AI readiness: are the foundations in place for AI in engineering?

Before making larger investments, organisations should examine whether data is accessible, responsibilities are clear and suitable working practices exist.

The following model considers these conditions separately. Concrete questions show where a sound foundation exists and where decisions are still needed.

Dimension 1: data

Whether an AI application can work effectively depends on the data that is both available and permitted for use. Three questions provide an initial assessment:

• Accessibility: Is the relevant data structured and accessible, or is it spread across file storage, inboxes and undocumented knowledge?
• Quality: Is the data current, complete and consistent enough for the intended use?
• Classification: Is it clearly defined which data external AI services may process and which data must remain internal?

A sound data classification is often the missing foundation. Without it, approvals and restrictions rely on individual decisions that are difficult to apply consistently across the team.

Dimension 2: systems and tooling

AI support needs to fit the existing working environment, including development tools, ticketing, documentation and CI/CD. Three questions indicate how well the technical foundations are prepared:

• Do the relevant systems provide suitable interfaces through which the required context can be made available in a controlled way?
• Is there a test or staging environment where AI-assisted changes can be assessed without putting production systems or data at risk?
• Is it clearly defined which systems an AI tool may read and where it is permitted to make changes?

Missing interfaces or test environments are often a sensible place to start. Improvements in these areas strengthen engineering and operations regardless of how AI is ultimately used.

Dimension 3: team

New tools do not automatically create a shared way of working. Without preparation, individual routines emerge that complicate reviews and collaboration. Teams should therefore clarify three points:

• Does the team share an understanding of which tasks AI tools can support and where their use would not be appropriate?
• Are there shared conventions for the context provided, instructions given to the tool and the handling of AI-generated code?
• Can the existing review process handle a higher volume of changes without reducing review depth or quality?

Review capacity limits the achievable benefit. If changes are produced more quickly, the team must still have enough time and technical attention available to assess them properly.

Dimension 4: governance

Technical governance begins with clear decisions and responsibilities. Three questions are particularly relevant for day-to-day use:

• Who approves AI tools, and which technical and organisational criteria guide that decision?
• Who is responsible for code produced with the support of a coding agent?

Responsibility remains with the team member who adopts the change and introduces it into the codebase.

• Are decisions about AI use documented so that their rationale, ownership and safeguards can be understood later?

Regulatory considerations

Readiness also includes identifying which regulatory checks may be required. Our EU AI Act risk checker uses your answers to indicate possible triggers, roles and requirements.

The result is not a legal assessment. Deadlines and further context are covered in EU AI Act: deadlines and obligations.

Self-assessment

Rate each dimension as clarified, partly clarified or open. Our AI readiness self-assessment guides you through eight statements in about three minutes.

The result provides a view of the current position and suggests next steps. Two patterns occur frequently:

Pilot without organisational adoption: Data and systems are partly prepared, while shared practices and ownership are still missing. These areas should be addressed before broader adoption.

Formal governance without approved practice: Rules exist, but data, systems and permitted working methods remain unclear. A practical approval process based on data classification provides the next step.

Conclusion

The four dimensions show whether a planned use of AI has a sound foundation. They also identify which technical or organisational questions should be resolved before the next investment.