AI is accelerating the development of interfaces: how to foster a coherent digital ecosystem
From customer-facing services to internal tools, AI is speeding up the creation of interfaces. The challenge is to develop them as a shared ecosystem, building on what the company has already established.
Zeroheight’s Design Systems Report 2026 finds that 56 per cent of the 147 professionals surveyed are experimenting with or using AI in their design systems, up from 46 per cent previously. Ten per cent have integrated it into their processes: there remains a gap between experimentation and structured use.
Websites, apps, e-commerce platforms and customer portals are parts of the same digital ecosystem, developed at different times by different teams and suppliers.
Added to these are those created within the organisation: dashboards, data-viewing tools and prototypes. These can be generated within language model-based (LLM) assistants by asking in natural language to generate code: ‘vibe coding’.
The customer has a single relationship with the company and expects consistent behaviour across services. Internal staff also benefit from consistent tools.
AI rapidly adds pages and features, but without shared references, each interface can introduce differences: a redesigned form, an action that changes meaning. Production accelerates and the workload increases to keep the whole system consistent.
When these interfaces are put into actual use, responsibilities, checks and maintenance must be clarified. A rule that changes must be updated across the various products.
Many companies have already compiled the recurring elements of their services—such as forms, buttons and menus—along with the rules for using them. This body of work is known as a design system: it links products to a shared experience and enables teams and suppliers to reuse pre-defined solutions.
With AI, this body of work can also guide part of the automated production process. However, the tools must understand which element to choose, which variants are permitted and when a new decision is required.
Some of this information is implicit: people acquire it through experience. For an AI tool, it must become an accessible and sufficiently precise reference. Otherwise, a plausible interface simply adds another exception.
This is what emerged from the assessment conducted by the Orvian group for an organisation with multiple products and teams, analysing five libraries covering design, code, documentation, adoption and governance.
On the assessment scale, the architecture of the variables organising colours, spacing and other shared values scored 92 out of 100. The context of the components, including descriptions and usage guidelines, scored only 19, however.
These scores relate to two dimensions of the same system; they are not percentages of savings. The investment was there, but the information needed to use it correctly with AI was incomplete.
The assessment, which lasted four weeks, distinguished between what needed to be retained and what required action: documenting conditions of use, linking design and code, and clarifying responsibilities. The findings became priorities and a roadmap shared with the team.
This is a managerial issue: if design and implementation do not align, what guides the work? Who decides whether a discrepancy should be corrected or approved as an evolution?
Applying an approved solution and introducing a new one entail different responsibilities. Making this boundary explicit allows us to assign specific tasks to tools and to return decisions that change the system to people.
This is where AI-ready Design System, the service through which Orvian makes the company’s design system usable in AI workflows, either building on the existing foundation or supporting its development within a dedicated scope.
The process begins with a dialogue between those who design, develop and maintain the system. The assessment identifies gaps and discrepancies between design and code; a limited pilot project is selected, with evaluation criteria defined prior to implementation.
The assessment compares equivalent activities before and after, using the same model and tools: reuse of solutions, deviations, corrections. Is an unforeseen request reported to the team or resolved using an unapproved approach?
The pilot assesses the impact of the intervention, whilst also taking into account time, staff workload and the maintenance of references. This evidence helps determine whether to roll it out.
The handover to the team serves to keep the rules up to date and ensure the verifications are repeatable. The objectives are a recognisable presence for marketing, less duplication for technology, and continuity between services for customers and internal tools.
The tools will change: components, rules and decisions remain company assets, which can also be utilised by subsequent teams. Each new interface can thus contribute to the growth of the ecosystem, offering a recognisable and sustainable experience over time.
