How to Turn AI-Generated Designs into Live, Editable Websites

Practical AI: Episode 27

From AI Prototype to Live Website: OpenClaw, Agents & the Software Selloff

Watch on YouTube

Published: February 6, 2026

TL;DR

This episode features guest Shanee Moret demonstrating how she uses OpenClaw, an open-source AI agent, to automate sales emails, create video clips, and manage tasks autonomously. The hosts discuss how Claude Co-work’s vertical plugins triggered a massive software stock selloff, then demonstrate converting an AI-generated website design into a live, editable site using PageMotor in one click.

Table of Contents

About This Show

Practical AI is a weekly live show (Fridays 11am CT) hosted by Olga Pechnenko and Chris Pearson that cuts through AI hype to deliver news, trends, and hands-on tips for builders and founders. Unlike technical AI podcasts, Practical AI focuses on business applications and ROI—what actually works, what’s hype, and what you can implement Monday morning.

What You’ll Gain

  • Understand how OpenClaw works through a real user’s experience. Shanee Moret walks through her setup process, integration with Notion and Gmail, and how her AI agent “Ash” now handles sales follow-ups, video clip creation, and task automation autonomously.
  • Learn why software stocks crashed after Claude Co-work launched vertical plugins. The episode breaks down how AI shifting from “feature” to “replacement” triggered a trillion-dollar repricing of the entire SaaS sector.
  • Discover the prototype-to-production gap that most AI demos ignore. Chris demonstrates taking a Claude-generated website design and converting it into an actual live, editable website with user management and content controls.
  • See AI cancer detection results from the largest randomized trial ever conducted. A Swedish study of 100,000+ women found AI-supported mammography caught 27% more aggressive cancers while cutting radiologist workload by 44%.
  • Gain insight into AI funding patterns with nearly $20 billion raised in a single week, including Waymo’s $16 billion round and Cerebras’s $1 billion for AI chips competing with Nvidia.

Biggest Takeaway to Implement: If you have repeatable business processes—especially sales follow-ups, content repurposing, or task management—start experimenting with agentic AI tools like OpenClaw. The setup takes 30-40 minutes, and as Shanee demonstrates, the ROI can be immediate: autonomous sales that close while you sleep.

Free PageMotor and Practical AI Updates:

Frequently Asked Questions

What is OpenClaw and how does it work?

OpenClaw is an open-source AI agent that runs locally on your computer and connects to messaging apps like Telegram, WhatsApp, and iMessage. Unlike chatbots confined to one environment, OpenClaw can access your files, send emails, browse the web, and execute tasks autonomously—functioning like a digital employee you communicate with via text. Read more below.

Why did software stocks crash after Claude Co-work launched?

Claude Co-work’s vertical plugins for HR, sales, and marketing triggered a market repricing because investors realized AI is no longer just a feature within software—it’s becoming the replacement for software. Companies like HubSpot and Salesforce saw massive selloffs as the market questioned whether traditional SaaS has a future. Read more below.

How long does it take to set up OpenClaw?

Shanee Moret, a self-described non-technical user, completed her setup in 35-40 minutes. The process involves running a one-line install command, connecting to Telegram, and adding API keys for your preferred AI model (Claude, GPT, or Gemini). Subsequent setups would take approximately 20 minutes. Read more below.

What can OpenClaw actually do for a business?

In the episode, Shanee demonstrates OpenClaw sending personalized sales follow-up emails that generated multiple “yes” responses within hours, creating video clips from YouTube videos in under two minutes, posting content to Notion, and even sending text reminders to family members. The agent learns your style and stops asking for approval once it masters a task. Read more below.

How did AI improve breast cancer detection in the Swedish study?

The MASAI trial of 100,000+ women found AI-supported mammography detected 27% more aggressive cancers, reduced interval cancer diagnoses by 12%, and cut radiologist workload by 44%—all without increasing false positives. Read more below.

What is the difference between an AI-generated website and a live website?

AI tools like Claude can generate beautiful HTML/CSS designs (prototypes), but these are static files—you cannot edit content, add users, or manage pages without technical skills. PageMotor converts these designs into “living objects” with a full content management system, user authentication, and drag-and-drop editing. Read more below.

How much did AI companies raise this week?

