Awkward conversations, AI inspection uses, and AI projects gone wrong.

Welcome to another edition of Grounding AI. We are catching up from last week spent at TechCon365 in Washington, DC. where I conducted an AI for the non-AI professional workshop to enable information workers to get up to speed on AI. Let's get to it!

Don’t Let Awkward Conversations Cost You Great Employees

Recently, we were demoing one of our AI-based solutions to a client. The client executives were hoping to use the solution to automate most of the manual work of a large team. An awkward conversation between Tom, the team manager, and executives ensued, when Tom asked, “What will my team do now?” The executives seemed surprised by this question.

In less than 30 seconds, I feel the executives lost the trust of a great employee as they stumbled through an unplanned answer. I wouldn’t be surprised if Tom updated his resume when he got off the call. I certainly would have.

It is imperative that the human aspects of this technological change be in the forefront of discussion. AI has the potential to radically change how we do business but only if we address the human aspects as well.

I disagree with those executives who feel AI will allow them to replace humans for tasks. My contention is that if you do this, you get no gain from the technology, other than cost savings. You also lose the cumulative knowledge that the replaced employee has.

Instead, imagine what your existing workforce could do if efficiently enabled with AI enriched capabilities? The true value is in giving your employees superpowers to enable you to thrill customers and outperform your competitors.

We talked quite a bit about these topics in my AI for non-AI professionals workshop as I feel it’s important to fully understand what’s at stake here. If you would like to join the conversation around this, I’m re-running the workshop as a paid webinar on Sept 9th at 8am PT. Register here: https://marqueeinsights.com/ai-for-non-ai-pros-webinar/

Inspection Use Cases of AI

I’ve talked to a lot of companies about how they would like to use AI in their organization. One of the most common uses I’ve encountered is inspection. Vision AI can see more than the human and can compare accurately to known state to spot differences. Let’s talk about three use cases of which I’ve worked with clients to investigate.

Windmills

Likely you’ve seen those massive windmills in wind farms around the world. It turns out they get damaged by wind, hail, lightning, etc. and require frequent inspection. Drone technology was an amazing step forward in inspecting this infrastructure as one person can run a program to have the drones fly round the structures and video them. However, humans had to review the video to note problems and log tickets. It’s very time consuming and prone to errors.

Using Video AI, a model could be trained as to what the normal state of the infrastructure should be and can “watch” the video to spot discrepancies. We could layer more AI on top of this discovery to figure out the type of issue and execute workflows to log tickets automatically. As connectivity continues to improve in the remote areas of these windfarms, streaming video from drones could be used for this activity.

Hospital rooms

The speed in which you can reset a hospital room directly affects the effective capacity of a hospital to serve patients. If you are waiting on a person to inspect the room once a person is moved from it, there can be delays and the person may miss something. Likely the person then must manually enter the information to log tickets.

What if you had a robot that is automatically dispatched to do a reset scan of the room? It could use Vision AI to spot discrepancies, then other AI can be used to classify and act upon the needed items. When integrated with existing systems, this could shorten the time needed to turn over the room.

You could also use this same approach to scanning hallways and public areas for maintenance. AI could potentially bring any public-serving facility to the level of cleanliness found at a Disney property.

Aircraft

When planes are highly utilized, inspection of the current state vs the ideal state is extremely important. We’ve all been on flights delayed due to unexpected maintenance. Human inspection is always performed and will catch a number of items. However, we are limited to the frequencies of light that we can see.

Using Vision AI and cameras that could see at wavelengths outside of our vision, a more thorough inspection may be possible. Minute cracks could be found before they become bigger issues. You could also tie this inspection process with the current part supply data to quickly figure out if and when an issue can be resolved.

In each of these three scenarios, we are decreasing response time to incidents. There are several other scenarios like this that you should consider in your own organization. These apex problems tend to cause delays later in your business processes.

AI projects gone wrong?

Suddenly, all the AI project news is souring, and the reasons being given are not surprising. It’s not AI that’s the problem but rather project issues that have been around for over a decade. While we advocate crawl, walk, run approaches to AI projects, but some have chosen to start with skydiving instead. While you may be successful, this can lead to some risk management challenges as well.

According to Gartner, up to 30% of AI projects will be cancelled by year’s end due to poor data quality, inadequate risk controls, escalating costs, or unclear business value. ?https://www.gartner.com/en/newsroom/press-releases/2024-07-29-gartner-predicts-30-percent-of-generative-ai-projects-will-be-abandoned-after-proof-of-concept-by-end-of-2025 ??The Register reports similar challenges in Copilot implementations in their article found here: https://www.theregister.com/2024/08/21/microsoft_ai_copilots/

These are not new challenges specific to AI. Every business intelligence project of the last 10 years has met one or more of these issues. The difference with AI is you meet these problems faster. If you are a SpaceBalls movie fan, we liken BI projects as causing “chaos at light speed” on occasion. With AI, the risk is that you could create “chaos at ludicrous speed.” We’ve advocated the need to create smarter data via medallion approaches, certification processes, and security reviews for over a decade. It’s even our company tag line, “Smarter data, better decisions.”

I think there are two forces at play here which are creating these situations. The first is the perceived opportunity to reduce cost “easily.” ?I feel this is because many of these projects are being approached as outsourcing projects, where the company is outsourcing work to a machine instead of another country. Our contention is the projects that will succeed eventually are the ones focused on enabling employees to do so much more, not on reducing headcount.

The second is the desire to be seen as market leading where core data and security hygiene processes have been ignored. If you’ve ever watched Kitchen Nightmares or Bar Rescue TV programs, you know you can’t be successful if your organization isn’t organized or managed effectively. Perversely, organizations that have recently experienced a ransomware attack may be best positioned for an AI project, simply because they’ve recently reviewed their security and data protocols.

If you are looking to pursue an AI project, there are risk management and planning approaches that should be used that are specific to AI projects. We recommend that clients view a year-long project as 4 quarter-long projects, simply because the underlying technology will change in that time, and you should plan for course corrections.

Thanks for reading!

I hope you’ve found value in these discussions. If you know of someone who can benefit from discussing AI from a unique perspective, we’d appreciate it if you could pass this newsletter on. Thank you for reading!