.png)
Goals
Johnathan Axwell
Johnathan is the CEO of GreenBox Manufacturing, with headquarters in London, a factory in the Midlands and a logistics fleet covering the UK. Jonathan runs a medium-sized enterprise producing sustainable packaging. His business is complex, he manages high energy consumption at the factory (Scope 2), a nationwide delivery network (Scope 1), and a massive web of material suppliers (Scope 3). As a London-based CEO, his main goal is to prove his company is Net Zero.
To reduce the company’s carbon footprint by 15%, without hiring extra accountants agencies
Frustrations
To speed up the processing time, adding all bills to the carbon reports in an easy way
A AI Agentic that acts as a carbon COO flagging anomalies when processing transactions
Data fragmentation nightmare, with data coming from very different sides, very difficult to process real-time data
Slow processing banking bills, fuel cards, supplier invoices and matching to carbon reports
Struggles with AI Agentic banking confidence, too important to take AI error risk
1
2
London based business owners repeatedly described the UK Sustainability Reporting Standards as "inaccessible" and "designed for scientists, not shopkeepers." The confusion between direct emissions (Scope 1) and indirect value chain emissions (Scope 3) creates a psychological barrier.
Business owners spend upwards of 20 hours a month manually cross-referencing paper receipts, fuel cards, and utility bills against carbon conversion spreadsheets, these hours are stolen directly from revenue generating activities. This manual process is not only exhausting but also highly susceptible to human error.
Context gap in banking data, fuels a deep mistrust in fully autonomous systems. While business owners are eager to offload manual work, they expressed significant anxiety about granting an AI Banking total control over their financial narrative. Banking transactions are often ambiguous, misclassify data, leading to legal inaccuracies or missed tax incentives.
3
4
Invisible emissions, where decentralised data such as staff working from home or fragmented commuting patterns, creates significant gaps in the reporting narrative. Business owners struggle with the lack of transparency in capturing these "hidden" emissions, as standard banking data doesn't account for the heating and electricity used in an employee's private residence or the specific mode of transport used for a business trip.
Cognitive Overload
Carbon Data Acuracy
Trust AI Banking
Carbon Workflow Leaks
The Core Challenge Is Balancing High-Stakes Automation With Absolute User Control
UX Challenge
In the 2026 UK regulatory landscape, business owners face severe legal and financial penalties for incorrect carbon reporting. Therefore, a "set-and-forget" AI cannot operate entirely in the dark. The UX challenge is to transition the user from a manual data-entry clerk to an executive AI editor without causing "verification fatigue."
Cognitive overload: Most business owners feel like they need a PhD, drowning themselves in manual data entry to understand and compliance with "Scope 1, 2 and 3" emissions reports. On average companies and directors spend 20+ hours a month tracking, understanding and processing their carbon footprint reports from energy bills, fuel receipts, travelling and supplier invoices.
The Bridge
The Bridge Is A Fintech And Climate Tech Hybrid Approach Solution, Combining Agentic AI & SME Carbon Emission reports Dashboard Systems.
Project Overview
New AI Way Of Reporting
SME’s Carbon Emissions
The Bridge combines Agentic AI & SME Carbon dashboard systems that don't just show data but take actions, like automatically audit enterprise bank transactions to generate legally required carbon footprint reports in real-time, categorises them into carbon scopes and drafts ready reports in a simple way.
The Problem
Carbon Emissions Reports Can Be Tricky And Challenging
Streamlined energy and carbon footprint reporting comes with unique challenges:
5 months
June 2026
Role: UX Lead
Carbon data accuracy: The manually cross-referencing and spreadsheet conversions, adds confusion between direct emissions in scopes 1, 2 and 3. This manual process is not only exhausting but also highly susceptible to human error, that creates a psychological barrier when categorizing bank transactions.
Overwhelming regulations: Evolving environmental legal carbon emission reports requirements (such as bidding criteria for public sector contracts) require detailed carbon reduction plans, imposing high administrative costs on smaller businesses.
Carbon emissions workflow leaks : Invisible emissions, where decentralised data such as staff working from home or fragmented commuting patterns, creates significant gaps in the reporting narrative. Business owners struggle with the lack of transparency in capturing these "hidden" emissions, as standard banking data doesn't account for the heating and electricity used in an employee's private residence or the specific mode of transport used for a business trip.
