👋 Welcome back to AI for SME Success, your weekly dose of practical AI insights that matter to small businesses.
This week:
- Google dropped 100 AI announcements at I/O. We cover the top 5 for SMEs.
- A step-by-step workflow a firefighter used to build a shopping app with AI.
- New examples of workflows AI can’t handle well.
- Practical takeaways from the Anthropic co-founders’ interview with Oprah.
- Soon this social network will filter AI-generated slop with AI classifiers.
Let’s dive in! 👇
📣100 Google AI Announcements
5 Top New Google AI Features for Small Business Owners
Last week was huge for Google. At its annual Google I/O conference, the company made 100 announcements covering new products, features, and functionalities, both planned and launched.
Source: https://io.google/2026/
In this section we feature the top announcements that are most relevant to small business needs and available now:
- Pomelli Brand Books and Websites. Pomelli, the AI tool that helps SMEs create on-brand content, can now generate brand books and websites in minutes.
- New Gemini Omni. Google’s first AI video model you can edit conversationally. Available now inside YouTube Shorts Remix at no cost. Limited to short-form by the 10-second cap.
- Stitch Real-Time Design and Voice. Google’s website and app design tool now lets you design by voice or text and watch the AI build alongside you in real time. Brand voice can be preserved.
- New Google Flow Agent. Your AI partner in Google Flow plans and reasons through multi-step creative tasks: brainstorming, creating, editing, batch edits across assets.
- New AI Search Experience. AI Overviews and AI Mode merged into one experience, powered by new Gemini 3.5 Flash. The new Search box accepts text, images, files, videos, and Chrome tabs. Critical for how customers find your business.
My top pick: Pomelli for web pages.
📲 How to Create an App with AI: Step-by-Step
All you need is one paid AI tool and three months of part-time effort.
Joe Poynton, a 44-year-old firefighter from the UK, built an iPhone app called Grocery Flow that learns the layout of your grocery store as you shop. The app then reorganizes future shopping lists to match the most efficient route through the store.
He had never taken a coding course. The entire project took about three months of part-time work between shifts and family obligations. The story was featured in Business Insider.
Poynton’s toolbox:
- Claude (free tier). Coding help, debugging, prompt refinement.
- Gemini (Pro). Planning, troubleshooting, app-development guidance.
- Xcode (free). Apple’s app-building software.
- Swift (free). Apple’s programming language.
His step-by-step playbook:
- Asked Gemini to plan workflow, costs, and tools.
- Had Gemini generate Swift code files for the app.
- Uploaded the code into Claude for feedback and improvements.
- Used Gemini to build, Claude to troubleshoot.
- Turned rough iPad sketches into app graphics with AI assistance.
- Tested the app with friends, family, and online users.
- Submitted the app to Apple’s App Store review and launched it.
⚠️ Where AI Went Wrong
Not every task is AI-ready.
Here are two examples published last week:
- Inventory management: Reuters reports Starbucks scrapped its AI inventory tool nine months after rolling it out to more than 11,000 stores across North America. The app kept miscounting and mislabeling items on shelves, failing to recognize syrup bottles and confusing similar types of milk.
- Investment: New NBER research found chatbot-built portfolios give you no real edge over passive index investing. AI picks skew heavily toward large-cap tech because LLMs lean on media coverage when recommending stocks, producing less diversified results.
💡 Takeaway from Anthropic Co-Founders’ Interview with Oprah
AI capabilities aren’t planned. They’re discovered.
Last week Dario and Daniela Amodei, the brother-and-sister co-founders of Anthropic, sat down with Oprah.
Among many other things, Dario explained that when they build a new model, they test it across many applications. “How good is this thing at translating French to English? How good is this at recommending a stock portfolio? How good is it at finding bugs in software code?”
That last one revealed something striking: their newest model, Mythos, beat human software engineers at finding exploitable vulnerabilities.
This is the key for AI users to understand. Traditional software was built to solve one specific problem. Your accounting software was for accounting.
AI is fundamentally different. An AI model can flop at one task and dominate the next. The same model that nails your customer email might botch your inventory tracking. You won’t know its strengths and weaknesses unless you test.
Takeaway: Never hand your AI a mixed bag of tasks. Pick one repeatable workflow and test it across different tools.
🧹 LinkedIn is Fighting AI Slop
Soon AI Classifiers will filter AI-generated posts and comments.
LinkedIn COO Dan Shapero announced the platform is actively cracking down on AI slop. Watch his full update here.
Key points:
- Building AI classifiers to flag posts that are “really generic, not representing a unique point of view from the person.”
- Applying classifiers to comments, “making sure that comments aren’t people using AI just to summarize a post, but to offer a unique perspective and point of view.”
- Using classifiers to limit distribution of generic content outside the user’s network.
LinkedIn is not alone. Fighting AI slop sits on YouTube’s list of key 2026 priorities. Meta is already labeling AI-generated media across Instagram and Facebook.
Takeaway: Posts with your real stories and points of view. Use AI to edit and sharpen.
Thank you for reading today’s edition!
If this issue was valuable, pass it along to a fellow business owner. I’d love to hear your feedback at natalia@nataliabrattan.com.
See you next week,