Ariad Assistant
Audit the different AI modes in the to build trust with hesistant users
Client.
BlueSentry
Tools.
Figma
Timeline
2 Sprint
Role.
UX/UI Designer
Problem
Convincing developers to use AI for client work. Learning how to create an agentic tool for a set of users who are accustomed to deterministic results, hestitant to use AI, and need to ramp up trust to achieve user adoption. Although my overall goal was to increase project managers and engineer productivity by introducing agents as a "digital workforce" to take on admin and technical task, I focused primarily on utilizing the existing AI chat and optimize its rule in building trust and setting up the user for trust.
Process
I spent one sprint on conducted user interviews and did usability testing with the current AI chat. I learned that usrs were confused with the dual chat interface and first-time users had trouble with the agent configuration. I created an audit presentation. Then for the second sprint, I prototyped and got buy-in for the design changes. The following are the key concepts I narrowed in on:
Separate the identify of the two chats.
Brand identity of ariad assistant as the ariad logo, separate from the agent that the user created. Global Ariad Assistant to play more collaborative role in helping the user configure their agents. Chat selected in the dropdown will be the saved state
Global Ariad Assistant to play a more collabrative role
-So the next suggestion is to find opportunities for the AI to adopt more roles to be helpful to the user. One of these modes is called creator where it's not invoked by the use and it's embedded AI content that the user encounters. So when the user adds a description of the agent, team of agent, or workflow in the starting modal, the first thing they should see in the chat is a genarative text that they can copy and paste into the system prompt and instructions. The user doesn't have to use it but this way they don't have to go to chat gpt or gemni to create these config. This replaces the previous flow where after the user creates a new agent, there is leftover context from another discussion.
The next mode of opportunity is a collaborative mode which is when the ariad assitant and the user work together toward a goal. We want to treat the global ariad assitant more as a partner that can help users configure their automations properly and to help troubleshoot. It has this functionality already but it's not utilized like how one of the interviewers mentioned and that's because we just need to tweak the setup of the ariad assistant to show the system prompt and instructions unprompted, to always have follow-up messages, and engage in a conversation more with the user. Having them available in this 3 panel approach instead of the modal overlay is best so that the user can easily fix their configuration. In the slide above, I have two pathways where a user on the top right manually copies the system prompt. Then in the image below that, the user asks ariad to add the prompt for them..
I assessed th
My Process
I assessed th
Outcome
-Configure this agent to have more confidence signaling language, e.g. say I don’t know if they don’t know
-Configure Ariad Assistant to not create fake data
-Add a disclaimer on the bottom
