Video Post: AI is not new

Key Takeaways

• AI in business predates LLMs by decades: Enterprises have quietly used machine learning (ML) for years to power fraud detection, e-commerce recommendation engines, and supply chain logistics.
• Practical automation beats novelty: Even in 2016, a Japanese cucumber farmer used TensorFlow on a Raspberry Pi to automate cucumber sorting—saving 8 hours of manual labour per day.
• Consumer gimmicks don’t drive P&L: Generating novelty images doesn’t make a business money; the real upside of the LLM wave is executive attention on automating operational bottlenecks.
• Compound small, measurable wins: Stacking multiple targeted AI projects that each deliver 10–20% operational improvements creates an outsized impact across the entire organisation.

---

Why Enterprise AI Is Evolution, Not Magic

1. Proven Machine Learning Use Cases Behind the Scenes

Long before generative AI headlines, production machine learning models were solving high-value commercial problems:

• Fraud detection: Scoring transactions in milliseconds to block anomalous activity.
• Recommendation engines: Suggesting relevant products to lift e-commerce conversion.
• Logistics & supply chain optimization: Routing deliveries and forecasting inventory demand.

2. The 2016 TensorFlow Cucumber Sorter Lesson

When Ben joined Google in 2018, one of Google Cloud’s classic 2016 case studies featured a Japanese cucumber farmer with no formal software engineering background. His mother spent 8 hours a day manually sorting cucumbers by size, shape, and quality. Using TensorFlow on a Raspberry Pi connected to a camera and servo motor, he built an automated classifier that sorted cucumbers automatically—reclaiming a full 8-hour workday every single day.

3. Compounding Practical AI Wins Across Your Organisation

The generative AI boom has woken boards up to what AI can do, but profitable engineering isn’t about novelty prompts. It is about identifying repetitive, high-friction workflows across your business and applying the right AI and machine learning models—whether classical ML, computer vision, or production AI agents—where we can tie fees directly to verified business outcomes.

---

Full Video Transcript

So here’s the thing about AI: it’s not in any way new. Most businesses have been using it behind the scenes for years to solve real-world business problems—things like fraud detection, recommendation engines when you’re shopping and it’s suggesting similar products you might be interested in, and optimizing logistics. All different kinds of businesses have been optimized with AI; it just used to be called ML (machine learning).

When I joined Google in 2018, there was a case study from 2016 of a Japanese cucumber farmer who built an AI model on TensorFlow and ran it on a Raspberry Pi to optimize his cucumber farm. His mum used to spend 8 hours a day sorting cucumbers. This guy was not from a tech background, but he developed a solution with a camera and motor that sorted the cucumbers automatically—saving his mum 8 hours every day she sorted cucumbers.

When LLMs came out, there was a lot of realization in the consumer space that AI can achieve impressive things, but businesses aren’t really going to make any money from generating a photo of a cat on the moon or Chuck Norris fighting dolphins—which is funny, but not very profitable.

The good thing about the LLM craze is that it has brought attention from businesses to practical ways they can optimize and save money with AI. If in 2016 a Japanese cucumber farmer could save 8 hours per day—an entire day of labour for one person—with AI, there are definitely ways that most businesses can benefit from this. It’s just a matter of uncovering those bottlenecks and applying the right AI solutions to them.

We’ve been working with a number of customers on projects that deliver incremental, double-digit improvements—and when you stack several of these targeted projects together, they have an outsized impact across the organisation and across the industry. I’d love to chat to anyone who has a business problem they think is solvable with AI and see how we can help solve it. Thanks!