MDS AI strategy

AI is now part of the delivery layer

MDS is no longer interpreted only by designers and developers. It is increasingly executed by AI as part of modern delivery workflows.

That changes the role of the design system. MDS can no longer function primarily as guidance for humans to interpret. It must also work as a system that can direct, constrain, and support execution at machine speed.

Ambiguity does not scale

Where MDS leaves room for interpretation, AI will make the call.

At AI speed and volume, vague or implicit guidance becomes inconsistency by default. That creates quality risk, compliance risk, and a growing gap between intent and delivery. This is not simply a tooling issue. It is a structural weakness in how standards are defined and applied.

Review cannot keep up with generation

AI can generate more UI than human-led review can realistically govern.

That means quality, accessibility, legal, and brand standards cannot depend primarily on downstream checks. They need to be built into the point of creation, so the right decisions happen earlier and more consistently.

Rules

Define the decisions that should not be left to interpretation

The first job of MDS is to remove avoidable ambiguity.

High-level guidance still has value, but critical decisions cannot remain implicit as output becomes faster and more automated. MDS must define the rules that govern component usage, interaction behaviour, accessibility expectations, and layout constraints.

Making those decisions explicit and enforceable reduces variation, removes repeated decision-making, and creates a shared baseline for how Maersk digital products are designed and built.

Outcome: consistency by default, not by exception

Quality and compliance

Set a non-negotiable quality bar

Today, too much quality depends on education, individual judgement, and late-stage review. That model does not scale.

MDS must define a clear quality and compliance bar, and make deviation easier to detect earlier in the process. This means moving beyond describing what good looks like and becoming clearer about what is acceptable, what is not, and where standards must hold.

Human judgement still matters, but it should be applied where it adds the most value, not spent compensating for preventable ambiguity.

Outcome: higher quality, faster delivery, and lower risk at scale.

Workflow

Embed MDS into how work gets done

As teams increasingly work with AI-assisted design and development tools, MDS cannot sit on the sidelines as guidance or a component library alone.

It must function as a shared design and build language that shapes how digital experiences are created, structured, and validated across roles and tools. The goal is not simply better documentation. It is to make MDS active in the workflows where decisions are made.

That shifts MDS from something teams consult after the fact to something that actively supports creation, review, and validation in flow.

Outcome: stronger adoption, fewer gaps between guidance and delivery, and better alignment with how teams actually work.

Shift 1: Structure the source of truth

Strengthen MDS as a single source of truth that can be consumed by both humans and AI.

This means consolidating content, reducing ambiguity, increasing precision, strengthening examples, and extending the core where needed to support emerging interaction patterns. The aim is to remove the gaps that still depend on interpretation and make MDS a more dependable system of record for digital experience decisions.

This is likely to include work such as:

• centralising website and guidance content into a more structured source of truth
• tightening guidelines and best practices across components, accessibility, forms, search and filter, feedback, content, and iconography
• improving examples so they reflect Maersk use cases rather than generic scenarios
• reducing the separation between website guidance, pattern documentation, and API documentation so MDS operates as a more unified source of truth
• extending the core where needed to better support emerging AI interaction patterns such as chat

Shift 2: Embed standards into workflows

Move MDS closer to where design and build decisions are made.

This means improving how standards are consumed in tools and workflows, increasing machine-readable structure, and making MDS rules easier to apply in practice. It is also where compliance becomes more operational, through stronger codification of areas such as accessibility, layout, branding, and adherence to the MDS source of truth.

This is likely to include work such as:

• transforming source content for different audiences and execution contexts
• improving support for AI-assisted design and engineering workflows
• reducing reliance on manual interpretation by making standards easier to consume programmatically
• increasing the extent to which MDS can guide or validate conformance to defined rules

Shift 3: Expand MDS beyond today’s core

Extend MDS into adjacent areas where stronger structure would improve consistency, reuse, and control.

This may include broader reusable assets, experience rules, business-specific components, and more direct ways for teams to design and build with MDS where current tools are too limited or too generic for Maersk’s needs.

This is likely to include work such as:

• extending the source of truth beyond the current core into areas such as templates and data visualisation
• supporting more business-specific components or reusable design intents
• supporting more consistent implementation of legal and regulatory requirements where these can be reflected in reusable patterns, rules, or guidance
• improving how customer research, internal user insight, and personas are structured and reused so they can better support experience quality and the continued evolution of MDS
• creating more direct ways for teams to design and build with MDS where existing tools constrain quality, control, or evolution

Dependencies

• Sustained leadership backing
  This direction depends on continued support for MDS as strategic infrastructure, not just a component library or service function, and on enough organisational agency to influence standards and decisions across Maersk beyond its current place in the hierarchy.
• Cross-functional alignment
  Progress depends on alignment across design, engineering, and product, especially where standards, workflows, and ownership boundaries need to evolve together.
• Access to enterprise AI capability within business constraints
  Some parts of this direction depend on enough access and capacity to explore, validate, and apply AI-supported workflows at meaningful speed.

Principal risks

• The market settles differently than current AI expectations suggest
  AI adoption may prove slower, narrower, or more uneven than current narratives imply. If so, parts of the framing may age faster than the underlying direction. The strategy remains sound if it continues to strengthen MDS as a source of rules, quality, and workflow support, rather than depending on a single AI scenario.
• Key decisions remain too distributed
  If standards, quality expectations, and compliance continue to be decided and interpreted in multiple places, teams will continue to create local guidance and working practices around MDS. That will make inconsistency harder to reduce and weaken MDS as the default source of truth.
• Loss of credibility through over-extension
  If MDS expands its scope or narrative faster than the product, content, and workflow support can sustain, the strategy risks feeling broader than it is deep. Progress depends on staying focused on the areas where MDS can add clear value first.