👋 Welcome back to AI for SME Success, your weekly dose of practical AI insights and updates that matter to small businesses.
This week, we’re diving into “vibe” designing with Google’s latest tools: what’s new, what works, and real examples you can use.
We also break down fresh AI adoption data that proves building with AI is a massive competitive advantage right now.
On the quick wins side, we cover the underrated power of delimiters in prompting, how they clean up your AI outputs, and how to get AI to add them to your prompt.
You’ll also get a quick breakdown of the best tools for turning messy handwriting into text.
🛠️ Vibe Coding and Vibe Design Updates, Strategies and Examples
Google Updates
Google just dropped two major upgrades: Stitch for design and AI Studio for building apps. Both tools are prompt-driven, no technical skill required, and fully inside the Google ecosystem.
Released March 18, the reimagined Google Stitch lets us design custom client portals, operational dashboards, applications and landing pages by simply describing what we want. The AI handles layouts, visuals, and brand consistency automatically. It supports voice input and live previews. The market noticed, and Figma’s stock dropped 8.8% the day of the announcement.
You can design apps with Stitch, then build them in AI Studio. AI Studio turns designs and prompts into “production-ready applications.” March updates include user authentication, real-time multiplayer, external libraries, and secrets management. Previously, this kind of back-end work would take a developer weeks to put together. Now, non-technical users can prompt for it.
The centerpiece of the update is Antigravity, Google’s coding agent released in November. Antigravity no longer just writes code. It understands your entire project, including chat history, file structure, and change history, and builds with full context.
What Can You “Vibe Code?”
Small businesses are using vibe coding to solve both internal and external needs.
In “Meet the Companies Vibe Coding Their Own CRMs” (The Wall Street Journal), published earlier this month, small businesses are shown building custom CRM systems with AI to better fit their workflows and reduce reliance on traditional software vendors.
Similarly, “I Built a Website for My Daughter Using AI Tool Bolt” (Business Insider) demonstrates how simple prompts can produce fully functional, customer-facing websites.
Vibe Coding Strategies
In “I Joined a Vibe-Coding Workshop to Learn How to Build Apps in 2 Mornings. Here Are My 5 Biggest Takeaways“ (Business Insider), Lee Chong Ming shares the following vibe-coding strategies:
- Don’t get stuck fixing a broken app. Move on to another one!
- The best way to learn vibe-coding is by trying and experimenting.
- Your first prompt is the most important, get it right!
- Don’t be loyal to one AI model. Learn and use different ones.
📊 2026 AI Adoption Statistics
Global AI Adoption
Interesting statistics from February 2026 reveal a staggering divide in the global AI adoption landscape. Out of 8.1 billion people, nearly 6.8 billion never used AI:
- Never Used AI (84%): 6.8 billion people have yet to interact with an AI tool.
- Free Users (16%): 1.3 billion people use free chatbots, likely for basic search or curiosity.
- Paid Users (0.3%): Only 15–25 million people globally pay for premium AI access.
- Use for Coding and Innovation (0.04%): A tiny fraction are using coding and advanced frameworks to innovate.
For those who has access to AI, it is easy to forget how exclusive the “power user” circle actually is on a global scale.
AI Adoption by SMEs
The latest data on AI adoption by SMEs in the US comes from a Goldman Sachs survey of 10,000 small businesses. The key theme is: SMEs are leaning in.
- Early-Stage Adopters (67%): The vast majority are actively experimenting with AI tools.
- Core Integration (14%): This group has moved past the “experiment” phase to fully embed AI into its core operations.
According to the MNP Digital Report: The Business of AI 2026, which recently surveyed 250 Canadian midmarket SMEs, most organizations are further along than you’d expect, but “further along” covers a wide range:
- Operational (48%): Actively using gen AI with solutions in production.
- Active (21%): Experimenting with tools like Copilot, ChatGPT, or Claude for individual productivity.
- Systemic (17%): AI embedded across departments with measurable ROI.
- Awareness (10%): General interest, no structured plan yet.
- Transformational (4%): AI as a true strategic differentiator.
Takeaway
Although AI tools have reached professional-grade operational quality, the percentage of SMEs using AI for design and innovation remains small. This creates an immediate opportunity to outpace the competition by building custom systems rather than just prompting.
🪄Little-Known Trick to Improve Prompt Clarity
Does the format of your prompt actually matter?
According to a recent study by Systima, it does, in about 25% of cases. Proper use of delimiters is covered in most major prompting guides, including:
- Anthropic’s Prompt Engineering Interactive Tutorial
- Gemini Prompt Design Strategies
- ChatGPT Prompt Engineering Best Practices
These guides describe appropriate use of delimiters like XML tags, asterisks, dashes, hash symbols, brackets, colons/labels, backticks, chevrons, dividers, and parentheses.
Here’s an example of using XML-style sections from Gemini Prompt Design Strategies.
But ….. who has time to format every prompt?
You don’t have to.
The trick is to only reformat when the AI misunderstands you:
- Start a new chat.
- Paste your original prompt with this instruction: “Add delimiters to show critical instructions, constraints, reusable templates, examples, non-modifiable data, and variable data.”
- Review the result. The reformatted version will reveal exactly where your original prompt was ambiguous.
- Start a fresh chat with the improved prompt.
After a few AI reformatting exercises, you’ll start recognizing what different delimiter do, and naturally work them into your prompts.
From Paper to Digital in Seconds. No Typing Required.
Did you know AI is very good at transcribing handwritten notes? I asked Claude, ChatGPT, and Gemini to transcribe my workout plan, and here are the results.
The original document was messy and filled with abbreviations, acronyms, and heavy edits.
While all tools made several mistakes, Claude’s transcription was nearly perfect.
Thank you for reading today’s edition!
If this issue was valuable, pass it along to a fellow business owner.
Also, I’d love to hear your feedback, questions, or topic suggestions at natalia@nataliabrattan.com.
See you next week,
Natalia