[Case 01]

Accenture Gen-Ai Navigator

Enterprise Internal Application
[ Impact ]
Designed an internal GenAI platform adopted by 200+ SMEs within 3 weeks of launch directly enabling $2M+ in new client pitches and positioning Accenture's AI practice for a $3B investment.
Designed an internal GenAI platform adopted by 200+ SMEs within 3 weeks of launch directly enabling $2M+ in new client pitches and positioning Accenture's AI practice for a $3B investment.
[Project Overview]

Accenture's internal knowledge ecosystem is vast accelerators, solution frameworks, case studies, benchmarks, SME insights accumulated over decades. But that knowledge was fragmented across hundreds of tools and repositories, locked behind manual search, and dependent on individual tribal knowledge.

For every new client pitch, consultants were effectively starting from scratch: searching the same repositories, relying on the same overbooked SMEs, and rebuilding context that already existed somewhere in the system. World-class expertise was being lost to logistics.

[My Role]

Senior Product Designer

[Platforms]

Internal Enterprise Platform

News

News

[Problem Statement]
How might we design an AI-powered navigator that helps Accenture teams rapidly curate, contextualise, and present solutions for clients using both the internal knowledge base and generative AI
Consultants and subject matter experts operate in fast-paced, high-stakes client environments where attention is limited and decisions are made in real time. While Accenture had a wealth of internal knowledge library, this information was fragmented across multiple systems, decks, and tools, making it difficult to access when it mattered most.

Existing solutions required too much searching, interaction, and cognitive effort, often pulling experts away from the conversation instead of supporting it. The problem was not a lack of information, but the lack of a clear, usable way to surface the right insights at the right moment without breaking the flow of client engagement.
Consultants and subject matter experts operate in fast-paced, high-stakes client environments where attention is limited and decisions are made in real time. While Accenture had a wealth of internal knowledge library, this information was fragmented across multiple systems, decks, and tools, making it difficult to access when it mattered most.

Existing solutions required too much searching, interaction, and cognitive effort, often pulling experts away from the conversation instead of supporting it. The problem was not a lack of information, but the lack of a clear, usable way to surface the right insights at the right moment without breaking the flow of client engagement.
Knowledge spread across hundreds of tools with no unified entry point
Multiple clicks, duplicated workflows, inconsistent outputs per team
No way to translate internal knowledge into client-ready narratives in real time
Heavy dependency on SMEs for every discovery cycle
[ Users & Context ]

The primary users were senior ( Subject Matter Experts ) SMEs and executives responsible for leading client conversations, identifying pain points, and proposing tailored solutions. These users typically operate in high-pressure environments such as live meetings and workshops, often while managing multiple tools, documents, and inputs simultaneously.

In this context, the platform was not a primary workspace but a supporting system used alongside presentations, notes, and real-time discussion. Speed, clarity, and trust were critical, while complex interactions or dense visual elements risked becoming distractions.

The primary users were senior ( Subject Matter Experts ) SMEs and executives responsible for leading client conversations, identifying pain points, and proposing tailored solutions. These users typically operate in high-pressure environments such as live meetings and workshops, often while managing multiple tools, documents, and inputs simultaneously.

In this context, the platform was not a primary workspace but a supporting system used alongside presentations, notes, and real-time discussion. Speed, clarity, and trust were critical, while complex interactions or dense visual elements risked becoming distractions.

[ Research ]
The most challenging constraint on this project was access.
SMEs and senior leaders — The primary users were impossible to schedule. Calendars were blocked, meetings were back-to-back, and structured interview sessions simply weren't happening.
Rather than wait, I proposed a different approach to my manager: let me go to them.

This led to CONTEXTUAL INQUIRY
I spent a full day embedded with the SME team sitting in their offices, sharing working spaces, observing from the corner of conference rooms during live sessions. No script, no interview guide. Just observation.

What I saw changed the product direction entirely.
Every single SME was running two screens. A primary screen for the meeting, slides, video calls, notes. And a secondary screen, constantly, where they were searching for data, pulling up case studies, cross-referencing solutions all while trying to stay present in the conversation.

The secondary screen was their lifeline, and it was completely undesigned for that purpose.
[Research Insights ]

They didn't need a better dashboard.
They needed a glanceable secondary interface information available at a glance, requiring zero navigation while in flow. This observation shaped every major design decision that followed:

Secondary screen as the primary use context
Large, scannable typography over dense data
In ambiguous technical environments, design with assumptions, not after certainty.