The week saw nearly $20 billion in AI funding across 66 companies. Major rounds included Waymo’s $16 billion at a $126 billion valuation, Cerebras’s $1 billion for AI chips, and ElevenLabs’ $500 million for voice AI. Read more below.

Practical AI: From AI Prototype to Live Website

Key Definitions

What is an AI Agent?

An AI agent is software that can take autonomous actions on your behalf—not just answer questions, but actually execute tasks like sending emails, managing files, browsing websites, and coordinating with other tools. Unlike chatbots that require constant back-and-forth, agents can complete multi-step workflows independently once given instructions.

What is OpenClaw?

OpenClaw (formerly Clawdbot and Moltbot) is an open-source AI agent created by Peter Steinberger that runs locally on your computer. It connects to messaging apps like Telegram, WhatsApp, and iMessage, allowing you to delegate tasks via text. The agent can access files, send emails, browse the web, and execute shell commands—essentially acting as a digital employee.

What is the SaaS to “Service as Software” shift?

Traditional SaaS (Software as a Service) sells access to tools that humans operate. The emerging model—sometimes called “Service as Software” or “SAS with one A"—uses AI agents to deliver outcomes directly, potentially eliminating the need for many specialized software subscriptions. This shift is driving the current repricing of software stocks.

What is a vertical plugin?

A vertical plugin is a specialized module that gives AI agents domain-specific capabilities for industries like HR, sales, legal, or marketing. Claude Co-work’s vertical plugins allow the AI to handle industry-specific workflows autonomously, which is why investors see them as potential replacements for specialized SaaS tools rather than features within them.

Quotable Moments

“If it is a repeatable process that you are engaging in, it should be a machine.” — Chris Pearson on identifying automation opportunities

“Within a couple hours, we got multiple yeses. And I was like, this is crazy.” — Shanee Moret on OpenClaw handling sales follow-ups autonomously

“The problem with working with agents will definitely not be a lack of ideas… it’s almost just prioritization because they’ll always suggest something.” — Shanee Moret on managing agentic AI

“By the end of this year, I will be the world’s foremost expert on technical documentation.” — Chris Pearson on how AI has renewed his focus on documentation as the key to training AI employees

0:57 Weekly AI News and Trends

The Great Repricing

Claude Co-work triggered a $285 billion software selloff as investors reconsidered the entire SaaS market. Microsoft lost $360 billion in market cap in a single day despite reporting their best quarter ever. The market is pricing in a fundamental shift: AI as replacement, not feature.

The week’s news painted a picture of accelerating AI adoption across industries. Amazon announced they’re shipping free Alexa Plus to 200 million Prime members while simultaneously cutting 30,000 corporate jobs. Dow Chemical, a 127-year-old manufacturer, cut 4,500 positions specifically citing AI—signaling that AI labor displacement has moved well beyond Silicon Valley.

In infrastructure news, Nvidia’s $100 billion OpenAI investment plan has fizzled, with the CEO taking a cautious step-back approach. Meanwhile, xAI integrated with SpaceX to form a $1.2 trillion conglomerate—signaling the market’s preference for vertical integration. OpenAI launched Frontier, a management layer for onboarding and monitoring AI agent employees, though it’s less comprehensive than Claude Co-work’s approach.

Model Releases

GPT 5.3 and Claude Opus 4.6 launched on the same day. Claude 4.6 features a 1-million-token context window and enhanced agentic capabilities. Anthropic’s head of enterprise product described the shift as moving from “vibe coding” to “vibe working.”

AI Healthcare Breakthrough

The MASAI trial in Sweden—the largest randomized controlled trial of AI in cancer screening—found AI-supported mammography detected 27% more aggressive cancers while reducing radiologist workload by 44%. This represents AI delivering measurable life-saving outcomes at scale.

28:00 OpenClaw Deep Dive with Shanee Moret

Shanee Moret, who has nearly a million LinkedIn followers and 250,000 newsletter subscribers, joined as the show’s first guest to share her hands-on experience with OpenClaw. Her perspective as a non-technical power user provided a practical window into what’s possible with agentic AI today.

Setup Reality Check

Shanee completed her OpenClaw setup in 35-40 minutes as a non-coder. The process: run a one-line install command, configure Telegram integration, add API keys for your preferred LLM. She recommends the Claude Max subscription ($100-200/month) for best results—"you’re getting an employee,” she notes.

