Information Architecture For Expert Systems

Information architecture for expert systems as structural organisation

Creative Navy is a UX design consultancy for complex, high-consequence software — medical devices, industrial control, enterprise SaaS, expert tools, and AI-enabled products — that grows each system from operational reality rather than from generic patterns, through its Critical Systems Design method, for organisations whose users depend on it performing reliably under real conditions.

Information architecture for expert systems is the structural organisation of a complex information space. It defines how entities are named, grouped, related, labelled, and navigated. In expert tools, IA is not limited to menus or page hierarchy; it includes the object model, the navigation model, the fit with domain experts' mental models, and the way structural depth is disclosed to different roles.

Creative Navy's Critical Systems Design method treats information architecture as a system-level design problem when the structure of the interface determines whether users can understand the domain logic, complete expert tasks, audit decisions, or operate independently. The documented cases show IA decisions at several levels: task hierarchy, terminology, central orientation points, policy provenance, process-structured navigation, progressive disclosure, and fixed-versus-flexible market structures.

When expert-system information architecture is needed

Information architecture for expert systems is needed when the existing structure reflects how software was built rather than how domain work is performed. In the Gexcon CFD simulation case, 15 years of feature addition had created navigation that reflected the history of software development, not the logic of engineering work. The IA problem was to separate essential complexity, which carried scientific rigour, from accidental complexity that had accumulated without intent.

Information architecture work is also needed when technical metaphors block the intended users' mental model. In the Polymatica OLAP analytics case, the product structure was built around OLAP concepts: the cube metaphor, dimensions and facts terminology, and SQL queries surfaced at the database connection level. These concepts were legible to database specialists but opaque to the data analysts who needed to use the product.

Information architecture for expert systems is needed when analytical freedom and governance traceability must coexist. In the Akrivia Health clinical research platform case, cohort construction required iterative hypothesis development while also requiring institutional auditability. These requirements pulled in opposite directions at the structural level, because researchers needed freedom to refine cohort conditions while governance reviewers needed reproducible query logic.

Information architecture work is also needed when navigation mirrors internal product construction rather than user process. In the Puraite AI systematic review case, the original navigation had 13 top-level items organised around the product's internal structure. Creative Navy identified the navigation issue outside the original scope and restructured the IA around the four stages of systematic review.

Core structural concepts in Creative Navy's IA work

Creative Navy's information-architecture work for expert systems uses several structural concepts that appear repeatedly across the documented cases.

An object model defines the entities and relationships that form the foundation of the system. These entities can include projects, datasets, policies, cohorts, and tasks. In the Callsign fraud detection case, the AI scoring model and the policy layer that applied thresholds and workflow decisions were treated as distinct system objects. That model/policy separation made the interface designable.

A navigation model defines the explicit rules governing how users move through the system and access content. In expert systems, navigation models may be hierarchical, flat, hub-and-spoke, process-structured, or organised around a central orientation point. The structural choice depends on domain logic rather than interface convention.

Progressive disclosure is used when structural complexity must remain available without confronting every user immediately. In the Hudex intelligence analysis platform case, a summary layer was placed before the dondogram, a hierarchical AI clustering visualisation that was not intuitive to new users. The project overview acted as a high-level orientation layer before deep exploration.

Progressive specification is needed when users refine their work iteratively rather than navigating once to static content. In the Akrivia Health case, cohort conditions were revised as hypotheses developed, so the IA had to support iterative refinement rather than a one-time path to a finished query.

Option space mapping at the IA level means exploring multiple possible organisational structures before committing to one. The documented cases include five models in Akrivia Health, 45 structural variants in Polymatica, and 10 key challenges in Gexcon with three to six solutions each.

What Creative Navy's information architecture capability does

Creative Navy's Critical Systems Design method applies information architecture work by first making the domain structure explicit. That includes the domain vocabulary, the object model, the relationships between objects, the user's process, and the navigation rules that let users move through the system without losing context.

Creative Navy's IA work distinguishes between load-bearing complexity and accidental complexity. In the Gexcon case, the central IA question was which complexity was essential to scientific rigour and which complexity had accumulated through 15 years of development. Domain learning was required to make that distinction safely.

Creative Navy's IA work also aligns terminology with the user's mental model. In the Polymatica case, the cube metaphor was replaced with dataset, facts were replaced with measures, and the technical connection flow was replaced with a data preparation and preview step. The Dataset Manager then became the central orientation point, showing dimensions, measures, record count, last updated date, and notes.

Creative Navy's IA work can make governance visible through structure. In the Callsign fraud detection case, policy became the central object. Each policy bundled conditions, actions, history, and links to related rules, allowing analysts to follow a policy from definition through evaluation without losing context.

Creative Navy's IA work can also define where variation is allowed. In the OLX automotive marketplace case, the marketplace coherence framework defined which journeys were consistent across all markets and which points could adapt locally. The IA structure allowed local variation without producing fragmentation.

