Case Study
Designing an AI Business Co-Pilot at DreamHost
Transforming hosting from a technical control panel into a guided growth experience
Project Overview
Project
AI Business Co-Pilot Ecosystem (DreamHost)
Type
AI-Integrated Platform Transformation
Role
Co-Lead Product Designer
Timeline
2-Year Strategic Product Evolution
Scope
End-to-End Business Lifecycle System (Profiler → Planner → Liftoff → Advisor → Assistant)
Deliverables
Research · System Architecture · UX Strategy · Interaction Design · High-Fidelity UI · Prototypes
Post-launch impact
200K+
Customers reached
Reached in the first quarter after launch
90%
Bypassed curated prompts
Users preferred custom inputs over suggested ones
Limited repeat usage
Most users engaged only once
The tool served its purpose in a single visit
Fewer support tickets
Reduced setup confusion
Answers moved into the panel, reducing repeat questions
Overview
Buying hosting is often the first step in building something meaningful — a business, a personal brand, a new idea. But inside DreamHost's control panel, that excitement frequently gave way to uncertainty. Customers weren't just navigating WordPress setup. They were navigating unfamiliar decisions about strategy, content, audience, and launch readiness — all within a complex technical environment where mistakes feel high-risk.
As a UX/UI Designer at DreamHost, I co-led the experience design of a suite of AI-powered tools built to reduce that uncertainty. I led the design of Business Profiler, Liftoff, and DreamHost Assistant, and collaborated on AI Business Advisor. Business Planner was designed by my design partner and is included here to illustrate the full end-to-end ecosystem.
This work wasn't about adding AI features. It was about designing a scalable guidance ecosystem — one that builds trust, adapts to user context, and reinforces confidence at every stage of the business journey.
Research
Understanding Why Users Stalled After Purchase
To understand where friction emerged, we analyzed support ticket themes and recurring post-purchase questions, onboarding behavior and drop-off patterns, cross-functional insights from Support and Product teams, and qualitative observations of early user workflows. A clear pattern surfaced. After purchase, users weren't initially struggling with technical setup — they were struggling with strategic decisions.
Research Questions
The friction wasn't about features. It was about direction.
Insights & Problems
The Real Problem: Decision Paralysis After Purchase
Although DreamHost serves a wide range of customers, the friction consistently surfaced across two behavioral patterns.
Determined Daniel
I know this can work — I just need a clear game plan to make it real.
- High ambition, low technical confidence
- Fear of making irreversible mistakes
- Reassurance-driven behavior — Am I doing this right?
Starter Suzy
I'm excited about teaching — but I didn't expect setting up a website to feel this overwhelming.
- Comfortable with tools, but lacked structured strategic clarity
- Frequently rebuilt or revised due to misalignment
Shared Pattern
Both beginners and experienced users stall after setup — not because they lack tools, but because they lack clear direction on what to do next.
The problem isn't lack of tools — it's lack of direction.
This revealed an opportunity to design a more guided, decision-support experience instead of a purely technical control panel.
Journey & System Design
Journey & System Design
Designing AI as a Guided Lifecycle — Not a Feature
Concept & Reframe
Persona insights revealed a key gap: users weren't looking for answers — they were looking for direction.
A standalone chatbot wouldn't solve that. In a dense, high-stakes control panel, adding AI as another feature would increase complexity, not reduce it.
Guidance needed to be embedded into the experience — not layered on top.
This led to a fundamental reframe:
How might we design AI as a business co-pilot inside a high-stakes hosting environment — one that guides without overwhelming, adapts to experience levels, and builds trust over time?
Principles
Guides without overwhelming
Adapts to experience levels
Builds trust over time
The 5-stage lifecycle
System Model
A Connected System, Not Isolated Tools
Each module works on its own, but they share the same context — so users don't have to repeat setup at every step. Goals, business type, progress, and prior inputs carry forward across the system, making guidance more relevant over time.
