The Growth Podcast

Episode: How to Design with AI | The Complete Guide for PMs with Xinran Ma

Host: Aakash Gupta
Guest: Xinran Ma
Date: February 21, 2026

Overview

This episode is a comprehensive masterclass on “designing with AI” specifically geared for product managers (PMs) and product designers who are navigating the rapidly evolving landscape of AI-powered tools. Host Aakash Gupta welcomes Xinran Ma, author of the “Design with AI” newsletter and a leading expert on AI design workflows, to unpack the true meaning of designing with AI—not just prompting AIs, but intentionally architecting workflows, systems, and user experiences. Together, they explore practical frameworks, tools, and mindsets for ideation, prototyping, and evaluating risks for AI-centric products.

Key Themes and Insights

1. The Universe of "Design with AI" (02:02)

Xinran’s Mind Map of Designing with AI:
- Prompting: Not just "prompt engineering," but about clarifying asks and providing the right context.
- Ideation: Leveraging AI assistants to generate ideas that humans may overlook.
- Design & Prototyping: Using tools for faster, more diverse prototyping iterations.
- Conscious Design: Building thoughtfully with awareness of risks like bias, hallucination, and lack of empathy in AI.
- Quote [02:02] (Xinran):
  “There are some workflows that I can explain more. Last but not least is a part of the landscape that fewer people talk about...designing with AI and staying conscious means how can we bring more intentions and thoughtfulness into the things that we are designing.”

2. Prompting with AI: Beyond Frameworks (03:16)

Clarifying the Ask:
- Start by clarifying what you want to get out of AI (the “blank canvas” problem).
- Use AI itself to help gain clarity if you’re uncertain.
The Power of Context:
- Context engineering > prompt engineering.
- Provide only necessary context: user roles, needs, constraints, audience, and references.
- Don’t overstuff prompts with irrelevant details.
- Quote [03:28] (Xinran):
  “Prompting with AI, it’s just a way of how human interact with AI in order to get better results. And I try to make things as simple as possible from a design perspective.”
On Structure:
- It’s “an art”—balance between specificity and room for exploration.
- Sometimes a structured prompt, sometimes open-ended for divergent thinking.
- Quote [07:35] (Xinran):
  “Sometimes we want more control of the outcome and sometimes we want to give AI more space to brainstorm. And it's really a fine balance.”

3. Ideation & Prototyping Workflows (08:32)

Ideation:
- Effective when contextualized with business goals, user needs, and clear constraints.
- Use AI for both divergent and convergent thinking (e.g., brainstorm and then rank ideas).
- Always ask for examples and sources to verify AI-generated ideas’ credibility.
Prototyping:
- Specify clear instructions and context.
- Track design variations, navigate between them, and refine efficiently.
- Prototyping and ideation are now merged due to AI tools’ capabilities.

4. Risk-Aware, Conscious Design (14:56)

Risks in AI Design:
- Hallucinated, biased, or generic outputs.
- Lack of true user empathy—AI can’t “feel” so always keep humans in the loop.
- Importance of double-checking sources, including diverse perspectives, providing user control, and careful audit of AI-generated outputs.
- Quote [14:56] (Xinran):
  “Empathy with the people you’re designing for becomes even more important because...AI is not human, right? It does not really have enough empathy.”

Workflow Deep Dives

1. Workflow 1: Going from “Blue Sky” Idea to AI Prototyping

(19:00 - 41:00)

Steps:

Custom GPTs for PRD Prototyping:
- Xinran demonstrates a custom GPT built for generating a lightweight PRD (product requirements doc) tailored for AI prototyping.
- The GPT asks sequenced questions to clarify the product goal, target users, platform, user flows, and avoids generic, unnecessary items (like login screens).
- Results in a focused, markdown-formatted spec, ready to drop into AI prototyping tools.
- Quote [20:36] (Xinran):
  “Effective prototypes is about request what you want to get out of AI, the context and reference…for the very first prompt and very first prompt matters.”
Mock Run in Claude:
- Paste spec into Claude AI for a first visual mock run.
- Quick test to ensure the spec/prompt generates as intended before using heavier tools.
Try with Other Tools (Lovable, V0, Bolt, Google AI Studio):
- Each tool has strengths:
  - Lovable: Most well-rounded, high-fidelity output, frequent updates, paid.
  - V0: Similar quality; allows code editing on free plan.
  - Google AI Studio: Fast evolution, free, catching up on visual design quality.
  - Subframe, Magic Patterns: Niche, product-design focused.
  - Bolt, Replit: More full-stack, developer-centric.
- Quote [36:13] (Xinran):
  "Lavabowl, V0, and Bolt are similar…Lovable [is] well-rounded…V0 more accessible as you can edit code without upgrading."
Best Practices:
- Only provide context relevant to the goal.
- Brief, hierarchical markdown for robust AI comprehension.
- Mock with quick tools, then iterate in specialized tools.

