emberOS AI

emberOS 2025 AI Products UX Research & Service Design & Prototyping

Cutting a six-screen booking flow down to two.

Client: emberOS
Year: 2025
Role: Lead UX Designer
Duration: 14 weeks
Services: UX Research, Service Design, Prototyping

emberos set out to solve a problem most companies hadn’t named yet: large language models now describe your brand to customers, constantly, and nobody is managing what they say.

The product had to make that visible, predictable and fixable. Which meant designing an interface for something nobody had a mental model for. There’s no established pattern for “here is what AI thinks of you, here is where it’s heading, here is what to change” — no competitor to borrow from, no conventions users already hold.

It also had to carry five distinct agents — Scout, Pilot, Flow, Merchant and Echo — each doing different work against a shared Brand Knowledge Graph, without the product fragmenting into five tools wearing the same logo.

The challenge

emberos set out to solve a problem most companies hadn’t named yet: large language models now describe your brand to customers, constantly, and nobody is managing what they say.

The product had to make that visible, predictable and fixable. Which meant designing an interface for something nobody had a mental model for. There’s no established pattern for “here is what AI thinks of you, here is where it’s heading, here is what to change” — no competitor to borrow from, no conventions users already hold.

It also had to carry five distinct agents — Scout, Pilot, Flow, Merchant and Echo — each doing different work against a shared Brand Knowledge Graph, without the product fragmenting into five tools wearing the same logo.

The approach

Make the abstract measurable. The product’s core idea needed a number people could act on. Share-of-Prompt and the TAVI tiers — brand ownership, category dominance, intent expansion — gave the interface something concrete to show, and gave users something to move. Most of the design work followed from making those legible at a glance.

Design the uncertainty honestly. The platform predicts what AI will say next. A prediction shown with false precision destroys trust the first time it’s wrong, so forecasts had to read as forecasts — with the prediction record visible alongside them, so users could judge the system by its track record rather than by its confidence.

One loop, five agents. Rather than five separate products, the agents were designed as stages in a single loop: see, predict, shape, prove. A user should understand where they are in that loop at any point, not which tool they’re in.

Proof as a first-class object. “Foresight, with receipts” is the positioning, so the evidence trail — what was predicted, what was changed, what actually moved — needed to be a thing you could open and read, not a log buried in settings.

The outcome

Observable and verifiable things you could legitimately point to, if they were outcomes of your work:

The platform ships five agent products against one shared graph
A self-serve AI Brand Check as an entry point ahead of a sales conversation
Five pricing tiers from Lite to Enterprise, each a distinct product experience
emberos was selected for HorizonOS Labs Cohort 3 — 11 from 600+ submissions
Customers include Stagwell, Riot Games, Assembly and Monigle
SOC 2 Type 1