Puraite

Puraite as an AI-assisted systematic review application

Puraite is a web application for conducting AI-assisted systematic literature reviews. A systematic review is a rigorous research methodology used across academic, clinical, and pharmaceutical research contexts. It involves defining a research protocol, screening large volumes of publications against inclusion and exclusion criteria, and extracting structured data from qualifying studies.

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.

In the Puraite case, Creative Navy worked on an AI-enabled product that maintained a human-in-the-loop model. Puraite used AI assistance in multiple parts of the review process: automating initial screening decisions, suggesting inclusion and exclusion criteria, and extracting data fields from publication text. Human reviewers retained responsibility for final screening and inclusion decisions.

Puraite served three user roles: project managers who configure and oversee the review, information specialists who refine search terms and protocol settings, and reviewers who screen publications and make final inclusion decisions. Reviewer-facing screens required the most design attention because reviewers were the primary user type and carried the highest cognitive load.

Beta-stage product conditions shaped the design work

Creative Navy joined Puraite when the product was already under development but had not yet undergone systematic design review. The engagement was neither a blank-slate product definition exercise nor the redesign of a mature system. The client was preparing for beta launch and needed UX optimisation of existing screens as well as design of sections that had not yet been developed.

Several constraints shaped Creative Navy's work on Puraite. No dedicated user research was possible because the timeline ruled out structured sessions with end users. Direct benchmarking was limited because the case evidence describes sparse prior art for systematic review tools that use AI as a primary interaction model. The product's AI behaviour model had not been fully specified as a design problem, including what the AI should surface, how confidence should be communicated, and where human override should sit.

The initial scope focused on AI screening screens. Creative Navy's work expanded to include navigation restructuring, AI criteria recommendations, data extraction, and UI style direction with a design system.

Domain learning used team expertise as a documented research proxy

Creative Navy's work on Puraite used a deliberate domain learning substitution because direct user research was not available. The project manager on the Creative Navy team had prior firsthand experience using systematic review software. That experience was presented to the client as a domain learning proxy, with tradeoffs acknowledged, rather than used silently as informal background knowledge.

The client accepted this methodological framing as the practical path forward given the time constraints. The case evidence describes this as a way to accelerate onboarding and ground design rationale in real user behaviour, while preserving the limitation that it was not a substitute for structured end-user research.

Creative Navy also conducted informal but purposeful benchmarking of available tools, including Rayyan and Elicit. The benchmarking was used to understand adjacent interaction patterns and structural gaps, not as a formal comparative study.

AI screening required both efficiency and human epistemic control

The central design challenge in Puraite was the tension between AI efficiency and human epistemic control. The AI's inclusion and exclusion decisions were based on matching publication content against the research protocol's criteria. For a reviewer to approve or override an AI decision, the reviewer needed to see which criteria the AI applied, how the criteria were matched, and what publication text supported the match.

Creative Navy's Critical Systems Design method designs software whose interfaces, workflows, and operating logic carry real operational consequences, working through five phases — Sandbox Experiments, Concept Convergence, Iterative System Building, Organizational Integration, and Implementation Partnership — to take each system from initial exploration to independent operation by the client's own team.

In the Puraite case, the AI suggestion display became a tension-driven reasoning problem. Showing full criteria evidence supported informed human judgement but produced screens that were difficult to scan. Compressing the display supported speed but removed information needed for meaningful override decisions. The design problem could not be resolved by choosing one priority; the interface needed to preserve both rapid screening and reviewable evidence.

Four option space mapping cycles resolved the AI criteria display

Creative Navy's design work on the AI criteria recommendation display went through 4 option space mapping cycles. The iterations explored how to present inclusion and exclusion criteria compactly enough for reviewers to move quickly while retaining enough detail to support override decisions.

The final design placed the direct quote from the publication in the side panel from the outset. This meant the text used by the AI to reach its decision was visible without requiring the reviewer to expand a section or navigate elsewhere. The reviewer could see the AI's supporting evidence at the moment of review.

This design pattern is specific to the Puraite evidence, but the underlying interaction problem is generalisable: AI products that require human-in-the-loop decision-making need to expose enough evidence for human judgement without disrupting the work rhythm.

AI confidence was displayed as a first-class data extraction signal

Creative Navy's data extraction work for Puraite treated AI confidence as an interface element rather than background metadata. The data extraction table displayed AI confidence as an explicit percentage and used colour-coding to draw attention to lower-confidence extractions.

