Article · Governance

EU AI Act readiness starts with the system and your role

For a UK organisation, a useful starting point is the actual AI system, its intended use and the organisation’s place in the supply chain.

Start with use and role
Intended useApplicable obligationsYour role
Reviewed Full analysis · 4 minute read

At a glance

Start with scope, then plan the work.

The system

Define intended use and its connection to the EU market.

Your role

Identify the responsibilities your organisation actually has.

The workplan

Map applicable requirements to owners and evidence.

A generic checklist cannot settle an unclear scope.

An AI Act workplan can become very large very quickly if it begins with every requirement in the legislation. The first task is to establish which questions are relevant to the organisation and the system being considered.

A UK address does not settle the position. The Act can apply to providers placing systems or general-purpose models on the EU market and, in specified circumstances, to providers or deployers outside the EU whose system outputs are used in the Union. The exact arrangement and applicable exclusions matter. European Commission, Article 2 scope.

Make an inventory that describes the use

Record what the system does, who uses it, who is affected and where it is supplied or used. Include the underlying provider and the significant components of the application.

A list of tool names is a start, but it rarely answers the scope question. The same model can support an internal drafting tool or a product used in a much more consequential setting. The intended purpose and operating context need to be visible.

Keep the record small enough to maintain. Link to supporting documents rather than asking teams to copy them into several questionnaires.

Establish the role before assigning the work

Buying a finished service, integrating a model into a branded product and making a substantial change to a system can involve different responsibilities. The organisation needs to examine its actual activity rather than relying only on a supplier’s label.

Document the role assessment and any uncertainty. If the supply arrangement or product use changes, revisit that assessment. Legal interpretation can be focused on the points that materially affect the workplan.

This also improves internal ownership. Product, procurement and legal teams can see why a requirement is being considered and which part of the business needs to respond.

Classify the system carefully

Do not treat every AI application as high-risk, and do not assume a system is outside the relevant rules because it is small or called a pilot. Classification depends on the provisions that apply to the intended use and circumstances.

Begin with potential prohibitions, then examine the relevant classification and other obligations. Keep the reasoning with the system record. Where a conclusion relies on a condition, make that condition part of the operating decision.

Use the amended timetable

The timetable changed in 2026. The general application date remains 2 August 2026, while specified high-risk provisions have later dates under the amendment. A single “AI Act deadline” is therefore an unreliable basis for planning. European Commission AI Act overview.

Our dated UK and EU AI governance: a working map sets out the key milestones and the primary legal source. Use it alongside the system-specific assessment; do not treat a later high-risk date as a blanket deferral of all obligations.

Build evidence around the gaps

Once the relevant requirements are understood, connect each gap to an activity and an owner. Some work will concern documentation or organisational processes. Other work may require engineering changes, evaluation, supplier information or operating procedures.

A useful plan says what evidence is needed and why. “Improve transparency” is difficult to act on. “Review the user information for this feature against its actual capabilities and limitations, with the product owner responsible” is a task someone can complete.

Reuse existing records where they are suitable. A release process, supplier review or incident procedure may already provide part of what is needed. Avoid creating a separate AI process without understanding how it will be used.

Keep the review connected to change

The system will evolve. A new market, new users, a different model or an expanded purpose can change the basis of the assessment.

Identify who tracks those changes and how they reach the right reviewer. A readiness exercise has lasting value when the organisation can keep the position current after the initial report.

Next step: Start with the AI use-case brief and register, then identify the role and scope questions needing attention. Beeston Labs supports technical and organisational readiness, with specialist legal interpretation brought into the relevant decisions.

Research reviewed: 8 November 2026. This article explains a working approach; application to a particular system requires its facts and the current legislation.

Working through this in your organisation?

Explore the related service

A useful place to start

What decision is ahead of you?

Tell us what you are building, adopting or trying to understand.

Discuss your project