Use this 5-level AI maturity framework to identify gaps, prioritise initiatives and build a realistic roadmap — without skipping critical steps.
Most companies that come to us have the same story: they tried one or two AI initiatives, the results were underwhelming, and now they're not sure whether the problem is the data, the tools, the team or the strategy. The answer, almost always, is that they skipped steps.
Assessing your company's AI maturity isn't an academic exercise. It's the fastest way to understand where the real blockers are, which types of projects are viable today, and what sequence of investments maximises return without burning resources. Without this diagnosis, choosing between building an agent with LangChain or calling an OpenAI API is like picking an engine before knowing what road you're driving on.
Several AI maturity models exist — Google's, McKinsey's, MIT Sloan's. All have merit, but they tend to be too abstract for SMEs and growth-stage SaaS companies. The framework we use at Simmple is grounded in these models but simplified to be actionable in under a day.
The five levels are: Level 0 — No AI initiatives; Level 1 — Ad hoc experimentation; Level 2 — Structured pilots; Level 3 — Production and integration; Level 4 — Scale and governance; Level 5 — Continuous optimisation and AI-native operations. Each level has specific characteristics across four dimensions: data, technology, people and processes.
Assessing AI maturity isn't just about technology. Companies that fail to scale AI projects rarely fail because of a lack of tools — they fail because of poor data quality, no internal ownership, or processes that were never redesigned to work alongside automated systems.
That's why the assessment must cover four independent dimensions. It's common for a company to be at Level 3 in technology (Anthropic SDK already integrated into the product) but Level 1 in data (training data scattered across spreadsheets) or Level 0 in governance (no policy on what models can and cannot do with customer data).
The most useful diagnosis isn't a 100-question survey — it's a focused set of high-signal questions that quickly reveal where the blockers are. Below are the questions we use in early assessment sessions with clients.
For each question, the answer isn't binary. The goal is to place the company on a spectrum and identify the most critical gaps. An honest run through these questions with the leadership team typically produces more clarity than months of internal debate.
Once you have the diagnosis across dimensions and levels, the next step is turning it into an actionable roadmap. The most common trap here is trying to fix everything at once — especially when leadership gets excited about the possibilities. Our recommendation is to use a simplified version of the RICE framework to prioritise.
For each identified initiative, score it on: Reach (how many processes or users it affects), Impact (what efficiency or revenue gain it generates), Confidence (how certain the technical outcome is) and Effort (estimated implementation hours). Initiatives with the highest score and lowest technical risk come first — typically process automations with n8n or LLM integrations into tools the team already uses.
Self-assessment has a structural problem: companies tend to overrate themselves in dimensions where they've already invested and ignore dimensions where there's no visibility. A company that built a chatbot with the OpenAI API might classify itself at Level 3 in technology — but if that chatbot has no logging, no response quality evaluation and isn't connected to real company data, it's closer to Level 1.
Another frequent mistake is confusing tool access with organisational capability. Having a GitHub Copilot subscription for the engineering team doesn't mean the company has AI maturity in software development — it means the team has access to a tool. Maturity is measured by the ability to extract consistent, scalable value, not by access to technology.
The output of a good assessment isn't a report — it's a list of decisions. For each gap identified, there should be a clear decision: resolve internally, bring in external help, defer or consciously skip. Many companies get stuck because the assessment surfaces too many problems at once. The solution is to sequence, not to solve everything simultaneously.
A realistic roadmap for a company at Level 1–2 typically has three phases: in the first 60 days, a low-risk pilot with clear metrics (for example, automating support email triage with a GPT-4o-based agent); in the first six months, expanding the pilot or launching a second use case using company data; in year one, building the data and governance infrastructure that supports scale. This pace is slower than most founders want, but it's what actually works.
Self-assessment has limits. It's useful for building awareness and aligning leadership, but tends to be less rigorous on the technical side — particularly around data quality, existing system architecture and compliance risk. An external audit makes sense when: the company has already tried one or two pilots without clear results; there's board or investor pressure to show a credible AI strategy; or when evaluating a legacy system migration that will affect AI adoption capacity.
What a good audit delivers isn't a list of tools — it's an honest analysis of where the real blockers are and which sequence of investments has the highest probability of return. If the output of an audit is just 'you need a data lake and an ML model', it probably didn't go deep enough.
An internal assessment using the 5-level framework described here can be completed in 2 to 4 hours with your leadership team. A more thorough external audit — covering systems, data infrastructure and processes — typically takes between 5 and 10 business days.
No. Companies at Level 1 or 2 can get real value from simple automations and LLM integrations in existing tools. The key is to start with use cases that match your current level, rather than skipping stages without the supporting infrastructure in place.
Digital transformation is a broader concept covering process digitalisation, cloud adoption and systems modernisation. AI maturity specifically measures an organisation's ability to adopt, scale and govern AI systems. It's entirely possible to be advanced in one and behind in the other.
Level 0 means the company has no active AI initiatives and lacks sufficiently structured data to support models. It's not a problem in itself — it's a starting point. The first step is usually data organisation and identifying a low-risk use case to build confidence.
It counts, but in a limited way. Ad hoc use of generative AI tools by individual employees sits at Level 1 (Experimentation). Moving up requires integrating those capabilities into workflows, with company data, governance policies and measurable impact metrics.
We recommend an adapted RICE scoring approach: evaluate each initiative by Reach (how many processes it affects), Impact (what efficiency or revenue gain it generates), Confidence (how certain the outcome is) and Effort (implementation hours). Items with the highest RICE score and lowest technical risk should come first.
Yes, and it's very common. A product team might be at Level 3 with automated data pipelines, while the finance department is still at Level 1. The assessment should be run by functional domain, not just at the organisational level — otherwise you'll miss the real gaps.
Próximo passo
If this article raised more questions than it answered about where your company actually stands, that's a good sign — it means you're asking the right ones. We can turn that diagnosis into a concrete roadmap.
Request an AI audit →