The Goal
Agentic AI Bank & EMS Carbon Footprint Reports
An Agentic AI dashboard that connects to enterprise banking, audits transactions in real-time, categories them into carbon heavy spending scopes, identifying the key wins per month to reduce their carbon footprint and generates a compliance ready to go report for government regulations and stakeholders in one click.
Understanding The User
User Research Summary
The objective of this research is to understand how UK SME's interpret, manage, and report emissions data under:
( Streamlined Energy and Carbon Reporting and UK Sustainability Reporting Standards )
The focus of my research was to find awareness gaps, reporting challenges, tool usability needs and decision making behaviour of the users. I employed a mixed methods approach to validate if business owners would actually trust an AI "Agent" to handle their legal reporting, and to find previous real life cases on how carbon emissions have been reported by enterprises.
Secondary research (desk research)
Sources included:
-
Analysis of British Business Bank sustainability reports and Gov.uk SECR guidance
-
UK Department for business and trade reports
-
Carbon Trust SME studies
-
CDP supply chain reports
Competitive, real-world tool analysis, qualitative research and case studies reviews (real examples)
I reviewed platforms used by SMEs:
-
Normative
-
Sweep
-
Persefoni
-
Unilever
-
Tesco
Initial Assumption
Assumption 1: Users want a "magic" button
that does everything in the background
Research Findings
Reality: Users felt anxious when the AI worked in secret. They feared being fined for AI wrong operating process
Design Impact
Added : Added an "agent Log" to show exactly which receipts the AI scanned and why it categorised them
Assumptions VS. Reality
Changing Ideas After Conducting Research
The pivot moments that changed the directions of my design, when challenged by real-life feedback.
Assumption 2: A "Conversational AI" (chat) is the best way to interact with the Agent
Reality: "chat fatigue." Users found typing questions tedious. They preferred "actionable notifications" that they could swipe or tap
Replaced the primary Chatbot interface with a "smart feed." The UI now behaves more like an Instagram feed of "audit tasks" than a WhatsApp conversation
Assumption 3: Business owners want deep educational content on ESG (environmental, social and governance)
Reality: They have zero time. They want "compliance in a box" with minimal reading
Replaced long articles with AI-generated "action snippets" (30-word summaries)
User Pain Points
The Specific User Needs
Understanding the most important needs by discovery and definition.
%201.png)
Carl is a business owner who has spent 12 years building a reliable logistics and van hire for construction services in South London. His primary aim is to automate the "paperwork nightmare" of carbon reporting so he can focus on his crew and his customers, ensuring his business remains a compliant, top-tier partner for the big corporate contracts that keep his vans on the road by deploying AI Agentic to do the heavy lifting for him.
Occupation: Business owner
Family: Married 3 children
Home Town:
London (UK)
Education:
Civil Engineer
Age:
45
Continue running his business more efficiently and gaining new contracts
Making strategic SME's reports decisions without manual complex math
Investing his valuable time in further development of his company
Processing thousands of paper receipts and invoices manually susceptible to human error
Complex to understand brake down of SME's reports difference between scopes 2 and 3
Using generic mixed off date formulas that derail the entire years footprint reports, confusing company’s expenses
Carl Murray
Goals
Frustations
Personas
Ideate The Solutions For The User Problems
Starting The Design
Paper Wireframes, The Initial Representation Of My Design Ideation Process
UX Design
The SME Command Center
The final design takes the approval stack (for Speed), the carbon ledger (for Trust), and the supply chain radar (for Scale) and merges them into a single, high-impact interface.
Digital Wireframes
The Core UX Philosophy Behind These Digital Wireframes Is Delegation Without Loss Of Control
These digital wireframes focuses on a dual-stream layout. This design pattern ensures the AI handles 99% of the administrative burden, keeping the human in control as the "Director" without overwhelming them.
Design System
Visual And Branding Orientation Of My Design
Crafting the look and feel for optimum user results, based on the visual design evolutions form UI to the high contrast tones elements and components.
Low-Fidelity Prototype
Foundational Layout That Defines How Complex Data Is Simplified Into An Accessible, Agentic Experience
The low-fidelity prototype strips away the aesthetic layer (gradients, maps, and illustrations) to isolate and test the core interactive skeleton. It establishes a grid system that maps cleanly to assistive technologies while prioritizing high scannability for busy SME operators.