Guided Lifecycle Flow
Guided Lifecycle Flow
Capture goals & business context
Turn goals into an actionable roadmap
Execute safely with progress tracking
Surface optimisation signals & next moves
Resolve friction across every stage
Shared Context Layer
Feedback loop: AI Business Advisor surfaces optimisation signals that feed back into planning — making recommendations smarter over time.
Connected Ecosystem
Tool Ecosystem Timeline
Five modules, powered by one shared context. Each step builds on the last — reducing repetition and improving guidance as users move through the journey.
Personalized foundation for every tool
Aligns all recommendations to your unique business goals.
Goals into clear, actionable next steps
Structured templates and dynamic analysis turn ideas into strategy.
Full website live in under a minute
AI content, performance-first design, zero coding needed.
SEO, research, and creative strategy in one place
Uncovers competitive insights and sparks product and brand ideas.
Always-on guidance at every touchpoint
Surfaces help exactly where users need it, without leaving their workflow.
Why this approach worked
"The goal was better decision-making — not automation. Guidance is embedded where users already work, reducing uncertainty without adding another interface to manage."
System map showing how guidance modules connect across the customer journey.
With the system model in place, the next step was translating it into screens and interactions.
Wireframes & UI Breakdown
Designing for Cognitive Safety at Every Phase
The visual system was built to reinforce calm, structured layouts with clear hierarchy, step-based progression, and transparent AI indicators. Every layout decision had to answer the same question: does this reduce uncertainty, or add to it? Wireframing this system wasn't primarily a visual exercise — it was a trust exercise.
WIREFRAMES
Phase 1 — Business Profiler: Cognitive Scaffolding
The Profiler's earliest wireframes used open-ended text fields — asking users to describe their business in their own words. We quickly discovered this replicated the very problem we were solving: a blank canvas that invited paralysis. We rearchitected the interaction model entirely around structured, sequential prompts with constrained inputs.
Step-based progression over a single long form.
Breaking input across stages reduced cognitive load and gave users momentum — each completed step signaled forward movement, not standing still.
Constrained choice over open text.
Dropdowns and multiple-choice reduced decision fatigue without removing personalization. Users felt guided, not boxed in.
Persistent progress indicators.
We tested without them — drop-off increased significantly. Progress visibility was a direct confidence signal.
Conversational microcopy.
Technical phrasing made even simple questions feel high-stakes. We iterated copy alongside layout until both felt like guidance, not interrogation.
Steps 0–3 · Welcome screen, website selection, business overview, and profile naming.
Steps 4–6 · AI Advisor chat, goals and priorities, and confirmation summary.
WIREFRAMES
Phase 2 — Business Planner: From Questioning to Guiding
(Business Planner was designed by my design partner and is included here to illustrate the full end-to-end ecosystem).
The Planner's challenge was different: we weren't collecting information — we were returning structured strategy based on it. The risk was overwhelming users with a wall of AI-generated output. We wireframed and tested three output models: a single consolidated plan, a card-based priority grid, and a sequenced task view with staged milestones. The sequenced view consistently performed best — not because it showed more, but because it gave users a clear sense of what came first.
The Planner marked the moment the AI shifted roles: from asking questions to making recommendations. Wireframing this transition — with clear visual hierarchy distinguishing AI-generated guidance from user-editable content — was essential to maintaining trust without creating dependency.
Fig. 2 · Business Planner wireframes showing AI-generated strategy output and user editing states
Wireframes
Phase 3 — Liftoff: Designing for Psychological Safety
Liftoff introduced a more nuanced design challenge: users were about to let AI generate a real WordPress website on their behalf. The goal wasn't just speed — it was confidence. Every interaction needed to reassure users that they were in control, even as automation increased.
Users were far more willing to proceed when the experience felt guided, reversible, and low-risk, rather than opaque or final.
Safe staging before publishing
We introduced a clear separation between creation and launch of the WordPress site. By framing the experience as a setup phase — similar to the guided WordPress installation flow — users could explore and refine without fear. Changes felt iterative, not permanent.
