OpenAI’s Agent Phone, Healthcare AI Updates, and How to Build Better Business Workflows
Dear Friend,
Last week in our Applied AI Session, we went deep into building workflows instead of just playing around with prompts. The biggest takeaway from the group was to start simple. Whether Hudson was using NotebookLM to instantly turn class modules into study guides, or Hailie was mapping out a multi-step newsletter process from Google Docs to Canva, the core lesson remains exactly the same.
The leverage happens when you identify a tedious task and build a detailed standard operating procedure around it. I do not use AI to simply write a little faster. I stack my work, setting agents loose on the heavy lifting while I focus on the final polish. And as Ameen and I emphasized in class, you must keep a human in the loop for review. AI is your brilliant intern, but you are still the boss.
This week’s news shows the entire tech ecosystem shifting toward this workflow-driven reality. Let’s look at what actually matters for your business.
This Week’s Stories
The App-Less Smartphone
What happened: OpenAI is reportedly developing a smartphone designed around AI agents rather than traditional apps. Instead of clicking icons to open isolated programs, users will ask the system to perform cross-platform tasks directly.
My Take: The interface of the future is conversational. If your business relies on users navigating a clunky mobile app, you need to start thinking about how your services can be accessed by an AI agent. The companies that make their data easily readable by these new systems will win the next era of mobile, search, and sales.
Healthcare and Voice Clones
What happened: Scientific American highlighted how AI voice technology and behavioral analysis are making their way into highly sensitive environments, including crisis and suicide hotlines. The tech is being utilized to analyze distress levels and offer real-time tactical support for the human operators managing these critical calls.
My Take: Healthcare AI is moving past simple administrative tasks and into direct patient interaction support. This shows that when implemented thoughtfully, AI does not replace the human touch in high-stakes environments. It actively supports the human operator, ensuring they have the best possible data to provide care.
Estimating the Real Cost of AI Power
What happened: Researchers at MIT have developed a faster way to estimate the massive energy consumption of generative AI models. This new framework allows data centers to accurately predict the power cost of running complex workloads before they even execute them.
My Take: many employees at companies where we deliver training report that energy usage and environmental impact is their a huge concern to them. Power consumption is the hidden tax on your tech stack. As businesses scale their AI usage, understanding the infrastructure costs and impact of these tools will become a core operational requirement. Start paying attention to the efficiency of the models you choose, not just their capability.
The Changing Reality of AI Regulation
What happened: Recent compliance reports show a massive divergence in AI regulation across the United States. While federal frameworks remain largely broad proposals, individual states are rapidly passing strict laws targeting automated decision-making and data transparency, particularly for high-risk systems.
My Take: You can no longer afford to ignore the regulatory landscape. If your company uses AI to make decisions about hiring, loans, or customer access, the legal exposure is real. Stop treating AI like a wild west experiment and start building proper governance policies into your standard operating procedures. Let us know if you want help with this. We’ve got some free resources and can help you build your internal and external AI use policies.
Video Scraping and the Legal Backlash
What happened: A prominent creator has filed a lawsuit against Runway AI, alleging the video-generation company actively bypassed technical safeguards to scrape massive amounts of video data. The suit claims Runway used proxy services and automated crawlers to evade detection and train their models without permission.
My Take: The honeymoon period of scraping the entire internet is over. We are seeing a distinct shift in the courts where reckless data acquisition is facing serious consequences. For businesses, this means you need to care deeply about the provenance of the tools you use. Stick to platforms that are transparent about their training data to protect your own commercial output.
Source: Benesch Law AI Reporter
Tool Picks of the Week
- Workflow & SOP Creator: We built this tool to help you systematically map out exactly how your tasks happen.
- Jasper.ai: This remains one of the most reliable platforms for consistent, brand-aligned writing.
- HeyGen: If you need to scale video outreach or internal training, this is the tool our community is actively testing to automate their video workflows.
