Building websites and managing teams with a friendly AI agent

I designed an AI-driven workspace that helps teams plan, build, and ship websites from one place. Think Framer-style page creation paired with Linear-style task clarity, guided by a cheerful agent that turns messy ideas into launchable work.

A playful AI workspace for websites, tasks, and launches

I designed an AI-driven workspace that helps teams plan, build, and ship websites from one place. Think Framer-style page creation paired with Linear-style task clarity, guided by a cheerful agent that turns messy ideas into launchable work.

Role

Product Design

AI Experience Design

Interaction Design

Workflow Testing

Product Strategy

Tools

Figma

Framer

Claude

Linear

Timeline

8 weeks

The Problem

Teams often jump between design files, task boards, docs, and chat just to get a website or project from idea to launch.

The Solution

Create a friendly AI agent that can turn prompts into page drafts, break the work into trackable tasks, suggest next steps, and keep everyone aligned from first concept to shipping.

AI Site Building Flow

The product combines a visual website builder with an agent-led creation flow. Instead of starting from a blank canvas, teams can describe a product, campaign, or landing page, then get an editable structure with sections, copy, tasks, and launch recommendations already organized.

Team frustrations

Starting a new client website is exciting, but it quickly becomes scattered. Content lives in docs, feedback lives in chat, and action items live somewhere else entirely.

Primary Frustration

Blank-canvas pressure slows teams down. People know what they want to launch, but translating fuzzy ideas into sections, priorities, and next steps takes too much energy.

Secondary Frustration

Design and delivery work drift apart. A beautiful page can still stall when copy edits, bugs, approvals, and publishing tasks are not connected to the canvas.

How might we help teams go from idea to published website without losing the thread?

What can we do?

We can introduce a friendly agent that drafts pages, explains its choices, and turns each recommendation into editable work instead of a mysterious black box.

What will improve?

Teams can move faster because the agent handles the first pass, keeps tasks attached to the work, and makes the next best action feel obvious.

Laying the Groundwork

Using our previous in-house web design work as the foundation, the agent was designed to help with page outlines, starter copy, design direction, launch checklists, and tiny bits of encouragement along the way. The goal was to make AI feel like a helpful teammate, not a command line.

Prompt-to-Page Drafts

A short brief can generate a structured page with sections, suggested messaging, and editable blocks. The user stays in control, while the agent handles the awkward first draft.

Agent Suggestions

The agent surfaces small, friendly recommendations such as “add proof here,” “tighten this headline,” or “create a follow-up task.” Each suggestion is contextual, actionable, and directly tied to the rest of the team management system.

Designing With the Agent

The agent appears inside the workflow as a cheerful guide. It can draft sections, create tasks, summarize feedback, and help the team decide what to do next without pulling them away from the page.

Fast Starts, Clear Choices

Instead of overwhelming users with a giant prompt box, the flow offers friendly choices: start from a goal, improve a section, fix a bug, or plan a launch. Each path keeps the experience focused.

Seamlessly Integrated with Your Project

The agent’s personality is upbeat and practical. It celebrates progress, but it also names blockers, asks useful questions, and keeps the team moving toward shipping.

Result

The prototype made the product feel less like two separate tools and more like one helpful workspace: a place where teams can design the site, track the work, and ask the agent what should happen next.

Task Management for Startups and Small Teams

Website projects create a messy blend of design tasks, content edits, engineering bugs, approvals, QA notes, and launch reminders. To keep that work visible, I designed a task system that lives beside the page builder instead of hiding in another tab.

Team frustrations

When launch work is disconnected from the website itself, teams lose context. A task like “fix hero” is only useful if everyone knows which hero, what changed, and who needs to review it.

What can we add?

Add a task manager that can link issues directly to pages, sections, comments, and agent suggestions.

How can we improve it?

Let the agent triage feedback, draft tasks, suggest owners, and group work into clear launch milestones.

Connected Planning and Shipping

Borrowing the clarity of modern issue trackers, the task manager keeps status, priority, ownership, and context close to the creative work. The agent helps tidy the queue so the team can focus on shipping.

Page-Linked Tasks

Tasks can be created from a section, comment, or agent recommendation. Each item remembers where it came from, so designers, writers, and developers can jump back into the exact page context.

Agent Triage and Safeguards

The agent can summarize duplicate feedback, flag missing details, and suggest priority without silently changing the plan. People approve the work; the agent keeps it organized.

Design System Improvements

I created a scalable interface system for prompts, agent messages, task cards, status chips, page previews, and launch states. The system needed to feel productive and calm, with enough charm to make planning feel less like homework.

Three key learnings

1. AI feels more useful when suggestions are tied to visible work and clear next steps

2. Website building and task planning work better when they share the same context

3. A friendly agent can reduce friction, but users still need control, transparency, and easy undo

Next steps

The next phase would expand the prototype with multi-page generation, richer task automation, and deeper launch readiness checks. I’d also explore how the agent learns a team’s voice, remembers project decisions, and helps turn scattered feedback into a clean path toward publishing.