Product and design work now happens inside fast loops.
AI workflows in Glare define how work moves from generation to decision. AI makes it easy to create output. Prompts can generate ideas, flows, and variations quickly. Teams can explore more directions in less time than before. What becomes harder is turning that output into something the team can evaluate, align on, and move forward with confidence.
Glare provides structure for that gap.
It organizes AI-driven work into a system that connects generation, evaluation, and decision-making so teams can move quickly without losing clarity.
Where Work Breaks
As more work is generated, evaluation becomes harder. Teams explore more directions, but comparisons are inconsistent. Feedback accumulates, but it doesn’t always lead to a clear decision. Work continues, but alignment becomes harder to hold.
As Ian Batterbee put it:
AI has made it easy to generate work quickly, but much harder to know what to trust.
The gap shows up between creation and decision. Teams can generate faster than they can decide.
The AI Workflow
AI workflows in Glare define how work moves from generation to decision.
AI makes it easy to create output. Prompts can generate ideas, flows, and variations quickly. Teams can explore more directions in less time than before. What becomes harder is turning that output into something the team can evaluate, align on, and move forward with confidence.
Glare organizes AI workflows around a clear system:
-
Prompts generate possibilities
-
Skills shape how those possibilities are evaluated
-
Agents coordinate the process
-
Participant input creates design signals
-
Teams make decisions with signals
Each part has a defined role. Together, they create a repeatable way to move from idea to decision without losing structure.
Prompts Generate
Prompts are a simple entry point for beginners.
They are simple ways to use AI with very little setup. A prompt can turn a rough idea into a concept, generate variations, or reframe a problem in seconds. They work with tools like ChatGPT or Gemini and expand the space of possible directions.
Prompts are fast, but they are temporary. Each prompt is a one-time instruction.
They are best used for:
-
Exploring ideas
-
Generating variations
-
Reframing questions
-
Surfacing assumptions
They create momentum, but not consistency.
Skills Structure
Skills bring structure to that output. A Skill is a reusable, versioned way of working. It defines how a task should be performed so it can be repeated across sessions, teams, and workflows.
The distinction is simple:
-
A prompt defines what to do once
-
A Skill defines how to do it consistently
Skills shape the work by:
-
Framing questions as decisions
-
Preparing concepts for testing
-
Defining what should be measured
-
Structuring how options are compared
-
Guiding how results are interpreted
They turn exploration into something the team can rely on.
Agents Orchestrate
Agents coordinate how the work moves. In Glare, Ray acts as the orchestrator. It manages the overall goal, determines what needs to happen next, and applies the right Skills at the right time.
The relationship is clear:
-
Agents manage the goal and sequence of work
-
Skills provide the repeatable steps to execute it
For example, an agent working toward evaluating a product direction might:
-
Apply a Skill to structure a test
-
Use another Skill to analyze responses
-
Compare outcomes across options
This allows the system to move from intent to execution without losing structure.
Human in the Loop
Signals and decisions are grounded in people.
Design signals in Glare are created from participant input using UX metrics. They reflect how real users respond to concepts, flows, and experiences. This input gives the system something real to evaluate and compare.
Decisions are made by teams. Product, design, and engineering interpret those signals, weigh tradeoffs, and decide what moves forward. In the future, agents will progressively used to make decisions.
Decisions define:
-
What moves forward
-
What gets refined
-
What gets stopped
Signals make those tradeoffs visible so teams can commit with confidence.
The flow is direct:
-
Participants provide input
-
Signals make that input visible
-
Teams interpret what it means
-
Decisions determine what happens next
AI supports the work. People determine direction.
How It Comes Together
When these parts connect, the workflow becomes repeatable.
Work moves through a clear structure that turns exploration into action.
-
A prompt generates an idea or direction
-
A Skill shapes that idea into something testable
-
An agent coordinates the steps needed to evaluate it
-
Participant input creates a signal
-
The team uses that signal to decide what moves forward
Each step builds on the last. The output of one step becomes the input to the next.
This loop can run within a single session or extend across a larger cycle. As it repeats, decisions become easier to make and easier to carry forward.
What This Produces
This workflow produces outputs that are immediately usable:
-
Signals tied to real user behavior
-
Consistent comparisons between options
-
Decisions teams can align around
-
Clear next steps that can be executed
These outputs do more than summarize the work.
They:
-
Reduce uncertainty
-
Make tradeoffs visible
-
Give the team a shared reference point
-
Keep work moving with direction
<table xmlns="http://www.w3.org/1999/xhtml" style="min-width: 50px;"><colgroup><col style="min-width: 25px;"><col style="min-width: 25px;"></colgroup><tbody><tr><td colspan="1" rowspan="1"><h3><strong>Where This Works Best</strong></h3><p>AI workflows are most effective in situations where the work needs both speed and structure.</p><p>They are especially useful when:</p><ul><li><p>Multiple directions need to be explored and compared</p></li><li><p>Decisions need to be made quickly without losing clarity</p></li><li><p>User input can be gathered to inform direction</p></li><li><p>Teams need a shared way to evaluate options</p></li></ul><p>In these conditions, the workflow helps teams move forward without relying on opinion or fragmented feedback.</p></td><td colspan="1" rowspan="1"><h3><strong>Where It Needs Support</strong></h3><p>The workflow depends on a few conditions to hold.</p><p>It works best when there is:</p><ul><li><p>Access to participants or realistic user feedback</p></li><li><p>Clear metrics or criteria for evaluation</p></li><li><p>Consistent use of Skills across the team</p></li></ul><p>When these are in place, signals carry through the system and decisions remain stable.</p><p>When they are missing, the system still produces output, but the connection to decisions weakens. Comparisons become less reliable, and alignment becomes harder to maintain.</p></td></tr></tbody></table>
Try This Now
Start with a single decision.
-
Create a signal from collecting user input
-
Choose a prompt to generate new insights
-
Apply a Skill to structure how you evaluate them
-
Collect a small set of user responses
-
Decide what moves forward
You don’t need the full system to start. One signal is enough.
