👋 Welcome back to AI for SME Success, your weekly dose of practical AI insights that matter to small businesses.
Today we cover:
- How to make the best use of Claude Fable 5
- Why corporations reverse AI job cuts
- Google updates for creating visuals
- How to mine Reddit for business insights
Let’s dive in! 👇
💼 Fable 5: Capability, Cost Optimization and Prompting
Anthropic’s most capable model is back and tops the client-ready benchmark.
Fable 5 runs on Anthropic’s most capable tier, It was briefly suspended on June 12 to comply with US export controls, and then the access was restored on July 1. Here is what we need to know about it.
The Most “Client Ready” Model
Most AI benchmarks are exam-style: math problems, coding puzzles, or multiple choice questions. Many models now ace many of those, but they fail to produce economic value. The Remote Labor Index (RLI), a benchmark created by CAIS, measures end-to-end economic output, reviewing how the work product is client-ready. Here is where Fable stands on the RLI.
Reference: Remote Labor Index results published by CAIS and Scale (July 1, 2026).
How to Optimise Fable 5 Cost
Fable 5 is one of the most expensive models, roughly double standard ChatGPT 5.5 and Claude Opus 4.8.
The trick many users have found: pair it with a cheaper model. Use Fable for planning and quality control, a cheaper model for execution. Mitchell Hashimoto runs Fable as architect, GPT 5.5 for execution, Fable again as judge. Shalini Goya uses it for writing, planning, skills, and goals. Aniket Panjwani plans and improves his highest-value projects first.
How to Prompt Fable 5
Here is the full prompting guide for Fable 5. A few important takeaways:
- Assign your hardest problem. According to Anthropic: “simpler workloads tends to undersell its capability range.“
- Tell it why, not just what. The model “tends to perform better when it understands the intent behind a request.“
- Retire your old mega-prompt. Start short; add rules only when something breaks. Old prompts are “often too prescriptive… and can degrade output quality.”
🔄 Big Companies Reversing AI Job Cuts
AI errors and shortfalls accumulate over 6 to 18 months.
Ford just rehired 350 veteran engineers to fix vehicle quality problems. The company assumed feeding AI its design requirements would produce a quality product without experienced oversight. It did not.
Similarly, Klarna and Commonwealth Bank of Australia rehired customer service representatives cut during AI automation pushes.
IBM automated most of its routine HR requests with AI, then announced in February 2026 it would triple entry-level hiring. Its CHRO noted that without junior hires, the talent pipeline needed for sensitive situations dries up.
According to a 2026 Careerminds survey, only 8.4% of HR leaders said their AI-driven restructuring delivered exactly what was promised. Orgvue’s 2025 survey found 55% of business leaders who cut jobs for AI admit those decisions were wrong.
Takeaway: AI mistakes and hidden problems tend to accumulate over time. Large companies notice them within six to eighteen months of AI deployment. Small businesses have less tolerance for errors. Be cautious about initial AI gains.
🎬 Google Updates for Creating Visuals
Google shipped budget image and video models plus 60-second video overviews for NotebookLM.
On June 30, Google released three updates that drop the cost for images, video, as well as video overview clips:
- Nano Banana 2 Lite, its fastest and cheapest image model, generates a 1K-resolution image in about 4 seconds for $0.034 (roughly 29 images per dollar). It comes with a Nano Banana prompting guide.
- Gemini Omni Flash generates and edits 10-second video clips through plain-language instructions at $0.10 per second of output.
- NotebookLM Short Video Overviews turns your uploaded documents into 60-second vertical videos.
NotebookLM Short Video Overviews is the most useful update for small business owners. You can use it through notebooklm.google.com, but not within Google Gemini. For now, it requires a Google AI Pro or Ultra plan and generates in English only, with free access promised soon. These overviews can be a starting point for your reels, which you can edit and blend with real videos in your video editor.
🎯 How to Mine Reddit for Business Insights
Step 1. Find Reddit threads relevant to your business
Your can search within Reddit or Google
site:reddit.com YOUR TOPIC after:DATE before:DATE
Example:
site:reddit.com AI small business after:2026-06-04 before:2026-06-30
Google will return URLs to conversations on your topic. Choose the most relevant one and open it.
Step 2. Extract the entire conversation
You can copy every comment from a thread by adding /.json to the end of the URL.
It will look like this:
Step 3. Analyze the entire conversation with AI
Copy this messy data into the AI tool of your choice and tell it what you want to find.
You can look for pain points, product ideas, hooks and topics for your social media, things that entertain your audience, trends, or selling points.
Do this regularly!
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!