AI Readiness in Practice: Insights from Medium and Large Organisations Across Europe
AI Readiness in Practice: Insights from Medium and Large Organisations Across Europe is the most recent report compiled by Antal in partnership with MakoLab. It shows that only 3% of the respondents report AI autonomously executing selected business processes in their organisations today. At the same time, 93% expect its autonomy to increase within the next two years.¹
What lies between current practice and that expectation is the demanding organisational task of providing AI with reliable knowledge, defining its authority and being able to account for what it does.
AI Readiness in Practice examines whether the knowledge, controls and operations underpinning current AI deployments are capable of supporting greater autonomy.
Deployment is widespread. Maturity isn’t
The results of the research reveal 85% of the respondents stating that their organisation is deploying AI in selected business areas or broadly across the organisation. Only one in three rates their organisational AI maturity as high.
More than half describe moderate maturity, with AI used in some teams. However, coordination between them is limited.
Accountability is defined in most of the organisations represented in the study, most commonly across functions or through a business owner. As more work is assigned to AI, those arrangements need to be clear about who can authorise actions, intervene and answer for the outcome.
For leaders planning the next stage, the question becomes more specific. What can the organisation confidently entrust to AI?
Three tests of AI maturity
In his commentary for the report, Jerzy Wiśniewski, MakoLab’s Data & AI Service Line Manager, frames maturity by way of three tests:
Context: is AI working on the basis of the organisation’s own knowledge or just from general knowledge about the world?
Control: are the organisation’s policies enforced or merely written down?
Evidence: can the organisation explain and reproduce an AI-generated outcome months later?
As AI gains authority to act, postponing each test becomes increasingly questionable.
Greater autonomy raises the cost of unresolved questions
At present, the most common division of labour is AI executing actions under human supervision, as reported by 38% of the respondents.
When AI supports an individual task, a member of staff can question an answer before using it. As a sequence of activities is assigned to AI, organisations need to settle several questions.
- Where in the sequence will that scrutiny happen?
- Which information should the system trust?
- Who owns the business rule it applies?
- At what point should it stop and seek approval?
These questions become consequential when an output becomes the input to the next action. In his commentary for the report, Karol Waszak, Antal’s Product Manager ReX, describes how an error that occurs early in an automated process can remain present through every subsequent stage and produce an outcome that appears entirely correct.
Connecting AI and internal knowledge is only the beginning
Retrieval-augmented generation connects generative AI to sources such as internal documents, policies and knowledge bases. In AI Readiness in Practice, 55% of the respondents report that their organisation uses RAG either extensively or in pilot projects. At the same time, only half agree that terminology and definitions are consistent within their organisation.
That combination exposes a practical difficulty. Retrieving an internal document is not the same as establishing that its definitions agree with those used elsewhere in the organisation. Connecting more sources may give AI more context but leave the question of which source takes precedence unresolved.
Can your organisation explain AI’s actions?
AI Readiness in Practice shows nine in ten respondents reporting some form of AI governance. However, only 20% strongly agree that their organisation can explain, reproduce and defend AI-generated outcomes during an audit, with a further 40% agree somewhat.
These findings relate to different capabilities. Establishing rules is one part of governance. Reconstructing a particular outcome requires those rules, decisions and records to remain traceable in practice.
This distinction matters well before an audit. When an unexpected result affects a customer or interrupts a process, the organisation will need enough evidence to investigate what happened and decide what to change.
Readiness also depends on what leaders can see
One of the report’s most revealing findings concerns uncertainty. Asked which operational data were sufficiently structured and reliable to support AI-assisted operations, 30% of the respondents did not know.
This is not to say that the underlying capabilities are missing. What it does show, though, is that the visibility of those capabilities is limited for some senior decision-makers.
The findings also raise a question about how investment priorities align with operational readiness. When identifying their organisation’s top three AI-related investment priorities for the next twenty-four months, 60% of the respondents selected AI agents. Just 8% selected observability and monitoring.
What should be ready before more is assigned to AI?
AI Readiness in Practice examines how AI is used, who is accountable for its outcomes and how the respondents assess the knowledge, governance and operational foundations behind it.
Download the report to explore how adoption compares with organisational maturity, the areas where the respondents report uncertainty and what greater autonomy will demand of the organisations preparing for it.
To apply the three tests to your own processes, knowledge and controls, book a working session with MakoLab’s experts.
Explore the research programme
AI Readiness in Practice is part of a research programme carried out by Antal in partnership with MakoLab. The programme has also produced four companion reports examining the foundations of enterprise AI:
About the research
¹ All the figures come from AI Readiness in Practice: Insights from Medium and Large Organisations Across Europe, a report compiled by Antal for MakoLab, October 2026. Antal conducted the CAWI survey between 16 July and 7 September 2026. More than 500 potential respondents were invited and 113 started the questionnaire. The final analytical sample consists of 40 respondents based in 11 European countries: Czechia, Denmark, France, Germany, Hungary, the Netherlands, Poland, Portugal, Spain, Sweden and Switzerland. They all hold positions at senior manager level or above and represent medium and large organisations with more than 200 employees.
All the organisations represented were already exploring or implementing AI.
The findings describe this defined group and should not be interpreted as estimates for the wider European business population. Percentages are calculated on the basis of n = 40 and rounded to the nearest whole number. Multiple-response questions do not total 100%.
Anna Kaczkowska
Content Marketing Specialist
Responsible for planning, creating and managing content
Jerzy Wiśniewski
Cybersecurity Architect
