AI workflow from the high-level

by Chris PehuraC-SUITE DATA — 2024/05/18

Email us at to see how we make your plans work.

When someone mentions AI to you, what do you think of?

Me, I think about my early experiences as a kid programming video games. There was one game where I wanted the hero to be swarmed by zombies. But the zombies weren’t that smart. They were very easy to evade. To fix that, I checked out a book out of my local library on game trees. I was from a small town. There wasn’t much selection on AI back then. Even in the larger cities like Winnipeg there wasn’t much selection. It’s not like today where all the resources are available with a Google search. From that book, I learned how to program game trees using BASIC on a COLECO ADAM. I wasn’t that old, I think about 10 at the time.

I was introduced to AI because of dumb zombies. What kept me hooked on AI is that AI is a puzzle. And I love puzzles. AI involved performance tuning. It involved optimization. That’s probably why I ended up with a degree in computer engineering and not computer science. Today, I’m a business engineer and I work with how AI fits within the company’s bigger picture. The one thing that I’ve seen with AI for all these years is that AI solutions are made up of specific components. It really didn’t matter which kind of AI it was. Traditional AIs, generative AI, and large language models. They all used those key components.

There are at least four typical components of an AI solution. Each component represents a group of activities. Optimal AI solutions focused on restricting the flow from left to right. This was to significantly reduce the complexity of the design. These components or stages do not need to be in the same order laid out in the diagram.


This component is a group of activities to help identify or assess something. Identify as in identifying a specific car part, position of a part, a person’s face. Assessments are a form of identification. You want to assess a patient when they go into a hospital. You want to assess their likelihood of developing later stages of cancer. For Identification, you cannot assume AI is accurate. This is what the red boxes are for. Those are things where the AI can go wrong. False positives, false negatives, or the AI receiving the wrong input. Rather than AI reclassifying, reidentifying, reassessing, you should move forward to the next component in the flow.

Risk Mitigation

This component mitigates and reduces the risk of damage when the AI is not behaving the way it should. If things get too bad with the risk, you can always have that intervention kill switch. Risk mitigation primarily involves laws, regulations, and reducing harm like in a hospital. When things are okay from a risk mitigation viewpoint, you move forward to the next component in the flow.


This component asks the question — “why is the AI in place?”. Is it there to help increase the value of something? Help with generating revenue? Is it there to reduce the costs? I always consider what harms an AI can have and focus on reduction of those harms. It’s not uncommon for companies to include litigation here. And for good reason when you see the big hiccups by big AI names. Once objectives are determined and things are routed according to what objectives need to be met, it’s time to move to the next component in the flow.


This component is where AI is delegating work and effort. This can be delegating to workers specific activities, tasks, or providing information like notifications, instructions, and recommendations. The AI could be redirecting instructions to agents, less intelligent AI, and robotics. For instance, telling robots to move one way when there’s no people around and then another way when there’s people around.

Today’s Challenges

The technology behind AI really hasn’t changed that much for the last few decades. The components of this workflow really hasn’t changed either. What has changed is the tech that supports AI. Tech like heavy data processing and the ability to store massive amounts of data. That’s what changed. That’s what makes our AI solutions today even possible. Our tech challenges for AI melted away. We have plenty of tech, plenty of options, plenty of data. Today, the problem is with skill sets.

We need people to know how to work with AI and how to properly design and delivery AI solutions. It’s getting more difficult to identify those people. It is not uncommon that you can talk to people who have PhDs in AI and they don’t know the basic fundamental concepts of AI. AI is a significantly big field and a lot of schools are focusing more on a breadth approach rather than the fundamentals of AI. There’s nothing wrong with schools doing the breadth approach. It gives students the ability to go in deeper when needed. It allows graduates to go into your company, take an off-the-shelf AI solution, and have it start working. But we need fundamentals. We need people in our companies to know how to do improvements and and further optimizations to make the AI work even better.

Key Take-Away

Everyone is getting those same AI off-the-shelf solutions. You, as a company, can’t remain competitive unless you can do a better job with the AI than the other guy. You must consider how to instill the AI fundamentals within your company — in your business people, your tech people, and especially your AI people. AI has raised the bar for all of us.

Email us at to see how we make your plans work.