2. Workflow 2: From Existing Experience to Divergent Design with Google Stitch

(41:30 - 57:32)

Steps:

Google Stitch for Idea Divergence:
- Use a snapshot of an existing design (e.g., “Ask Redfin” feature) and context-rich prompts.
- Stitch is uniquely positioned for early-stage, divergent design ideation—quickly generates multiple, diverse design options.
- Special “YOLO mode” in Stitch delivers wild, maximally divergent variants.
- Quote [50:47] (Xinran):
  “You can make it very refined or YOLO, which is going crazy…for divergent brainstorming.”
Evaluate and Iterate:
- Select and iterate on promising variations.
- Stitch allows for regenerating and varying designs, creatively adjusting layout, colors, etc.
Export to Prototyping Tools:
- Export selected variants into Google AI Studio for prototyping and interactive flows.
Notes:
- Stitch is currently free, but its long-term support as a Google offering is uncertain.
- Advanced Tip:
  - In Google AI Studio, don’t miss the “System Instruction” field for extra styling/context details (56:08).
  - Use the new “Annotate App” feature for collaborative, multiplayer feedback.
Comparison Table (Cursor, etc.):
- Cursor: More powerful and flexible but steeper learning curve, less friendly for non-technical users, slower than browser-based tools.
- For most, browser-based tools (Stitch, V0, Lovable) offer speed and accessibility.

Notable Quotes & Memorable Moments

Xinran on Prompting Philosophy [03:28]:

“Prompting with AI, it’s just a way of how human interact with AI in order to get better results. And I try to make things as simple as possible from a design perspective.”
On Empathy and AI [14:56]:

“Empathy with the people you’re designing for becomes even more important because...AI is not here human, right? It does not really have enough empathy with the people who are YouTube.” — Xinran
Custom GPT Construction [20:36]:

"Effective prototypes is about request what you want to get out of AI, the context and reference…for the very first prompt and very first prompt matters." — Xinran
Power of YOLO Divergence [50:47]:

“You can make it very refined or YOLO, which is going crazy…for divergent brainstorming.” — Xinran
Aakash Summing Up PM+AI Superpowers [59:37]:

“You don’t want to use AI to just do a worse job of what you would have done before. You want it to enable you with new superpowers. Because the designers and PMs who have new superpowers from AI, they will replace those who are just using AI generically.”

Segment Timestamps

00:00-01:35 — Introduction to the episode and guest; AI products designed wrong; why “prompting” isn’t the whole story
02:02-07:35 — What is "Design with AI?" Xinran's mind map and the universe of design with AI; essentials of prompting
07:35-12:29 — Specifics: balancing prompt structure, ideation workflows, convergent/divergent thinking
14:00-14:56 — Figma Make adoption trends among designers globally
14:56-17:59 — Conscious design: risks, empathy, human control, and high-quality insight sourcing
19:00-41:23 — Workflow 1 demo: Custom GPT for AI prototyping, best practices, plugging into Claude, Lovable, V0, etc.
41:23-57:32 — Workflow 2 demo: Google Stitch and Google AI Studio for divergent design, exporting, and collaborative feedback; tool comparisons
57:32-61:01 — Recap, additional resources (Xinran’s custom GPT, newsletter), closing thoughts

Additional Advanced Tips (56:08, 57:10)

Use system instructions in Google AI Studio for style and structure guidance.
Take advantage of new annotation/commenting and multiplayer features for collaborative AI-driven design review.
For non-technical users, browser-based tools are more approachable and conducive to iterative workflows.