Guided progression over full exposure
Early wireframes surfaced too much of the WordPress environment too soon, which caused hesitation. We shifted to a step-by-step, form-driven flow — domain, plan, setup — revealing WordPress complexity progressively and only when relevant.
Reversible actions with visible control
Undo and edit states were made persistent and visible, not hidden. Users needed to see that they could go back, adjust inputs, or refine their WordPress site at any time.
AI as a collaborator, not an authority
AI-generated content for the WordPress website was reframed as a starting point, not a final result. By positioning outputs as editable suggestions, users moved from passive acceptance to active participation.
Fig. 3 · Liftoff wireframes illustrating staging flows, reversible edit states, and progressive complexity layering.
WIREFRAMES
Phase 4 — AI Business Advisor: Designing for Continuity
Most AI tools reset with every interaction, forcing users to repeat context. For the Business Advisor, we designed a system that maintains continuity across conversations, turning the experience from isolated chats into an ongoing workspace.
Persistent history panel
Users can revisit past conversations and continue where they left off, reinforcing progress over time.
Context-aware interface
Relevant business information and prior inputs remain visible, reducing the need to re-explain goals.
Guided + flexible inputs
A structured prompt library supports users who need direction, while the open chat allows more advanced exploration.
Clear conversational structure
User input and AI responses are visually distinct, making interactions easier to follow and edit.
This approach positioned the Advisor not as a simple chatbot, but as a continuous thinking partner embedded in the product.
Fig. 4 · AI Business Advisor wireframes showing context continuity, conversation threading, and prompt structure
WIREFRAMES
Phase 5 — DreamHost Assistant: Trust Through Transparency
Users interacting with the Assistant were often already experiencing friction or frustration. The design focus shifted from guidance to reassurance — ensuring every interaction reinforced trust through clarity, transparency, and immediate support.
Visible source attribution
AI responses include inline links to relevant Knowledge Base articles, shown by default. Transparency becomes a functional reassurance, allowing users to verify information in real time.
Context persistence across the session
Full conversation threading ensures users never have to repeat themselves. Follow-up questions build naturally on prior context, reducing friction and maintaining continuity.
One-click human escalation
Access to human support remains visible and immediate. Even when not used, the presence of escalation reduces anxiety and increases confidence in the system.
AI Assistant experience showing verified responses, persistent conversation context, and seamless escalation to human support.
High-Fidelity & System Integration
One system. Five tools. Zero repetition.
The final designs weren't built as individual screens — they were built as a continuous experience. Every transition, every AI output, and every piece of guidance shares the same context, so users always know where they are and what comes next.
Prototype in Action
See the system behave as one
A short walkthrough showing how AI outputs from one tool inform the next — demonstrating context continuity, not just screen transitions.
Interactive prototype demonstrating transitions between lifecycle phases and the persistence of user context across tools.
Design System Components
Built for AI-driven interactions
Each component was designed to handle uncertainty — surfacing AI suggestions without overwhelming users, and always keeping a human fallback within reach.
Custom components designed to support AI-driven interactions consistently across the ecosystem.
Final Outcome & Learnings
Key takeaway
Guidance is infrastructure, not a feature.
Impact comes from removing uncertainty at the right moment — not from adding more capabilities.
Before
- —Customers had to leave the flow to find answers
- —Guidance varied across steps and surfaces
- —Setup confusion often turned into support tickets
After
- ✓Guidance appeared in context, inside the panel
- ✓A shared system made help consistent end-to-end
- ✓Customers could self-serve more often during setup
Learnings
Guidance belongs in the flow.
Help works best at the moment of decision — not in a separate destination.
Consistency is a designed system.
Shared patterns made guidance feel reliable across every surface.
Design for the stuck moments.
The highest-impact changes came from where customers hesitated most.
What made it work
Guidance lived where the work happened.
In-panel help at key steps — no context switching required.
Reusable patterns scaled the system.
Shared components kept guidance consistent across the full journey.
Co-design produced a stronger outcome.
Shared ownership with my design partner built a more resilient ecosystem than either could alone.