describe as Transformational maturity. The difference is that smaller companies can move through these levels in months rather than years — if they have the right engineering capability at the right time.

The Honest Assessment

If you’re leading a team that’s considering AI, here’s a quick self-assessment:

Can you describe, in specific technical terms, what AI feature you’d build first and what data it would use? If yes, you’re at least Level 2. If you’ve shipped it, you’re Level 3.

Does your product generate proprietary operational data that improves with more customers? If yes, your data readiness is ahead of your product maturity — which is actually the best position to be in.

Could someone join your team today and ship an AI feature within 4-6 weeks using your existing data and infrastructure? If the answer is “yes, if they understood our domain” — you’re ready for the jump. If the answer is “we’d need to rebuild the data layer first” — you’ve got prerequisite work to do.

The Bridge Between Vision and Execution

The most valuable work in early-stage AI isn’t building a sophisticated ML pipeline. It’s building the first concrete thing that proves the thesis. A self-healing integration monitor. A field mapping suggestion engine. A contract parser that extracts SLA terms. Something small enough to ship in weeks, specific enough to evaluate, and connected enough to the product vision that it illuminates the path forward.

The frameworks from Gartner, Microsoft, and McKinsey all converge on one point: maturity isn’t about sophistication of technology. It’s about alignment between capability and business value. That alignment happens fastest when someone on the team can hold both the AI engineering skills and the product context in their head simultaneously.

That’s the bridge. Not a model architecture. Not a framework choice. A person who can look at the data flowing through your system and see what intelligence wants to emerge from it.

Tolics Engineering Mind