1
Wireframe Goals
Executive summary & quick actions
The UX goal: Eliminated complexity and user anxiety by immediately showing accurate regulatory reports.
Layout structure: A dominant full width header block reads "London SME’s report", text blocks indicating "carbon footprint reports” “Scope 1, 2 and 3” for real-time emissions with visual graphical data, and “ask AI agent” for agentic AI solutions.
The agentic AI sidebars: The AI Agent's task boxes, surfaces a prominent "approve flagged" button and a "generate SECR PDF reports" when legally safe for users, and executes pending approvals with a single tap, allows users to find details breakdown of carbon emissions for scopes 1,2 and 3.
2
The agentic AI sidebars: The AI agent's task boxes, are conversational container doesn't wait for input; it states its current activity: "scanning march data. Found 3 items requiring business context." It prompts the user with binary choice micro buttons (e.g., [office expense] / [client project]), instantly turning messy banking data into an auditable legal report.
Banking Transactions & Verification
The UX goal: The Bridge the "context gap" in banking data through transparent AI categorisation of transactions and legally generates safe legally required carbon government footprint reports.
Layout structure: The main 70% stream functions as an open banking accounts synchronisation and transaction ledger (e.g. Shell Fuel - £450). Next to each transaction is a lozenge indicating the AI’s classification (e.g., Scope 3: Category 1) or an amber/red indicator for [AI Flagged - unsure], alert.
3
Fleet carbon usage breakdown,
supply chain & logistics map radar
The UX goal: Map fragmented logistical data while plugging "carbon workflow leaks." It displays real-time emissions with graphical visualisation data and creates specific simulating solutions depending on each individual scenario for best business solution.
Layout structure: A centralised map widget displays across London, UK shipping corridors, coloured coded by carbon category, intensity and business units. Adjacent to the map is a vertical supplier leaderboard stacking with each fleet staff representative and fleet details for real-time accurate carbon emissions usage and their carbon efficiency scores.
The AI agent box serves predictive warnings based on real-time data shifts: "midlands factory line energy spike detected. Suggest shifting logistics to rail to save £500 in carbon penalties. Run simulation?" It features primary actions.
Usability Study Findings
The Findings Behind The User Usability Study
This usability study findings shows how the design iterated on user testing with Carl and Jonathan and other London’s SME owners and managers.
1
2
3
Users were terrified to hit "submit" because they didn't know if the report would actually pass their corporate clients' ESG procurement portals and government legal requirements.
Busy fleet owners abandoned the screen if they had to type or click through multiple drop-down menus while out on a work site.
One-Click to Legal Export & Verification
2
Bank Sync & Trust Deficit
The Complexity of Compliance Reports
The Approval of AI Flagged Transactions
1
Round 1 Findings
Tracking Invisible Work-Form-Home Leaks
When the AI automatically estimated home-energy emissions without warning, business owners felt the data was "made up" and legally risky.
While the map looked impressive, it was purely passive. Users expressed frustration that they couldn't immediately act on a "red zone" route or supplier.
Users like Carl experienced instant cognitive overload when seeing terms like "Scope 3 value chain." They spent minutes trying to figure out what to click.
SME owners hesitated at the connection screen. They feared giving an AI "control" over their financial data or that it would make mistakes without their knowledge.
Mapping Regional Logistics
Round 2 Findings
3
Mockups
The Mockups: Evolution Of The Dashboard
The dashboard design evolves through user-interface stages, showing a transition from a rich data visualisation platform to a fully accessible, compliance ready enterprise tool.
High-fidelity Prototype
The interactive AI Agentic Workspace
The high-fidelity prototype brings the tactical user experience to life by simulating
real-time environmental data streams and automated workflows. It serves as the definitive, click-tested validation of the product’s architecture before handing off to development.
.png)
Accesibility Considerations
Fundamental Rules For Accesibility
An analytical visual screens has been generated to demonstrate how the interface adheres to strict accessibility standards through an optimised reduction of:
1 Cognitive load via natural language,
2 Keyboard navigation & focus indicator predictability
3 High contrast semnatic colour schemes for low vision users& complex data accesibility
Ensure the dashboard is entirely operable without a mouse, accommodating individuals with permanent motor disabilities, repetitive strain injuries (RSI), or power users who rely solely on mechanical switches and keyboard interfaces. Every interactive data node, AI chat response button, and Open Banking ledger row is mapped into a logical, sequential focus order that strictly follows the spatial arrangement of the page layout (from top-left to bottom-right) Active elements feature a highly visible contrast colour, against the rest of the layout, Additionally, invisible "skip to content" hidden links are embedded at the root of the DOM, allowing users using the Tab key to bypass the complex Agentic Log ticker and jump straight to the central transaction verification grid.
