The What, Why, Where and How of the AI conversation
“We use AI.” “It’s AI-based.” “This is an AI-powered solution.”
If CEOs, CIOs, CFOs and their managers had a pound for every time they heard these statements, they could solve our national debt. Misleading AI declarations and vastly overstated claims are commonplace in today’s technology conversations. A technologist may have one understanding, a product manager another and a salesperson yet another. I personally witness misunderstandings on a daily basis that are caused by our imprecise, and often inaccurate, use of the term.
Below I’ve outlined some clear and focused guidelines to help you respond to the inevitable phrase, “We use AI.” In this article, I’ll use the “What, Why, Where, When, How” sequence to help facilitate conversations with your software vendors and consultants.
What is AI?
The concept of Artificial Intelligence (AI) was developed by scientists and mathematicians in the 1950s who explored the mathematical possibilities of machines making decisions with human-like intelligence. At that time, computers could not store commands, they just executed instruction sets. But the dream and concept of AI was alive and kicking: it was the ability of machines to perform human tasks.
Today, AI has become a term that is pervasive in our day-to-day business lives, overused (often incorrectly) and misunderstood by many. A key first step in evaluating AI solutions and vendor offerings is to ask how a vendor or consultant defines AI, in order to set the ground rules. Is it the simplest version of a machine performing human tasks? Is it predictive? Or are we talking about Skynet and self-awareness? (Terminator reference – I’m a big fan). Establishing this mutual definition will provide a solid grounding for all follow-up conversations and help to manage expectations.
Why Do You Use AI?
Adopting AI for the sake of its buzzword value is likely to be a waste of precious resources and provide little to no advantage if the technology is used in the wrong circumstances. AI should be implemented in situations where intense human involvement is a drain on the organisation, or where the consumption and analysis of large volumes of data can provide competitive advantage. Essentially, it should only be used if it will deliver a better business outcome.
Ask the question, “Why are you using AI?” Any software vendor should be able to outline the advantage that AI will bring over an alternative solution and clearly state how the business outcomes will improve as a result.
Where do you use AI and where is it hosted?
Understanding where in the process or business solution AI is used can be very telling and give you a feel for how well your vendor has grasped the requirements and expected business results. Does the model align with your understanding of the problem? Will it genuinely alleviate delays and errors and improve your productivity? It’s important to ensure that your hands-on business teams validate the Where as this will help guide both your project your results.
The second Where question focuses on another important question: where does your AI live? The rise of the cloud has made AI infrastructure a no-brainer, with cloud companies such as AWS and Microsoft offering application developers and data science teams access to unlimited horsepower and key components for rapid development and management. Is a cloud environment right for your organisation or do you need a containerised offering that can run on-premise?
How do you use AI?
The final question, How, ties all the previous questions and answers together. You don’t need a sledgehammer to crack an egg, and not all business problems require AI as a solution. Here at Ephesoft, we leverage AI in complex document tasks where it can replace laborious, time consuming and often inaccurate human tasks. For example, we use AI and supervised machine learning technology to identify and extract complex invoice table data. However, not every problem requires that type of solution or the overhead of AI processes. For example, if you need a system to classify high volumes of different document types, AI may be too slow. In this case, there may be non-AI processing techniques that are much faster. The How needs to make sense, and AI needs to be the right tool for the right job.
I hope that this article provides some insight into how to discern real AI from fluff and misconception, and that your next digital transformation project will achieve the business outcomes you need.