AI for Solopreneurs: The Best Tools When You’re a Team of One
By leveraging an affordable AI stack typically consisting of generative writing assistants like Claude or ChatGPT, workflow automation tools like Zapier, and knowledge management systems like Notion solopreneurs can reclaim many hours weekly and increase client capacity by up to 43%. This “team of one” approach focuses on automating high-volume, repeatable bottlenecks such as content production, research, and administrative coordination, allowing solo founders to compete effectively with larger teams for a modest monthly investment of $55–$80. Success requires a “depth before breadth” strategy, where users master a few core tools to ensure high-quality output while avoiding the risk of over-automation, which can erode the personal touch essential for client relationships. Ultimately, the most impactful AI implementation targets the specific tasks that consume the most time but generate the least revenue, transforming the solo business into a high-capacity, high-value operation.
Read the full article
Google Labs 🧪
How to set up, use, and master Google NotebookLM
NotebookLM is a private, AI-powered workspace that acts as your personalized research assistant by basing its responses entirely on the specific documents you upload.
1. Set up your Notebook: Create a new notebook and immediately give it a clear title so you can easily identify it on your dashboard.
2. Upload your sources: You can upload up to 50 sources per notebook (or up to 300 on the paid version). Each source can contain up to 500,000 words or be up to 200MB in size. Supported formats include:
- PDFs, Google Docs, text files (.txt), Microsoft Word (.docx), Markdown (.md), ePub, and CSV.
- Google Slides (up to 100 slides) and Google Sheets.
- Webpage URLs, public YouTube videos (it reads the text transcript), audio files (MP3, WAV, etc.), and images.
3. Discover new sources: If you need more information, use the Fast Research tool to search the web or your Google Drive based on a query, or toggle on Deep Research (available to users 18+) to have the AI browse hundreds of websites and compile a multi-page report for you to import.
Extract Insights with the Studio Panel
Once your sources are uploaded, the AI will automatically generate summaries for them. You can then open the Studiopanel to transform your materials into engaging formats:
- Turn your sources into a podcast-style discussion. You can customize the length and choose from formats like Deep Dive (in-depth discussion), Brief (a quick summary by a single host), Critique (evaluating your material), or Debate (exploring opposing perspectives). You can even use Interactive mode to “call in” and ask the AI hosts questions using your voice while the audio plays.
- Create a video slideshow featuring AI narration to present your topics visually.
- Instantly generate interactive Flashcards and Quizzes to test your comprehension.
- Generate Briefing Docs, FAQs, Study Guides, Timelines, and Mind Maps. Mind maps allow you to visualize complex connections and can be downloaded as image files.
Chat with Your Sources and Verify with Citations
The chat interface allows you to query your uploaded documents as if you have a research assistant by your side.
- Type your questions into the chat bar. You can tailor the output by using prompts like “explain the concept of superposition like I’m 10” or by adjusting your conversational style settings.
- If you want to study actively rather than just get answers, use the “Learning Guide” feature. It acts like a personal tutor by asking you open-ended, probing questions to help you break down problems step-by-step.
- When NotebookLM answers, it provides inline citation numbers (e.g., or) next to its claims. Always click these bracketed numbers—they will jump directly to the exact passage in your source document so you can read the context and verify the AI’s accuracy.
Take Notes and Organize Your Workspace
- The chat interface does not automatically save its outputs. If you get a useful answer, click “save to note”.
- You can take the notes you’ve saved and convert them into brand-new sources within your notebook.
- If you have 5 or more sources, NotebookLM can automatically categorize and label them by topic to keep your workspace organized, or you can manage these labels manually.
Join the Conversation
Got a burning question, a fresh take, or just want to share your latest AI wins? Hit us up at [email protected]. Your insights keep this community growing and thriving!
See you in the Lab,
-Nicole A. Donnelly
Founder, AI Smart Ventures
AI Strategy – AI Training – AI Consulting – AI Implementation
