Organisational AI maturity: from individual tools to measurable business value

What does organisational AI maturity actually tell us?

Many organisations are experimenting with AI tools, automating individual tasks and developing employee skills. But producing a document or report faster does not necessarily shorten order fulfilment times, reduce errors or improve profitability.

Assess AI maturity through a process and its results: does it address a business priority, is accountability clear, are reliable data available and can value be measured? Different processes within the same organisation may be at different adoption levels.

Six indicative levels of AI adoption

This article uses an indicative six-level framework to support a discussion about the current situation. It is not a certification standard or a validated scale of financial returns. Apply it to a specific process based on how that process operates today.

L0AI is not used systematically

The process follows established working methods. There is no agreed AI adoption objective or owner; individual employee experiments may remain invisible to the organisation.

L1AI is being explored

Employees try tools and build skills. Adoption largely depends on individual initiative, while its effect on the overall process is not yet assessed.

L2AI is applied to selected tasks

Trials have a defined scope, with agreed usage and output checking practices. Benefits for an individual task are not necessarily connected to overall process measures.

L3AI is integrated into a process

AI forms part of a consistent workflow. Responsibilities, working practices and human oversight are defined; effects on quality, delivery times or costs are monitored.

L4Human and AI workflows are coordinated

AI solutions or agents perform multistep tasks and pass results between systems. Operating boundaries, exception handling and oversight are established.

L5Autonomous operation is governed

Systems operate with greater autonomy in selected areas. Reliability and value are assessed continuously, while changes, permissions and human intervention remain controlled.

An important step: from individual tasks to a managed process

Moving from L2 to L3 in this framework represents a change in how work is managed: the question becomes whether AI improves the whole process as well as helping an employee. Reaching another level does not itself guarantee financial value.

Illustrative example: a sales team prepares proposals faster with AI, but customers still choose competitors because order fulfilment takes too long. If production planning is the main constraint, faster proposal writing does not resolve it. Start by examining the planning process, then decide where AI could help. This is a hypothetical example, not a client result.

Greater autonomy does not automatically mean greater value

AI maturity also depends on data, employee capabilities, security, clear accountability and outcome measurement. Microsoft’s agentic AI maturity model also assesses business strategy, governance, technology and organisational readiness. It is a separate framework provided for comparison, not the source of this L0–L5 scheme.

Not every process needs an autonomous agent. Where errors have significant consequences or tasks vary greatly, a solution with human oversight may be more appropriate. Choose a way of working that delivers value and whose risks the organisation can manage.

How Lean helps set the direction for AI implementation

Lean helps identify what most constrains business results, understand the causes and simplify workflows. AI can strengthen that change by helping process information, identify inconsistencies or respond to exceptions faster.

LeanAI starts with a business challenge that matters to the organisation’s leadership. The IMPACT method helps select a priority, test a solution and embed an effective way of working. AI implementation makes sense when the intended process and business outcomes are clear.

What would a justified next step look like?

Choose one process that matters to the business. Assess its current use of AI, identify the constraint on results and agree on a measure: order fulfilment time, quality, costs or another outcome relevant to the challenge.

Define a limited trial, an accountable owner and a way to check results. After the trial, assess value across the whole process, including operating costs and errors. Use that evidence to decide what comes next, rather than aiming for the highest level in the framework.