🚇 Automation Goes Underground

Hi, - welcome to the second edition of Not a Threat.

This week, I’ve been building systems to automate some of the production process for this newsletter.

My short-term goal is to release more than one a week, so I’m experimenting with tools to help me source and draft these insights.

With any luck, you’ll be hearing from me a lot more frequently!

These are the highlights I discovered over the past week:

• Faster, cheaper, safer - AI right beneath your feet

• Are your employees sabotaging your AI efforts?

• AI in the fields, not in the cloud

• An effective mindset for working with AI

Let’s get into it.

- Sam

🔥 THE GOOD STUFF

CONSTRUCTION AUTOMATION

🚇 AI-Powered Tunnel Fit-Outs: Faster, Cheaper, Safer

Background: While excavation in tunnel construction has become increasingly automated in recent decades, the mechanical and electrical (M&E) fit-out phase of these projects remains largely manual, time-consuming, hazardous, and difficult to scale. The Automated Tunnel Robotic Installation System (ATRIS) aims to change that.

The details:

• ATRIS uses AI-driven vision, robotic arms, and precision mapping to automate tunnel bracket installation.

• Testing showed a 40% increase in productivity and 30% reduction in fit-out costs.

• The system supports sustainability with projected 40% cuts in emissions from reduced equipment and travel.

• It improves safety by reducing manual tool use and lowering the number of workers needed underground by up to 70%.

• A digital twin platform, DATA-IS, captures real-time installation data, enabling faster handovers and fewer delays.

Bottom line: The tunnel fit-out stage was long considered too hands-on for automation. But with ~$50 billion of projects in various stages of construction in the U.S. alone, the potential upside of the technology is significant. The takeaway for leaders: don't assume your hardest problems are too difficult for AI to solve. Start with the pain point, then explore what’s possible.

CHANGE MANAGEMENT

🪓 Employees Admit to Feeling Threatened by AI, Sabotaging Deployment Efforts

Background: As businesses pour billions into AI, many face a stark reality: their workforce may not be on board. A survey by AI startup Writer found that 31% of employees admit to sabotaging their company's AI strategy, with younger workers leading the charge.

The details:

• 41% of Millennial and Gen-Z employees have engaged in AI disruption, often due to fears over creativity and job security.

• A lack of effective AI tools contribute to this behavior, with 35% of employees paying out-of-pocket for those they use at work.

• More than a quarter of workers have entered company information into non-approved tools.

• 72% of C-suite leaders struggling to transition to generative AI, and 34% labeling their efforts as a "massive disappointment."

• Despite challenges, over 90% of employees and executives remain optimistic about AI's potential.

Bottom line: Writer suggests that for AI strategies to succeed, leaders must 1) formalize and invest in a generative AI plan, 2) nurture internal AI champions who can drive change throughout your business, and 3) select vendors who support change management by providing education to employees and helping to shape the AI vision at your company.

AGRICULTURE

🌾 AI Cameras Help African Farmers Save Their Crops

Background: Esther Kimani, founder of Farmer Lifeline Technologies, is tackling one of African agriculture’s biggest problems: crop loss from pests and diseases. In Kenya alone, smallholder farmers lose up to 50% of their harvests each season to these threats—a challenge that affects 33 million farmers across the continent. Her low-cost AI system offers early detection, faster response, and real results.

The details:

• The system uses solar-powered cameras with built-in AI to monitor crops within a 600-meter radius.

• It detects threats in under 5 seconds and sends tailored SMS alerts to farmers, including treatment advice.

• SMS works on basic phones, ensuring the service is accessible to farmers without smartphones.

• It costs $3 per month, far cheaper than drone scans ($100/hr) or agronomist visits ($60+).

• Data from devices feeds a central dashboard that supports regional pest outbreak tracking by local authorities.

• Users report up to a 40% yield boost and a 30% reduction in crop loss.

Bottom line: Kimani’s model combines smart tech with radical accessibility. The real lesson for leaders everywhere: the cost of powerful AI is dropping fast, opening the door to high-impact, low-cost solutions in places and sectors that were previously out of reach.

AI STRATEGY

📘 4 Ground Rules for Working with AI from Co-Intelligence by Ethan Mollick

Background: Ethan Mollick’s Co-Intelligence is a practical guide to using generative AI. Drawing on his experience as an entrepreneur, Wharton professor and early AI adopter, he frames AI not as a replacement for human work but as a partner. In the book, he offers four clear principles for building that partnership.

The details:

1. Invite AI to the table: Use AI in everything where it’s not legally or ethically off-limits. The best way to understand its potential is to experiment. You’ll learn what it’s good at and where it’s still weird, wrong, or just plain useless.

2. Be the human in the loop: Even as AI improves, human oversight remains critical. You’re responsible for its outputs, and you need to catch its mistakes, steer its behavior, and ensure it aligns with values and norms.

3. Treat AI like a person (and define its role): Give it a role or persona - it’s not magic, just better when given context. Tell it to act like a lawyer, designer, or strategist, and it’ll return sharper, more relevant results.

4. Cultivate your own expertise: Don’t outsource your brain. AI can boost your abilities, but only if you bring something to the table - domain knowledge, critical thinking, and the skill of working well with the machine.

Bottom line: This isn’t a book about tools, it’s about mindset. Mollick makes a sharp case for hands-on learning, thoughtful oversight, and the value of human expertise in an AI-integrated world. For leaders trying to bring AI into their orgs, these four principles offer a no-nonsense starting point - but the book is well worth a read (and accessible for non-technical people like me!).

👩‍🎨 IS IT ART?

PAINTING

🤔 Which of these pictures was painted by a human?

Background: OpenAI recently launched native image generation inside ChatGPT, allowing users to create and manipulate images without switching tools.

The details:

• Users can now ask ChatGPT to generate high-fidelity images in a wide variety of styles, or by combining text and reference images.

• Image quality is a significant jump: accurate text in images, photorealistic lighting, layered depth, and stylistic nuance are all handled with surprising competence.

• Users can edit existing images with prompts like “remove background,” “make this into a comic,” or “add glasses to the dog.”

• It’s slow. Each image can take minutes to render, limiting use cases for now but OpenAI says speed will improve.

• I used one of these images with a basic prompt to “mimic the painting style, scene, and period of attire for the subjects.”

Bottom line: You tell me - is it art?

👋 THAT’S ALL, FOLKS

FEEDBACK

If you’ve made it this far, I’d love to know what you thought of this edition—or others. Good or bad, give me your feedback below! 👇

What do you think of Not a Threat?