Summary Table: Top AI Design Tools Discussed

| Tool | Best For | Strength/Notes | |---------------------|------------------------------------------|--------------------------------------------------------| | Lovable | High-fidelity prototyping | Best all-around, visual editing, paid | | V0 | Editable prototyping | Free code edit, aesthetic preference | | Google AI Studio| Fast, evolving, prototyping | Free, system instructions, catching up in design | | Stitch | Early, divergent ideation | Free, YOLO divergence, export to AI Studio/Figma | | Claude AI | Quick runs/checks | Lightweight mock, not for visual fidelity | | Cursor | Power users, full-featured builds | Steep learning curve, less for non-tech PMs/designers | | Bolt/Replit | Full-stack prototypes | Backend integration, more engineering required | | Magic Patterns | Design-focused prototyping | Free with podcast/newsletter subscription |

Final Takeaway

Designing with AI isn’t just about using the newest prompt template or tool—it’s about intentional workflows, high-quality input/context, risk awareness, and a cycle of divergent/convergent thinking. PMs and designers who consciously master these AI superpowers will lead the next wave of exceptional product teams.

For Xinran’s custom GPT, frameworks, and mind map, see links in the show notes and Aakash’s newsletter.

[For the demo mind map, custom GPT links, and visual references, refer to the episode's show notes and newsletter.]

The Growth Podcast

Episode: How to Design with AI | The Complete Guide for PMs with Xinran Ma

Host: Aakash Gupta
Guest: Xinran Ma
Date: February 21, 2026

Overview

Key Themes and Insights

1. The Universe of "Design with AI" (02:02)

Xinran’s Mind Map of Designing with AI:
- Prompting: Not just "prompt engineering," but about clarifying asks and providing the right context.
- Ideation: Leveraging AI assistants to generate ideas that humans may overlook.
- Design & Prototyping: Using tools for faster, more diverse prototyping iterations.
- Conscious Design: Building thoughtfully with awareness of risks like bias, hallucination, and lack of empathy in AI.
- Quote [02:02] (Xinran):
  “There are some workflows that I can explain more. Last but not least is a part of the landscape that fewer people talk about...designing with AI and staying conscious means how can we bring more intentions and thoughtfulness into the things that we are designing.”

2. Prompting with AI: Beyond Frameworks (03:16)

Clarifying the Ask:
- Start by clarifying what you want to get out of AI (the “blank canvas” problem).
- Use AI itself to help gain clarity if you’re uncertain.
The Power of Context:
- Context engineering > prompt engineering.
- Provide only necessary context: user roles, needs, constraints, audience, and references.
- Don’t overstuff prompts with irrelevant details.
- Quote [03:28] (Xinran):
  “Prompting with AI, it’s just a way of how human interact with AI in order to get better results. And I try to make things as simple as possible from a design perspective.”
On Structure:
- It’s “an art”—balance between specificity and room for exploration.
- Sometimes a structured prompt, sometimes open-ended for divergent thinking.
- Quote [07:35] (Xinran):
  “Sometimes we want more control of the outcome and sometimes we want to give AI more space to brainstorm. And it's really a fine balance.”

3. Ideation & Prototyping Workflows (08:32)

Ideation:
- Effective when contextualized with business goals, user needs, and clear constraints.
- Use AI for both divergent and convergent thinking (e.g., brainstorm and then rank ideas).
- Always ask for examples and sources to verify AI-generated ideas’ credibility.
Prototyping:
- Specify clear instructions and context.
- Track design variations, navigate between them, and refine efficiently.
- Prototyping and ideation are now merged due to AI tools’ capabilities.

4. Risk-Aware, Conscious Design (14:56)

Risks in AI Design:
- Hallucinated, biased, or generic outputs.
- Lack of true user empathy—AI can’t “feel” so always keep humans in the loop.
- Importance of double-checking sources, including diverse perspectives, providing user control, and careful audit of AI-generated outputs.
- Quote [14:56] (Xinran):
  “Empathy with the people you’re designing for becomes even more important because...AI is not human, right? It does not really have enough empathy.”

Workflow Deep Dives

1. Workflow 1: Going from “Blue Sky” Idea to AI Prototyping

(19:00 - 41:00)

Steps:

Custom GPTs for PRD Prototyping:
- Xinran demonstrates a custom GPT built for generating a lightweight PRD (product requirements doc) tailored for AI prototyping.
- The GPT asks sequenced questions to clarify the product goal, target users, platform, user flows, and avoids generic, unnecessary items (like login screens).
- Results in a focused, markdown-formatted spec, ready to drop into AI prototyping tools.
- Quote [20:36] (Xinran):
  “Effective prototypes is about request what you want to get out of AI, the context and reference…for the very first prompt and very first prompt matters.”
Mock Run in Claude:
- Paste spec into Claude AI for a first visual mock run.
- Quick test to ensure the spec/prompt generates as intended before using heavier tools.
Try with Other Tools (Lovable, V0, Bolt, Google AI Studio):
- Each tool has strengths:
  - Lovable: Most well-rounded, high-fidelity output, frequent updates, paid.
  - V0: Similar quality; allows code editing on free plan.
  - Google AI Studio: Fast evolution, free, catching up on visual design quality.
  - Subframe, Magic Patterns: Niche, product-design focused.
  - Bolt, Replit: More full-stack, developer-centric.
- Quote [36:13] (Xinran):
  "Lavabowl, V0, and Bolt are similar…Lovable [is] well-rounded…V0 more accessible as you can edit code without upgrading."
Best Practices:
- Only provide context relevant to the goal.
- Brief, hierarchical markdown for robust AI comprehension.
- Mock with quick tools, then iterate in specialized tools.

