Your Organisation Uses AI. But How Mature Is It Really?

One reason maturity has become difficult to assess is the speed at which the benchmark is changing.

In 2023 and 2024, giving thousands of employees access to a capable enterprise AI assistant could itself feel like a major step.

Since then, the conversation has moved rapidly toward enterprise search, retrieval from proprietary data, AI coding environments, multimodal systems, agents, model routing, reusable skills and increasingly sophisticated human and AI workflows.

Microsoft's 2025 Work Trend Index found that 82 percent of leaders considered the year pivotal for rethinking core aspects of strategy and operations. It also found that 81 percent expected agents to be moderately or extensively integrated into their company's AI strategy within the following 12 to 18 months.

The point is not that every enterprise needs to adopt every new AI capability immediately.

The point is that the benchmark itself keeps moving.

This produces a strange situation for leaders.

An organisation can be significantly better at AI than it was six months ago while becoming relatively less mature compared with the market.

ServiceNow and Oxford Economics offer a striking example of this effect. Their 2025 Enterprise AI Maturity Index surveyed nearly 4,500 executives across 16 countries and 11 industries. The average maturity score fell from 44 to 35 in one year, and fewer than 1 percent of organisations scored above 50 on their 100 point scale. Their research attributed the decline partly to the pace of AI innovation moving faster than organisational capabilities such as skills, governance and cohesive adoption strategies.

That does not necessarily mean enterprises became worse at AI.

It means the definition of maturity became harder to satisfy.

You can be improving at AI and still be falling behind.

This is why a maturity assessment cannot only ask how much progress an organisation has made internally.

It must eventually ask how that progress compares with the world around it.

There is another distinction worth making.

Many waves of digital transformation began by converting existing manual processes into digital ones.

A paper form became an online form.

A physical approval became a digital approval.

A spreadsheet became a cloud application.

The underlying process often remained recognisable even after the technology changed.

AI creates an opportunity to go further.

If a process was originally designed around the limitations of human labour, manual information retrieval, fixed software interfaces and scarce expertise, simply inserting an AI prompt into that process may leave most of its original constraints intact.

The better question is not always

Where can AI help with this process?

Sometimes it is

If we designed this process today with AI available from the beginning, would we design it this way at all?

This distinction is already visible in organisations seeing stronger results from AI.

McKinsey found that organisations it classified as AI high performers were nearly three times as likely as others to report fundamentally redesigning workflows when deploying AI. Fifty-five percent of high performers reported this kind of redesign compared with 20 percent of other respondents. McKinsey identified workflow redesign as one of the factors most strongly associated with meaningful organisational impact from AI.

This is the shift from putting AI inside work to redesigning work around AI.

There is another reason executive leaders struggle to answer questions about AI maturity.

The organisation itself is uneven.

Engineering may already be building agents that interact with development tools and internal systems.

Marketing may use ChatGPT or Claude heavily, but primarily through individual prompting.

Operations may have deeply automated processes.

Finance may have access to a narrow set of approved tools.

Legal and compliance may still be establishing policies around what data can enter which models.

One business unit may have a sophisticated internal AI platform while another is experimenting independently with consumer applications.

All of these realities can exist inside the same enterprise.

IBM's 2025 CEO study provides some evidence of how fragmented this environment can become. Half of the 2,000 CEOs surveyed said the speed of recent technology investments had resulted in disconnected or piecemeal technology within their organisations. At the same time, 68 percent identified an integrated enterprise-wide data architecture as critical for cross-functional collaboration, while 72 percent considered proprietary data important to unlocking generative AI.

This matters because AI maturity is not simply a question of which model an organisation has purchased.

It is also a question of whether AI has access to the right context, whether systems can work together, whether employees can build reusable capabilities, and whether governance allows all of this to happen safely.

An enterprise can have thousands of AI users and still have very little shared AI capability.

That distinction is easy to miss.

If 5,000 employees independently become 20 percent more capable with AI, the organisation has gained something meaningful.

But if the knowledge of how those employees use AI remains entirely individual, the company may struggle to turn that improvement into an organisational capability that can be reused, governed and scaled.

At Kai, we believe AI maturity has to go beyond counting licenses or asking employees whether they use generative AI.

It needs to examine how AI works across the organisation.

That includes how employees get access to AI tools and whether adoption is coordinated or fragmented.

It includes governance and how new AI tools are reviewed before they are used with company information.

It includes what employees and teams have actually built, whether that is reusable prompts, assistants, workflows, agents or more sophisticated internal systems.

It includes the degree to which AI can securely work with organisational information rather than relying on employees to manually provide context every time.

It includes how AI outputs are reviewed and how human oversight changes based on the importance or risk of a decision.

It includes whether teams deliberately choose different models for different requirements rather than defaulting to whichever tool happens to be available.

Most importantly, it includes the workflow itself.

Has AI made the old process slightly faster, or has the organisation started changing how the work gets done?

These are very different signals of maturity.

The conversations behind this research kept returning to one practical problem.

Even leaders who recognise that AI maturity goes beyond tool access struggle to answer a simple question.

Where do we actually stand?

A conventional maturity model can provide a score.

But a score without context has limited meaning.

Imagine an organisation scores 70 today.

Is 70 strong?

Perhaps.

But what if comparable organisations average 45?

Or what if they average 80?

Now imagine the organisation improves from 70 to 76 over the next six months.

That sounds encouraging.

But if its peers move from 75 to 88 during the same period, the organisation has improved internally while losing ground externally.

This is why we designed the Kai AI Maturity Survey as a live benchmark rather than only a static assessment.

👉 You can explore it here: https://kaitrust.com/ai-maturity-model

The survey looks at how deeply AI is embedded across an organisation, including governance, workflows, data access, model choices and what teams are actually building.

Participants can establish where they stand today and then compare their maturity with the broader benchmark as more organisations participate.

Over time, the goal is not simply to answer

Are we better at AI than we were before?

It is to answer

Are we changing quickly enough compared with the organisations around us?

There is one final reason we believe maturity should be measured through capabilities rather than specific technologies.

Today's frontier will become tomorrow's baseline.

Chat interfaces gave way to copilots.

Copilots are increasingly being joined by agents.

Agents will become more connected to enterprise systems.

New approaches to context, orchestration, tools, skills and human oversight will continue to emerge.

Microsoft already describes a progression from individual AI assistance to human and agent teams, and eventually toward processes where humans provide direction while agents operate larger portions of the workflow.

The exact technologies will change.

A durable maturity model therefore cannot define an AI-native enterprise as one that happens to use today's fashionable architecture.

Instead, we think the more enduring capability is this

How quickly can an organisation absorb a new AI capability and turn it into a safe, repeatable and scalable way of working?

That is a much higher bar than providing access to an AI assistant.

It requires technology, but also governance, data, organisational learning, workflow design and leadership.

And perhaps most importantly, it requires organisations to stop thinking of AI as another software rollout.

The leaders in this transition will not simply give employees better tools.

They will continuously reconsider how their organisations should work now that intelligence itself has become increasingly available as a capability.

The question for enterprise leaders is therefore no longer simply whether their teams use AI.

It is whether the organisation is learning, adapting and redesigning itself quickly enough to keep pace with what AI makes possible.

Is your organisation simply using AI, or is it genuinely becoming AI-native?