Leicester, UK /
3:45 PM
Leicester, UK /
3:45 PM
AI Energy Platform (NDA)
Redesigned how building managers read energy performance.
Redesigned how building managers read energy performance.
replacing a wall of numbers with one clear score.
The context
"A wall of numbers. One clear answer."
Managers didn't want more data. They wanted one clear answer.
I was the sole product designer on a B2B AI platform helping commercial building owners manage energy cost, consumption, and carbon performance. The early direction leaned on charts and technical metrics, accurate, but not useful to the people who had to act on them.
A handful of user interviews reframed the entire brief. That insight shaped everything that followed: a single building score, budget-aware onboarding, and AI insights written in plain English instead of raw numbers.
I built the design system that made all of it possible — the decisions behind each piece are what follows.
I was the sole product designer on a B2B AI platform helping commercial building owners manage energy cost, consumption, and carbon performance. The early direction leaned on charts and technical metrics, accurate, but not useful to the people who had to act on them.
A handful of user interviews reframed the entire brief. That insight shaped everything that followed: a single building score, budget-aware onboarding, and AI insights written in plain English instead of raw numbers.
I built the design system that made all of it possible — the decisions behind each piece are what follows.
ROLE
Solo Product Designer
TEAM
Product manager & engineering
TIMELINE
6 months to MVP
PLATFORM
Web app (B2B SaaS)
DOMAIN
AI · Energy · Real estate
ROLE
Solo Product Designer
TEAM
Product manager & engineering
TIMELINE
6 months to MVP
PLATFORM
Web app (B2B SaaS)
DOMAIN
AI · Energy · Real estate
Tools
Tools
THE CHALLENGE
"More dashboards weren't the answer"
Commercial building owners and managers were sitting on huge amounts of disconnected data - energy cost, consumption, equipment readings - with no reliable way to tell whether any of it meant something was actually wrong.
The product brief I inherited leaned into that instinct: more charts, more metrics, more visibility. My job was to design the interface for an AI system meant to help people manage their buildings better. The early direction risked adding to the noise instead of cutting through it.
Commercial building owners and managers were sitting on huge amounts of disconnected data - energy cost, consumption, equipment readings - with no reliable way to tell whether any of it meant something was actually wrong.
The product brief I inherited leaned into that instinct: more charts, more metrics, more visibility. My job was to design the interface for an AI system meant to help people manage their buildings better. The early direction risked adding to the noise instead of cutting through it.
THE APPROACH
"I know my cost. Tell me what to do about it."
Before designing anything, I ran interviews with the people who'd actually use the product day to day — building managers and portfolio owners, not engineers. Two moments changed the direction of the whole project.
Two reactions kept coming up, in different words, from different people. Managers already knew their costs — what they wanted was for the system to tell them what they should be paying, and how far off they were. And more than once, walking someone through a screen full of technical metrics, the response amounted to the same two words: so what? Not hostility — genuine confusion about what a number meant for their building or their budget, unless someone explained it.
Before designing anything, I ran interviews with the people who'd actually use the product day to day — building managers and portfolio owners, not engineers. Two moments changed the direction of the whole project.
Two reactions kept coming up, in different words, from different people. Managers already knew their costs — what they wanted was for the system to tell them what they should be paying, and how far off they were. And more than once, walking someone through a screen full of technical metrics, the response amounted to the same two words: so what? Not hostility — genuine confusion about what a number meant for their building or their budget, unless someone explained it.
THE PROCESS
"From seven metrics to one entry point"
Those two conversations reframed the design direction completely. Instead of surfacing every available metric with equal weight, I proposed a single number — a Building Score — as the entry point to the whole product: one answer to the question every manager actually had, how are we doing? The score combined cost against budget, consumption against an AI-generated baseline, and carbon and operational performance into one figure.
But a B2B energy platform isn't only used by someone glancing at it once a day. It's also often left running on a screen in front of building operators monitoring conditions continuously. Designing one interface for both of those attention spans — a ten-second check and an all-day watch — became the real problem behind the problem.