2. Workflow 2: From Existing Experience to Divergent Design with Google Stitch

(41:30 - 57:32)

Steps:

Google Stitch for Idea Divergence:
- Use a snapshot of an existing design (e.g., “Ask Redfin” feature) and context-rich prompts.
- Stitch is uniquely positioned for early-stage, divergent design ideation—quickly generates multiple, diverse design options.
- Special “YOLO mode” in Stitch delivers wild, maximally divergent variants.
- Quote [50:47] (Xinran):
  “You can make it very refined or YOLO, which is going crazy…for divergent brainstorming.”
Evaluate and Iterate:
- Select and iterate on promising variations.
- Stitch allows for regenerating and varying designs, creatively adjusting layout, colors, etc.
Export to Prototyping Tools:
- Export selected variants into Google AI Studio for prototyping and interactive flows.
Notes:
- Stitch is currently free, but its long-term support as a Google offering is uncertain.
- Advanced Tip:
  - In Google AI Studio, don’t miss the “System Instruction” field for extra styling/context details (56:08).
  - Use the new “Annotate App” feature for collaborative, multiplayer feedback.
Comparison Table (Cursor, etc.):
- Cursor: More powerful and flexible but steeper learning curve, less friendly for non-technical users, slower than browser-based tools.
- For most, browser-based tools (Stitch, V0, Lovable) offer speed and accessibility.

Notable Quotes & Memorable Moments

Xinran on Prompting Philosophy [03:28]:

“Prompting with AI, it’s just a way of how human interact with AI in order to get better results. And I try to make things as simple as possible from a design perspective.”
On Empathy and AI [14:56]:

“Empathy with the people you’re designing for becomes even more important because...AI is not here human, right? It does not really have enough empathy with the people who are YouTube.” — Xinran
Custom GPT Construction [20:36]:

"Effective prototypes is about request what you want to get out of AI, the context and reference…for the very first prompt and very first prompt matters." — Xinran
Power of YOLO Divergence [50:47]:

“You can make it very refined or YOLO, which is going crazy…for divergent brainstorming.” — Xinran
Aakash Summing Up PM+AI Superpowers [59:37]:

“You don’t want to use AI to just do a worse job of what you would have done before. You want it to enable you with new superpowers. Because the designers and PMs who have new superpowers from AI, they will replace those who are just using AI generically.”

Segment Timestamps

00:00-01:35 — Introduction to the episode and guest; AI products designed wrong; why “prompting” isn’t the whole story
02:02-07:35 — What is "Design with AI?" Xinran's mind map and the universe of design with AI; essentials of prompting
07:35-12:29 — Specifics: balancing prompt structure, ideation workflows, convergent/divergent thinking
14:00-14:56 — Figma Make adoption trends among designers globally
14:56-17:59 — Conscious design: risks, empathy, human control, and high-quality insight sourcing
19:00-41:23 — Workflow 1 demo: Custom GPT for AI prototyping, best practices, plugging into Claude, Lovable, V0, etc.
41:23-57:32 — Workflow 2 demo: Google Stitch and Google AI Studio for divergent design, exporting, and collaborative feedback; tool comparisons
57:32-61:01 — Recap, additional resources (Xinran’s custom GPT, newsletter), closing thoughts

Additional Advanced Tips (56:08, 57:10)

Use system instructions in Google AI Studio for style and structure guidance.
Take advantage of new annotation/commenting and multiplayer features for collaborative AI-driven design review.
For non-technical users, browser-based tools are more approachable and conducive to iterative workflows.

Summary Table: Top AI Design Tools Discussed

Final Takeaway

For Xinran’s custom GPT, frameworks, and mind map, see links in the show notes and Aakash’s newsletter.

[For the demo mind map, custom GPT links, and visual references, refer to the episode's show notes and newsletter.]