Waiting for legal and tech teams to resolve LLM constraints would have stalled the project indefinitely. I designed with explicit, stated assumptions and validated iteratively instead.

The best enterprise AI interface stays out of the way.

The most impactful decision wasn't a feature we added it was what we removed. Restraint and trust signals outperformed every dashboard we explored.

Minimal interaction depth look, don't click
[Process Followed]
01
Think
Ideation Research Mental Modals Behavioural Modals Competitive Analysis
02
Make
03
Check
01
Think
Ideation Research Mental Modals Behavioural Modals Competitive Analysis
02
Make
03
Check

This project coincided with the earliest commercial wave of large language models, and the ambiguity was real. Parallel workstreams from legal, tech, and compliance teams were running simultaneously with unanswered questions:

How much internal client data could be shared with an LLM?
Could we access models via API or only via local servers?
What guardrails were needed to reduce hallucination risk?

A traditional double diamond process assumes you can define the problem, then solve it. Here, the problem kept shifting as technical constraints evolved. So I shifted too working with explicit assumptions, fast iterations, and progressive commitment rather than waiting for certainty that wasn't coming.

Trust requires transparency, users need to see where information comes from
AI value must be expressed in business outcomes, not technical specs
Conversational entry points reduce cognitive load
Users need clarity, not volume
[ Design Goals ]

The design focused on enabling experts to stay present in conversations while accessing relevant information seamlessly.

  • Reduce cognitive load during live usage

  • Surface insights and assets instantly

  • Support confident, real-time decision-making

  • Maintain a high level of clarity and trust

  • Fit naturally into existing consultant workflows

[ Approach ]

Given the executive nature of the users and limited availability for structured research, the design process prioritised contextual understanding and iterative validation. Inputs from stakeholders, usage patterns, and real working environments informed design decisions.
The approach emphasised practicality over novelty focusing on how the platform would be used in real moments rather than how it appeared in isolation.

[ Solution ]

The final solution was a streamlined, lightweight GenAI interface designed to function as a glanceable assistant rather than an immersive application. Key elements included:

  • A minimal, distraction-free layout

  • Large, readable typography for at-a-glance consumption

  • “Battlecard” style summaries with key talking points and assets

  • Simplified navigation and reduced interaction depth

  • Voice-first and hands-free interaction where appropriate

  • The design intentionally removed unnecessary visual complexity to prioritise speed and usability in live settings.

[Outcome]
200+ SMEs adopted the platform within the first 3 weeks of launch no mandated rollout, fully organic.
$2M+ in new client pitches enabled directly through Navigator-generated outputs.
Contributed to Accenture's broader $3B AI acceleration investment, including partnerships with Anthropic and OpenAI.
[Key Learnings]
Access constraints are a research brief, not a blocker.

When I couldn't interview users, I observed them instead contextual inquiry surfaced the dual-screen behaviour that no interview would have, because users weren't even conscious of it themselves.

In ambiguous technical environments, design with assumptions, not after certainty.

Waiting for legal and tech teams to resolve LLM constraints would have stalled the project indefinitely. I designed with explicit, stated assumptions and validated iteratively instead.

The best enterprise AI interface stays out of the way.

The most impactful decision wasn't a feature we added it was what we removed. Restraint and trust signals outperformed every dashboard we explored.

Access constraints are a research brief, not a blocker.

When I couldn't interview users, I observed them instead contextual inquiry surfaced the dual-screen behaviour that no interview would have, because users weren't even conscious of it themselves.

In ambiguous technical environments, design with assumptions, not after certainty.

Waiting for legal and tech teams to resolve LLM constraints would have stalled the project indefinitely. I designed with explicit, stated assumptions and validated iteratively instead.

The best enterprise AI interface stays out of the way.

The most impactful decision wasn't a feature we added it was what we removed. Restraint and trust signals outperformed every dashboard we explored.

[ Users & Context ]

The primary users were senior ( Subject Matter Experts ) SMEs and executives responsible for leading client conversations, identifying pain points, and proposing tailored solutions. These users typically operate in high-pressure environments such as live meetings and workshops, often while managing multiple tools, documents, and inputs simultaneously.

In this context, the platform was not a primary workspace but a supporting system used alongside presentations, notes, and real-time discussion. Speed, clarity, and trust were critical, while complex interactions or dense visual elements risked becoming distractions.

Back

Back

Select this text to see the highlight effect