Sprinkle some AI into your Projects
The popularity (and hype) around AI-infused services had many business users asking: “What could we do with AI in our projects?”
News stories are filled with the promise and the fears of AI technology, but for those in smaller businesses, it is sparking conversations about what it is and what it may do for their work. We’ll start by reviewing some of what AI is (and isn’t) and then share some practical examples that companies of all sizes can access right now to help with their projects. We’ll review some of the drawbacks too.
First, AI isn’t a single, all-knowing artificial mind. Today, it is a collection of technologies that are individually-suited to specialized tasks
A sampling of AI-type technologies:
Natural language models, such as ChatGPT, create meaningful text and respond to questions
Object recognition and processing models, such as those used in vehicles to recognize street signs, for example
Decision models, seeking patterns in large sets of data and using those to make predictions about the future or optimal decisions
Fuzzy-logic systems that consider more than yes/no to come up with adaptable responses to problems like, how to most efficiently keep this building at a comfortable temperature
So, what does this mean for your projects?
At the least, it saves your team time by automating some tasks. However, the benefits go beyond that by reducing data errors, allowing your team to focus on more important items, and paving the way to improve how you interact with your customer, suppliers, and other stakeholders on the project.
There are mature AI-type solutions that can help your project with tasks such as:
Recognizing key pieces of text on invoices and automatically entering those into your project controls/accounting systems, rather than relying on user input. These solutions don’t need invoices to follow a strict format; they learn to recognize multiple styles of documents and extract necessary information. Invoices that aren’t processed to acceptable levels of reliability are separated out for review by a human user.
Monitoring project risk factors and creating a score or KPI that helps your team understand whether risk levels are acceptable or may be changing on the project
Sorting through large quantities of inspection data, for example, looking for data points that don’t fit acceptability criteria or deserve a more detailed inspection
Writing project documentation, standards, and contract terms that are clear & understandable, and cover situations you may not have considered
However, there are some key drawbacks to keep in mind
Most AI-based solutions “train” off of existing data and have bias. This might mean a document-processing model might fail on sets of documents that are too different from the ones it has seen in the past, or worse, a system fails to recognize a risk factor because it wasn’t present in any of the past data it used to create its model.
AI-based systems struggle with separating truth from fiction. Language models that use the entire internet as their learning grounds suffer from this problem when there are large amounts of conflicting and/or misinformation available online for a topic. While these models are not consciously lying, trusting them too blindly can lead users to accept automated answers as fact.
Confidentiality and commercial use aren’t a given. Many AI-based solutions use the information provided by users as additional training data, but what if that data is proprietary or sensitive? It is possible for that same information to emerge in undesired places, particularly if the model is public. This means you need to either accept this risk or understand how the information you feed to the solution is used, and that means sorting through the licence agreement. While you’re at it, make sure that agreement fits with your intended use of the system, to make sure you’re not unknowingly stepping into copyright or licencing issues.
It’s a rapidly-evolving area, but businesses that are getting educated and trying out solutions are learning what a fantastic set of tools are available in the world of AI.