Global Web Index.
Jan 2026-present
GWI was evolving from a service-led operating model toward enterprise self-serve AI, supporting clients including Meta, Amazon and Omnicom.
As adoption grew, fragmented permissions, user management workflows and operational processes created increasing complexity across the platform. Critical tasks still relied on internal teams, limiting scalability and slowing the transition to self-serve.
I led discovery, workshops, system modelling and prototyping to align teams around a shared direction and define the governance foundations for scalable enterprise self-serve AI.
Impact snapshot
Questions nobody could answer confidently
When I joined, there were fundamental questions nobody could answer confidently:
- Why had certain features been built?
- What assumptions had actually been validated?
- Who were the primary users?
- How did systems connect behind the scenes?
Over time, workflows had evolved across multiple teams and codebases, creating complexity that lived largely in people’s heads.
Before designing new experiences, I needed to understand the system itself.
Current Organisations management in GWI Admin: the operational surface before hierarchy and access foundations were prioritised.
Before designing anything, the priority was to create clarity.
Building a shared understanding
I worked across Product, Engineering, Customer Success, Legal, Analytics and AI teams to understand how enterprise user management operated in practice.
Through workshops, interviews and stakeholder mapping, I created a shared view of the ecosystem and surfaced previously hidden dependencies.
Discovery artefacts
Stakeholder mapping
Who owned which workflows, who needed to be involved and who needed to be informed.
Proto-persona workshop
Lightweight profiles to give us direction on information to validate with users.
Interviews
13 interviews across RevOps, Enterprise CS, Product, Engineering, AI, and Legal.
Service mapping
How work moved between systems and owners.
“A lot of it lives in our heads.”
Mapping a fragmented ecosystem
Research revealed that enterprise customers had evolved organisational structures, ownership models and access requirements that weren't represented within the product.
Organisations, teams and users lacked clear relationships
Access and permissions relied on operational workarounds
Critical workflows depended on manual intervention
Business impact
The operational model created significant overhead across the business, making self-serve increasingly difficult to scale.
- ~500 Hours spent on admin support in six months
- £36m Enterprise revenue affected by operational dependencies
- 100% Manual user-management changes
Modelling the system before designing the interface
Many of the challenges weren’t caused by interface design.
They stemmed from unclear relationships between organisations, users, permissions and access models.
To create a shared language across teams, I mapped the entities and relationships underpinning future self-serve experiences.
Future vision
Future admin system, rapidly imagined with AI tools
- Parent–child hierarchy — a clear roll-up view of accounts
- A way to view and manage seat capacity
- An overview of users and access within organisations
- A way to track subscriptions to spot opportunities for engagement and reduce churn
From future-state vision to Alpha reality
As priorities evolved, the focus shifted toward launching an Agentic Alpha product within ~two months,
Rather than delivering the entire future-state vision, I identified the minimum governance foundations needed to support safe self-serve experiences.
In scope
- Self-serve for the new product
- Role-based access
- User access management
- API / MCP usage and metering
Out of scope
- Full system redesign
- Enterprise-grade organisational modelling
- Backend-heavy dependencies
Design
The Alpha experience focused on enabling core self-serve workflows while maintaining appropriate governance and ownership.
Invite users flow · currently in development.
Balancing safety with simplicity
As AI capabilities expanded, governance became increasingly important.
I explored multiple approaches to permissions and guardrails, balancing flexibility, usability and implementation complexity.
Ideation
Idea 1
Audiences created and managed entirely within Octopus.
Idea 2
Soft guardrails: Recommended and Approved labels to guide selection without blocking access.
Idea 3
Separate library and catalogue views with explicit add and remove from library.
Final direction
Using AI to move faster
AI tools were used to support speed and understanding:
- Rovo AI (Atlassian) To uncover buried engineering decisions
- Secoda AI To understand gaps between systems and business costs
- ChatGPT Create an agent to store research artefacts and synthesise findings
- Figma Make Quick ideation and concepts
- Cursor To build prototypes with Figma MCP and push to repo
Developing a design system that’s not just good for humans but good for machines.
Alongside those tools, I worked with the design team to evolve a staging design system as the foundation for a future product-wide rollout, so we could move fast prototyping with AI. This involved structuring tokens and components in a way that was readable by AI agents and linking to documentation and guidelines.
Tokens · variants · components
Establishing a foundation for scalable self-serve
Self-serve account settings
Designing systems, not just interfaces
This project went beyond designing screens. It focused on:
- Creating clarity in a complex system
- Balancing long-term architecture with short-term delivery
- Aligning teams with different priorities
The work is ongoing, but it established a strong foundation for scaling GWI’s platform.