The key insight from Shanee’s experience is that OpenClaw operates fundamentally differently from chatbots. When she asked it to create 25 Notion posts, it requested only her Notion API key, then completed the task in three minutes. Traditional AI assistants would have taken 40+ minutes with constant permission requests.

Sales Automation Results

Shanee tested OpenClaw on a list of leads who had previously shown interest but hadn’t purchased. Within a few hours, the agent generated personalized follow-up emails that resulted in multiple sales. The first draft was “97% sounding like me,” she reported, with the second iteration indistinguishable from her own writing.

Perhaps most striking was the video clip demonstration. At 12:39 AM, Shanee asked if OpenClaw could cut video clips. By 12:41 AM—two minutes later—she received finished clips from her podcast, delivered directly to her Telegram with a message: “Here’s the clips. They’re on your desktop.”

Shanee’s approach to agent management reveals an important principle: the challenge with agentic AI isn’t generating ideas—the agent will always suggest improvements—it’s prioritization. She recommends staying focused on mastering specific use cases before expanding, and being explicit about approval requirements if you work in regulated industries.

49:27 Cupcake Website Design Demo

The second half of the episode tackled a problem everyone with AI-generated designs faces: the gap between a beautiful prototype and a functional website you can actually use. Chris demonstrated the full workflow using a fictional cupcake business.

Prompt to Production

The initial prompt: “My mom has a cupcake business and needs a cute and fun website.” After two iterations—adding a more playful font and hero section improvements—Claude generated a complete responsive design with animated buttons, hover effects, and a “cupcake matchmaker quiz” feature in under 10 minutes.

The critical distinction Chris emphasized: what Claude generates is HTML/CSS/JavaScript—essentially a picture of a website. It’s static, meaning any change requires regenerating the entire design. PageMotor converts this into “living objects” that can be manipulated individually through a visual interface.

The demonstration showed changing a phone number from a simple admin panel—a task that would require a full redeployment with static HTML. This matters enormously for agencies delivering client work and for anyone who needs to maintain their site without technical skills.

The Documentation Insight

Chris’s key revelation: AI has transformed documentation from a human training tool into the foundation for training AI employees. Well-documented processes can be executed by machines “millions, billions, possibly trillions of times with unreal accuracy.” This renewed focus on documentation is becoming his competitive advantage.

1:23:29 AI Funding News and Trends

The funding segment revealed nearly $20 billion flowing into AI companies in a single week—the second-largest week in the past ten. AI continues to dominate overall venture funding, with the United States accounting for the vast majority of investment.

Week’s Top Rounds

Waymo: $16 billion at $126 billion valuation for robotaxi expansion to 20+ cities. Cerebras: $1 billion at $23 billion for AI chips 56x larger than Nvidia’s GPUs. ElevenLabs: $500 million at $11 billion for voice AI with celebrity voice licensing. Bedrock Robotics: $270 million for AI and robotics in construction. Positron: $230 million for energy-efficient AI inference hardware.

Three new unicorns emerged this week: Waymo, Cerebras, and Goodfire all crossed the $1 billion valuation threshold. The pattern suggests investors are betting heavily on physical AI (robotaxis, robotics), AI infrastructure (chips, compute), and voice/audio AI as the next major platforms.

What the Data Tells Us

Physical AI is having its moment with robotics and autonomous vehicles attracting massive capital. The AI chip arms race is intensifying as companies seek alternatives to Nvidia. Voice AI has reached enterprise scale with ElevenLabs generating $330 million ARR. And notably, AI safety is finally getting real funding.

1:36:31 Key Takeaways

Olga’s takeaway: “I’m going to get OpenClaw.” The combination of Shanee’s non-technical success and clear business results made the case for personal AI agents impossible to ignore.

Chris’s takeaway: The elevated importance of documentation. Every critical business process needs to be codified not for training humans, but for training AI. The feedback loop of running processes through AI reveals holes in documentation that were previously invisible, enabling continuous refinement toward perfection.

The broader theme: we’ve crossed a threshold where AI agents can genuinely replace work, not just assist with it. The companies that adapt—by documenting processes, experimenting with agents, and rethinking their software stack—will thrive. Those waiting for the dust to settle may find there’s no dust, just a new landscape.

Keep Learning