I also ran usability sessions specifically on the AI-generated insights, testing plain-language explanations against raw technical output. The pattern held every time: people trusted and understood the system more when it explained why something mattered, not just what the number was.
Those two conversations reframed the design direction completely. Instead of surfacing every available metric with equal weight, I proposed a single number — a Building Score — as the entry point to the whole product: one answer to the question every manager actually had, how are we doing? The score combined cost against budget, consumption against an AI-generated baseline, and carbon and operational performance into one figure.
But a B2B energy platform isn't only used by someone glancing at it once a day. It's also often left running on a screen in front of building operators monitoring conditions continuously. Designing one interface for both of those attention spans — a ten-second check and an all-day watch — became the real problem behind the problem.
I also ran usability sessions specifically on the AI-generated insights, testing plain-language explanations against raw technical output. The pattern held every time: people trusted and understood the system more when it explained why something mattered, not just what the number was.
THE SOLUTION
"Simple at a glance. Complete on demand."
The score became the entry point, not the whole product. Above the fold, the homepage brought together everything a manager or operator needed in one view:
The score became the entry point, not the whole product. Above the fold, the homepage brought together everything a manager or operator needed in one view:
An interactive, colour-coded map, letting the same screen represent a single building or scale up to a whole portfolio of sites — an owner managing several properties could see at a glance which ones were fine and which needed attention.
The Building Score, front and centre, as the instant answer.
Live data across every utility type — electricity, water, chilled water, gas — each shown with a short-range forecast, so a spike could be seen coming rather than discovered after the fact.
An alerts panel graded by severity (critical, major, minor), so attention went to what mattered first.
An interactive, colour-coded map, letting the same screen represent a single building or scale up to a whole portfolio of sites — an owner managing several properties could see at a glance which ones were fine and which needed attention.
Live data across every utility type — electricity, water, chilled water, gas — each shown with a short-range forecast, so a spike could be seen coming rather than discovered after the fact.
The Building Score, front and centre, as the instant answer.
An alerts panel graded by severity (critical, major, minor), so attention went to what mattered first.
Below the fold, the detailed historical charts and figures remained available for anyone who wanted to dig deeper — the same screen worked as a quick daily check for an owner and a continuous monitoring tool for an operator, without either one fighting the other for space.
Below the fold, the detailed historical charts and figures remained available for anyone who wanted to dig deeper — the same screen worked as a quick daily check for an owner and a continuous monitoring tool for an operator, without either one fighting the other for space.
I also designed budget-aware onboarding, where users set their own cost and consumption targets up front, so every insight afterwards was measured against a number the user actually cared about — and rewrote the AI-generated insights to read like a colleague explaining a problem, not a system logging a value.
I also designed budget-aware onboarding, where users set their own cost and consumption targets up front, so every insight afterwards was measured against a number the user actually cared about — and rewrote the AI-generated insights to read like a colleague explaining a problem, not a system logging a value.
Underpinning all of it, I built the design system from scratch — components, tokens, and patterns that let the product scale across this many data types and views without every new screen being designed from zero.
Underpinning all of it, I built the design system from scratch — components, tokens, and patterns that let the product scale across this many data types and views without every new screen being designed from zero.
THE IMPACT
"A system that outlasted the sprint"
DAYS
DAYS
The design system cut the design phase of each feature sprint from two weeks down to four days, giving the team significantly more time for development and testing. The first release shipped in six months.
The design system cut the design phase of each feature sprint from two weeks down to four days, giving the team significantly more time for development and testing. The first release shipped in six months.
Looking back, the clearest lesson was how much of the project's success came before a single screen was drawn. The two interviews that reframed the brief mattered more than any interface decision that followed. If I did it again, I'd push for that kind of grounding research even earlier — before a spec exists at all, not after.
Looking back, the clearest lesson was how much of the project's success came before a single screen was drawn. The two interviews that reframed the brief mattered more than any interface decision that followed. If I did it again, I'd push for that kind of grounding research even earlier — before a spec exists at all, not after.