AI intelligence platform

ModelSight is an enterprise AI intelligence platform designed to help brands understand where they of sit across large language models.

timeframe

UX design, UI design, user flows, data presentation

The problem

As AI platforms become a growing source of product discovery and brand evaluation, businesses need visibility into how they're represented within AI-generated responses. However, the underlying data is inherently complex. Users needed to analyse multiple brands, markets, models, topics, prompt types and AI platforms simultaneously, often across large datasets and different reporting periods. Existing views made it difficult to understand performance drivers, compare competitors, or find meaningful insights without significant effort. The challenge wasn't simply displaying more data. It was helping users answer questions such as: - Why has visibility changed? - Which competitor is outperforming us? - Which AI platform is driving results? - Which market or product needs attention? - What actions should we take next? without overwhelming them with filters, tables and metrics.

Our hypothesis

If we could simplify how users explore complex AI performance data through intuitive filtering, clearer information hierarchy and contextual summaries, users would be able to uncover insights faster and make better decisions without needing specialist knowledge of the underlying datasets. By reducing cognitive load and surfacing the most relevant information first, we believed we could transform ModelSight from a reporting tool into a decision-making tool

What we did

Working closely with product managers, developers and stakeholders, I helped design an experience that balanced flexibility with usability.

Personas
Four personas, one platform: digitas admin, digitas data team (view-only), client admin, client (view-only) each with different permissions and dashboard views.

Data Design
Rather than dumping data all at once, we layered it including scorecards up top, trends and comparisons beneath, detailed tables for anyone who wants to go deeper.

Filtering
Users needed to filter across market, model, brand, topic, ai platform and time all at once. Instead of exposing everything up front, we built consistent filtering patterns that reduced invalid combinations and made the impact of each selection clear.