Accessibility considerations ensure that complex visualisations such as maps, charts, and analytics are supported by text colour-blind-friendly indicators so that all users can access and interpret the data effectively. To counter this, every status indicator combines distinct geometric shapes, explicit text label icons, and a colour contrast to all of the backgrounds and text of the dashboard, this guarantees that users checking their dashboard in bright light can easily distinguish a critical alert from a completed task.
Accesibility Considerations
To ensure the platform is accessible to neurodivergent users and business owners experiencing high stress or cognitive fatigue, the UI actively minimises jargon and technical complexity. Standard carbon accounting platforms bombard users with abstract legal terminology which can cause immediate executive dysfunction or cognitive overwhelm. The dashboard addresses this by utilising a Natural Language Layer that translates complex compliance frameworks into plain, actionable English. This layout strategy reduces the working memory required to complete an audit, making mandatory legal reporting equitable for users with ADHD, dyslexia, or varying levels of functional literacy.
1
Reducing Cognitive Load via Natural Language
Keyboard Navigation & Focus Indicator Predictability
2
3
High-Contrast Semantic Colour Schemes for Low-Vision Users, and Complex Data Accessibility
Efficiency, Automation & Sutainability
Takeaways
The Business & Environmental Impact
The implementation of the EcoDirector AI Command Center fundamentally shifts how SME's approach sustainability from a dreaded compliance chore to a frictionless, automated asset.
90% Reduction in Cognitive Load & Manual Labor: By combining Open Banking data feeds with proactive AI agents, the platform entirely replaces the need for spreadsheet hunting. The user changes roles from a tedious data-entry clerk to an overseer—simply reviewing and approving pre-categorized tasks.
Corporate Contracts: For users like Carl, the platform safeguards their primary revenue stream. Having an audit-ready, UK SRS-compliant report ready in clicks gives tier-1 corporate clients the precise data they demand, cementing the SME’s status as a preferred supplier.
Strategic Profit Protection: For scale-up CEOs like Johnathan, the dashboard acts as a digital "Carbon COO." By mapping supply chains and logistics corridors in real-time, it moves carbon accounting out of historical archives and into forward looking strategy instantly flagging energy leaks and modelling real-time financial ROI on electric shifts or energy cunsumption alternatives.
What I learned
Traditional AI applications often wait for a human user to type a prompt. This case study proved that for busy SME owners, a blank chat box is a barrier. The design succeeded by shifting the interaction model to proactive delegation. The AI does the heavy lifting in the background and presents the user with concrete, binary choices (e.g., [Approve] / [Edit]), drastically cutting time-to-task.
When dealing with massive datasets like nationwide logistics operations or complex accounting regulations (Scopes 1, 2, and 3), the designer’s primary job is curation. Grouping information into a 70/30 dual-stream layout data feed vs. action queue proved that users don't need to see all the data at once ,they just need to see what requires their attention right now.
Designing for accessibility in this project demonstrated that strict accessibility guidelines of high-contrast states layouts directly improve usability for all users. A fleet owner trying to verify an expense on a bright light relies on the exact same high-contrast and hit-target architecture designed for a low-vision or motor-impaired user. Inclusive design is simply good product design.
"This project proved to me that great design doesn't just look beautiful,
it bridges the gap between complex data and real human capability, making high stakes sustainability accessible to every single user."
Let's connect!
Cognitive Overload
Carbon Data Acuracy
Trust AI Banking
Carbon Workflow Leaks
Frustrations
Carl Murray
Frustrations
Jonathan Maxwell
Time to add value proposition by implementing the best possible solution.
It displays four initial paper wireframes showcasing different 'Agentic AI' UX patterns, from a simple dashboard interface to a complex scenario builder. The layout culminates in the final "SME Command Center" design direction, which merges the best concepts into a single, comprehensive dashboard that balances Carl's need for simplicity with Johnathan's need for strategic insight.
.png)
.png)