Case Study
Exploring how generative AI could support an ERP system by helping business admins compose, refine and distribute communications more efficiently.
At a glance
The problem
Business admins used an ERP system to manage and distribute communications, but the process had opportunities for AI-supported content generation, refinement and delivery.
The work
I explored AI interaction patterns, mapped AI-supported user journeys, created wireframes and iterated high-fidelity Figma prototypes alongside user testing.
The outcome
The team delivered an end-to-end AI-enhanced prototype, stakeholder presentation and Figma handover, with plans to continue the work into future development.
Role
UX Designer
Project type
Discovery sprint
Duration
3 months, 2024
Focus
AI-assisted communications
Prototype concept
AI-supported communication generation flow
The prototype showed how users could work with an AI assistant to create, refine and preview campaign communications before distributing them through familiar tools such as email.
Context
This discovery sprint explored how generative AI could be introduced into an ERP system used across a range of business contexts. One of the most valuable opportunities identified was supporting business admins who manage communication creation and distribution.
The aim was to identify where AI could improve the user experience, reduce manual effort and improve the quality of output, then produce an end-to-end prototype that could be used to communicate the concept and inform future build work.
Problem
ERP systems can support broad and complex business processes, but that flexibility can also make everyday workflows feel heavy. In this case, users needed to create and distribute business communications, sometimes to large groups or conference audiences.
The design challenge was to find a focused use case for generative AI that felt genuinely valuable, rather than adding a generic chatbot or disconnected AI layer. The experience needed to support content generation and distribution while keeping the user in control.
Explored how other products facilitate AI-to-human interaction, including conversational interfaces, assistant-style patterns and approaches to building trust.
Worked through possible directions and narrowed the focus to content generation and communication distribution as the most valuable area to develop further.
Created user journeys showing how AI could support the comms generation and delivery process from initial input through to review and distribution.
Moved from early wireframes into high-fidelity Figma prototypes, iterating the AI interaction, tone, personality and communication style alongside user feedback.
Decision
Why
The discovery identified several possible AI opportunities, but comms creation had a clear user need and a strong fit for generative AI.
Impact
Created a focused concept that was easier to validate, prototype and explain to stakeholders.
Decision
Why
User feedback became more positive once the AI had a more personable presence, with users describing it as approachable, cute and helpful without feeling excessive.
Impact
Increased willingness to engage with the AI and helped make the interaction feel more trustworthy and human.
Decision
Why
Business communications still require judgement, tone control and review, so the AI needed to support the user rather than fully automate the workflow.
Impact
Positioned AI as a collaborative helper for drafting and refinement, reducing the risk of users feeling detached from the final output.
Decision
Why
Because the AI had been given an identity and personality, future work would need consistency across visual design, tone and interaction behaviour.
Impact
Produced a basic visual guide to support future design and development work beyond the discovery sprint.
AI assistant identity
Creating a personable assistant to make AI feel more approachable
The assistant identity helped make the AI feel less abstract and more approachable, supporting a more conversational experience while keeping the user in control.
Outcome
Learning
This was my first time designing for AI, and it highlighted how much complexity AI introduces beyond the interface itself. I had to consider how users might assume the AI works, how much explanation it should provide and how to handle moments where the AI may be incorrect.
It was also a valuable opportunity to take ownership of my design process, progress quickly through iterations and support future delivery by thinking about consistency, personality and implementation from the start.
