Start with intent
Let people explain the outcome before asking them to configure a workflow.
WORKLY AI AGENT / RAYAFrom a request
From a request
to a repeatable skill.
Exploring an AI teammate that turns everyday conversations into organised work—with people in control.
Follow the story
MEET RAYA Workly’s AI teammate concept
- My role
- UI/UX · Conversational interaction
- Product
- Workly · Web workspace
- Focus
- Custom skills through chat
- Stage
- Concept & low-fidelity design
01 / THE OPPORTUNITYLess setup.
Less setup.
More useful action.
Project work includes recurring updates, task assignments, and follow-ups. This exploration asks how a person could describe that work in plain language and turn it into a reusable AI skill.
THE REPEATING TASK“Send the team’s
“Send the team’s
task update. Again.”
Collect information → write the update → share it → repeat.
THE DESIGN OPPORTUNITY“Raya, make this
“Raya, make this
a weekly skill.”
Describe the goal → define the details → review the skill → run it.
Build shared context
Clarify the schedule, data, and action through a guided conversation.
Keep control visible
Make created skills easy to find, inspect, edit, and stop.
02 / RESEARCH TO DIRECTIONStudy the patterns.
Study the patterns.
Find a clearer path.
The project research compared Motion, Artisan, and Enso. I used the comparison to explore three questions: how people discover agents, how they create skills, and how much control they keep.
Role-based agents & skill setup
The study examined specialised agents and the sequence of naming a skill, defining its trigger, and adding inputs.
Design direction
Make skill creation easy to follow.
Focused business workflows
The comparison explored how an agent can bring several steps of a business task into one workflow.
Design direction
Connect the conversation to a clear outcome.
A broad choice of agents
The study raised a discoverability question: how do people find the right capability when many options are available?
Design direction
Keep useful skills close to the conversation.
Summarised from the project’s research notes; this describes the design exploration rather than a current product audit.
See the research comparison
Enlarge ↗03 / THE CONCEPTUnderstand. Remember.
Understand. Remember.
Then act.
The proposed system has three layers. Separating them explains how a conversational request could become a useful action inside Workly.
From words to understanding.
The conversation identifies the goal and clarifies missing details—what to do, who it is for, and when it should happen.
THE REQUEST“Send my team a weekly task summary.”
From information to useful context.
Project history, current tasks, and workspace information give the request meaning. The right data supports the next step.
THE CONTEXTTeam leads · Pending tasks · Deadlines
From a clear goal to repeatable work.
A defined trigger and action turn the request into a reusable skill—one that people can review, run, edit, or stop.
THE PROPOSED SKILLMonday, 9 AM → Email the summary
THE INTELLIGENCE BEHIND THE INTERACTIONInside RAYA.
Explore artwork ↗PAST
What happened?
Completed tasks, earlier projects, and reports.
PRESENT
What is happening?
Active projects, ongoing tasks, and current activity.
FUTURE
What is planned?
Schedules, upcoming work, and forecast inputs.
This is a proposed product model. The case study does not claim a working AI implementation.
04 / TWO EVERYDAY STORIESShow the conversation.
Show the conversation.
Make the behaviour clear.
Storyboards explore what RAYA needs to ask, what the person decides, and how the resulting skill is described. Choose a scenario to follow the interaction.
STORY 01 / CXOA weekly update,
A weekly update,
built through chat.
The leader asks RAYA to email team leads every Monday. The conversation turns that request into a named skill with a trigger, selected information, and an action.
- Name the skill
Give the recurring task a recognisable name: Weekly Task Summary.
- Set the trigger
Define when it runs: Monday at 9 AM.
- Choose the inputs
Select team lead, pending tasks, and deadlines.
- Define the action
Send an email, then review the skill summary before testing it.
Enlarge ↗SKILL SUMMARYWeekly Task Summary
- When
- Monday · 9 AM
- Uses
- Team lead · Tasks · Deadlines
- Action
- Email team leads
STORY 02 / MANAGERA daily report.
A daily report.
A deliberate audience.
A manager asks for a daily progress report on Project X. RAYA clarifies the time, report content, and who should be able to receive it.
- Describe the report
Ask for a daily progress update for Project X.
- Choose a time
Schedule delivery for 6 PM each day.
- Clarify the content
Discuss completion, pending items, and deadlines.
- Confirm the audience
The manager chooses a private report rather than sharing it with the team.
Enlarge ↗REPORT SUMMARYProject X · Daily progress
- When
- Every day · 6 PM
- Delivery
- Inbox & dashboard
- Audience
- Manager only
05 / ACCESS & CONTROLThe same assistant.
The same assistant.
Different permissions.
The concept maps access to a person’s role. A conversational interface still needs clear boundaries around what someone can view, assign, or manage.
01All actions
CXO / Admin
Super access
02Assign · View · Manage
Team lead / Manager
Management access
03View · Mark tasks
Employee / Intern
Individual access
Control continues after creation.
The wireframes include Run, Edit, and Stop actions. A skill remains something the person can inspect and change.
06 / LOW-FIDELITY EXPLORATIONFrom dialogue
From dialogue
to a usable workspace.
Four early sketches explore the complete skill lifecycle. The workspace navigation stays familiar, while the centre changes from an introduction to chat, confirmation, and skill details.
Enlarge ↗
Enlarge ↗
Enlarge ↗
Enlarge ↗07 / REFLECTIONA clearer route
A clearer route
from intent to action.
This exploration connects competitor research, a proposed system model, role-based access, two user stories, and early wireframes. The central design idea is to make automation understandable through conversation and visible controls.