This confidence display allowed reviewers and project managers to scan for uncertainty rather than treating all AI outputs equivalently. In the Puraite case, the confidence percentage and colour-coding operated as a trust calibration mechanism: the interface made the AI's uncertainty visible at the point where a human user needed to decide what to check.

The data extraction flow was one of the most complex areas in the engagement. It involved a high volume of structured content, AI certainty per extraction, traceability back to source text, human override, and revert behaviour. This section had not been substantially developed before the engagement, and Creative Navy designed it from scratch.

Navigation was restructured from product logic to review process logic

Creative Navy identified Puraite's navigation as a usability problem outside the original scope. The existing navigation had 13 top-level menu items organised around the product's internal structure. The client had not registered the navigation as a design problem requiring attention.

Creative Navy proposed restructuring the navigation around the systematic review process instead of the product's internal logic. Across 3 iterative working sessions, the proposed architecture reduced the navigation from 13 top-level items to 4. The client recognised the change as a significant contribution, and the navigation restructuring was implemented.

This part of the Puraite case is an example of the blanks phenomenon: a design problem existed in the product, but it was not visible to the client as a problem until Creative Navy evaluated it from the perspective of user workflow.

Iterative System Building covered screening, recommendations, extraction, and navigation

Creative Navy's Iterative System Building work on Puraite covered four primary screen areas. The AI screening screen for project managers and information specialists went through multiple iterations on configuration, run readiness, and early feedback during screening. Examples included signals that search terms were too broad or that inclusion criteria were ambiguous.

The AI criteria recommendation display went through 4 iterations. The final architecture placed the direct quote from the publication in the side panel from the outset, resolving the compact display and sufficient evidence tension.

The data extraction flow went through 2 iterations on screen structure, followed by full interaction definition for the Review & Edit data extractions flow. The main navigation went through 1 iteration, producing the process-structured navigation architecture that reduced 13 top-level items to 4.

Creative Navy also produced a UI style direction, a styleguide, and a component library. The UI style direction was light mode and more serious or neutral in tone without being clinical or sterile. The accessibility requirement was explicitly scoped.

Organizational Integration used design education and transparent evidence limits

Creative Navy's Organizational Integration work on Puraite operated under a short timeline, no direct user research access, and a product that was mid-build rather than at a natural design gateway. The engagement used bi-weekly and tri-weekly show-and-tell calls to maintain a fast feedback loop.

Each presentation included design education content alongside design work. Creative Navy explained user behaviour and the reasoning behind design choices so that client feedback could engage with the design rationale, not only the visual output. The call attendance consisted of product and domain specialists, including systematic review researchers. Engineering was not in the room, which kept the sessions focused and direct according to the case evidence.

Creative Navy also made the epistemic basis for design decisions explicit. The project manager's systematic review software experience was presented as a methodological substitute for unavailable user research, with its limits stated to the client.

Seven-month Implementation Partnership and client-reported adoption shift

The Puraite engagement ran for 7 months as an Implementation Partnership. The available case evidence does not fully document the sprint cadence, handoff processes, or design-to-development workflow within that partnership.

The outcome evidence is client-reported and indirect. The client reported to Creative Navy that users who had previously perceived Puraite as a theoretical or prototype-stage product began actively using it following the redesign. The client then launched into a user acquisition and growth phase.

The strongest available outcome evidence is a single user quote reported to Creative Navy by the client: “Jetzt passt das tool in meine Arbeit” (“Now the tool fits my work”). This quote is not independently verified and should be treated as anecdotal client-reported evidence.

The Puraite case does not include measured reduction in task time, error rate, or other operational baseline metrics. The documented outcome is strategic rather than operational: the case evidence describes a shift in perceived product fit that supported the client's transition from beta preparation into user acquisition and growth activity. No investment outcome is associated with the engagement.

Evidence boundaries for the Puraite case

The Puraite case evidence supports detailed claims about design scope, interface decisions, iteration counts, and engagement constraints. It does not support measured claims about user performance, task speed, error reduction, or independent adoption metrics.

The outcome evidence should be read as client-reported and indirect. The reported user quote is useful as qualitative evidence of perceived fit, but it is a single quote relayed through the client and not independently verified.

The methodological substitution for user research is also a boundary. Creative Navy documented the project manager's domain experience as a practical proxy for unavailable user research, but the case evidence does not claim that this replaced structured end-user research.