I’m Trevor Hendricks (LinkedIn) and I’ve been designing and writing software for most of my life. I want to offer a practical viewpoint on Artificial Intelligence and its applicability to commercial laundry route operators in particular.
My view is this, humans create tools often called machines for their own purposes and evaluate their usefulness in terms of how well they meet the purpose. Simply put, we use tools and if they work for us, we keep using it until something better comes along.
AI is the latest and greatest of our tools and we must keep in mind that it’s a tool. Some of us are old enough to remember when computers threatened our lives, the robots were going to take over. Calculators were going to make everyone stupid. Television was going eat our kids’ brains. Social media is toxic. The internet is dangerous. All of this is true and false, in a given context.
Context, is how we select our tools. It defines what we’re talking about. To cut down trees we select tools like axes and chainsaws. Your calculator, Excel or AI isn’t going to add to the pile of lumber but what they can do is help add to the pile of money. That is if, the AI tool has been trained properly and if you know the right questions to ask. Effective use of the tool is incumbent on the user. Whacking a chainsaw against a tree as you would an axe is an indictment of the user not the tool. You have to know what you’re asking before any answer your tool gives can help you.
To make AI useful in a given context it must be trained. To do that effectively, all data that’s related to the topic must be available. The AI models commonly available are generally trained on data available from the internet. In this sense, AI is a search engine on steroids and will yield nothing more than what’s already available.
To be useful to businesses such as route operators, the AI will need access to data from that market. The more sources and detail the better. That data becomes the basis of the training the expert will provide the AI. The expert designs the model that represents the business, so naturally their experience is the utmost importance. Still the AI will only understand the route operator business; it won’t understand your business without your data. This is what differentiates AI from your assistant.
In part 2, I go into detail on what to look for when selecting and assistant.
In the first part, I talked about taking a business, such as a route operator, from generic AI to Assistant. In this part I’m going to drill down on how to do that.
In business, competitive advantage is the object of the exercise, how to get it and keep it. Profits and client retention are the indicators of success.
Context is how we select our tools. In this case, we’re talking about software. In order to qualify as an assistant the AI would need to be able to get answers to questions that you’ll ask. It should also be able to make suggestions. It should be able to use your current data. Take note, AI can generate incorrect or misleading responses which are called hallucinations. These errors can happen for various reasons, such as limited training data, poor model architecture, or ambiguous input.
Next you should evaluate the answers it gives to key questions. The more clearly you ask the question, the better the response will be. That’s because you’ll be reducing the ambiguity. The quality of the data and model that the assistant is trained on determine how specific an answer and suggestions it can provide. For example, an assistant trained only on publicly available data and models cannot be expected to provide answers specific to your business. Without your data, it cannot provide answers to simple data-based questions like what’s my daily collection average? Who is my best customer? It also cannot answer questions to unstructured data sources like your contracts or customer email feeds. Finally, if the model is not based on your business, the value of the answers is limited. Delivery routes, vending routes, logistics and commercial laundry route operations are similar to some degree however; when you’re relying on the answers for profits and client retention, you’ll want the assistant to be talking about your business.
Once your questions are answered, you’ll need to have the confidence in them to apply it to your business. That’s not as easy as it sounds because it requires commitment and action on your part. Knowing what to do is not the same as doing it.
In part 3, I go through the specific questions you should ask.
In the previous parts, I talked about what AI is and what to expect from it. Context is what we’re talking about. The data and model define the quality of the answers we can expect. The specificity of the questions you ask will determine how actionable the answers are.
In this part, I drill deeper on the key questions route operators usually ask to improve profits and client retention. Below I have listed some of the most common questions asked and the data sources that need to be accessed to get actionable answers.
Assuming the data is available, the model is used as a framework for formulating the answers. How well the expert who built the model understood the route operator business determines how the assistant will interpret the data. Poor models will talk about delivery and pick up routes rather than service and location-based revenue collection. Unless the model is based on the route operator business, contract settlement will be rudimentary.
Ideally, there will be mechanisms to update the model as well as the data. This ensures that the assistant can respond to changing business conditions.
In Part 4, I suggest questions commercial laundry route operators should ask.