The AI Cuss Jar

No doubt, there is much hype about AI. That’s not a bad thing, because excitement drives exploration and exploration drives innovation. From television to social media, advertisements for products “powered by AI” abound. Similarly, in many boardrooms and shop floors, “AI” has become a catch-all phrase—used so broadly that it loses meaning.

What if your organization had an “AI cuss jar”?

Every time someone says “AI,” they would have to put a dollar in—unless they can explain exactly what they mean. No buzzwords. No abstractions. Just a specific use case solving a specific business problem.

Beneath the hype of LLM (Large Language Models, including chatbots such as ChatGPT and Gemini), there’s real substance in Narrow AI use cases, and it’s often far more practical than people realize.

For example:

1. Machine vision for label verification. A camera positioned over a conveyor confirms that every product label is placed within tolerance—correct location, correct orientation, correct SKU—before the item leaves the line. No language models involved. Just image classification and rule-based thresholds preventing costly rework.
2. Predictive maintenance on critical equipment. Vibration and temperature sensors feed into anomaly-detection models that flag when a bearing is likely to fail. Maintenance is scheduled before downtime occurs, reducing costly emergency repairs and protecting throughput.
3. Demand forecasting using historical sales data. Time-series models analyze seasonality, promotions, and order history to improve purchasing decisions. The outcome? Lower carrying costs and fewer stockouts.
4. Automated invoice matching. Optical character recognition extracts invoice data and compares it to purchase orders and receiving records. Exceptions are routed to a human for intervention. Routine matches close automatically, shortening the cash cycle.
5. Dynamic route optimization for field service. Algorithms recalculate technician schedules based on geography, urgency, and traffic conditions—reducing drive time and improving first-time fix rates.

Notice something: none of these examples require a chatbot. None depend on generative text. They are targeted, measurable applications of data and automation aligned to operational goals. Chatbots are fantastic at drafting content, conducting research, and countless other clerical tasks, but embedding a LLM in the flow of a mission-critical business process could be unpredictable and risk-inducing.

Here’s a fun challenge for you: use your favorite chatbot (e.g. ChatGPT) to ask how AI can be used to solve a business problem without the use of a chatbot. For example, ask ChatGPT “what AI model would I use to determine if an invoice should be paid without using an LLM?”

When we encourage our teams to stop saying “AI” and start saying “machine vision to prevent labeling errors” or “predictive maintenance to reduce downtime,” the conversation changes. It becomes grounded, fundable, accountable, and predictable. We gain a clearer picture of the ROI and the path to improvement.

The AI cuss jar is not about limiting ambition. It’s about sharpening it. If we cannot describe the problem, the data, and the expected outcome in plain language, we should consider diving deeper before investing. Specificity is where the real value lives.

Join us at the Regional Technology Council (RTC) virtual lunch and learn later this month to dive beneath the surface of AI and understand how it is solving real world challenges with specific models.

https://www.eventbrite.com/o/regional-technology-council-inc-111793860201