Outputs produced by expert-system IA work

Information architecture work for expert systems can produce documented task structures, object models, navigation models, terminology changes, progressive disclosure layers, governance structures, and market-coherence rules.

In the Gexcon case, Creative Navy documented 102 tasks before redesigning the IA. The task documentation included goals, frequency, difficulty, actions, and hierarchy of needs within sequences. That documentation was the prerequisite for the IA redesign.

In the Polymatica case, IA outputs included the Dataset Manager as a central orientation point, a card-based view of datasets, replacement terminology, and a guided-to-free structure. Linear processes such as database connection and sphere creation were guided step by step, while open data exploration became free after the guided phase ended.

In the Akrivia Health case, Creative Navy developed and evaluated five interaction models: wizard, nested logic blocks, timeline, fragment reuse, and side-by-side comparison. These were structural hypotheses about how researchers think about cohort construction, not interface styles to select from superficially.

In the Hudex case, the IA output included a project overview layer before deep AI clustering exploration. The project overview showed high-level theme counts, material-source counts, and orientation information before users entered the dondogram.

In the Puraite case, the IA output reduced top-level navigation from 13 items to 4 items. The four items followed the four stages of systematic review rather than the product's internal structure.

Evidence from documented cases

The Gexcon CFD simulation case provides evidence for IA restructuring in a single working environment. The system had to support 102 distinct tasks inside one environment. After the IA change, active users per team increased from 1 to 3–4, client-reported. Time to first successful simulation changed from 4 days to 6 hours and is recorded as a measured outcome in the case evidence.

The Polymatica OLAP analytics case provides evidence for replacing technical IA with domain-appropriate IA. Creative Navy explored 45 structural variants across 10 challenge areas before convergence. Independent task completion increased from 2% to 56%, measured via product analytics. The documented mechanism was that users who could not form a mental model of the system could not complete basic operations, while a structure matching the user mental model supported completion.

The Akrivia Health clinical research platform case provides evidence for IA that aligns analytical freedom with auditability. Creative Navy developed five interaction models, ran six design cycles from first wireframes to interactive prototype, and conducted eight usability sessions with realistic tasks. Client-reported evidence states that governance reviewers could verify cohort logic without escalating to the research team.

The Callsign fraud detection case provides evidence for IA as governance architecture. Fraud rules had been scattered across database views and configuration tables. Creative Navy's IA work clarified the separation between the AI scoring model and the policy layer, and made policy the central object bundling conditions, actions, history, and related rules. The documented case states that this governance structure was what risk teams at Lloyds and HSBC could evaluate under SCA and PCI DSS requirements.

The Hudex intelligence analysis platform case provides evidence for progressive disclosure in complex AI output. The dondogram was the primary entry point and was not intuitive to new users. Creative Navy introduced a project overview layer before the dondogram and iterated that overview 20 times, the highest iteration count for any single component in the documented portfolio.

The Puraite AI systematic review case provides evidence for process-structured navigation. Creative Navy reduced top-level navigation from 13 items to 4 items, aligning the IA with the four stages of systematic review. The client described the navigation restructuring as one of the most significant contributions of the engagement.

The OLX automotive marketplace case provides evidence for defining fixed and flexible IA across markets. The IA problem was that no structural definition existed for which journeys were fixed across all markets and which could vary. The marketplace coherence framework defined consistent journeys and explicit adaptation points.

Boundaries and limits of this capability

Information architecture for expert systems does not mean simplifying expert work by removing all complexity. The Gexcon case shows that some structural complexity is essential because it carries the scientific rigour of the tool. The IA task is to distinguish essential complexity from accidental complexity.

Information architecture work does not require separate beginner and expert modes when a single structure can support different speeds and depths. In the Gexcon case, the beginner/expert tension was resolved through one structured IA pattern with appropriate depth on demand, not through two modes or a simplified entry path.

Information architecture outcomes vary by evidence type. Some outcomes are recorded as measured, such as Gexcon's time to first successful simulation and Polymatica's product-analytics task completion change. Other outcomes are client-reported, such as Gexcon's active users per team, Akrivia Health governance reviewers verifying cohort logic without escalation, and Puraite's client description of the navigation restructuring.

Information architecture is not only a navigation-layer activity. In the documented cases, IA included object modelling, terminology, governance provenance, process structure, central orientation points, progressive disclosure, and market-coherence rules.

What this produces

Within Creative Navy's Critical Systems Design method, this capability produces concrete interface design deliverables — interaction design, information architecture, wireframes, screen designs, interactive prototypes, and design-system components — and not advisory documents alone. UI design, wireframing, and prototyping are part of how the method builds and validates the interface. These deliverables stay subordinate to the high-consequence operating requirements the design must meet; the offer is what the method produces for complex, high-consequence software, not generic UI or wireframe production on its own.