If you haven’t implemented an artificial intelligence (AI) solution into your business yet, you may feel like you’re missing the boat. And in many ways, I’d agree with you. But is your business ready for artificial intelligence?
Some studies show that nearly 99% of companies are investing in AI in some way, shape or form. AI isn’t a “will we, won’t we” type of technology. AI will be the de-facto standard, much like an operating system or software, it will be embedded into every business technology in the not so distant future.
But that doesn’t mean you should just jump on the bandwagon for fear of falling behind. There are a lot of considerations to take into account before even dipping your toes in the AI water — or to carry through on my first analogy, to ensure you aren’t putting the cart (or wagon) before the horse.
Proper Planning of AI Implementation.
AI projects fail because of backlash due to a lack of proper planning and scoping. To ensure a successful artificial intelligence initiative, businesses need thoughtful preparation.
Take into consideration things like ensuring that AI doesn’t exist in isolation but is integrated into broader business processes are key to success.
What Questions Should You be Asking?
Plus, before rolling out any AI initiative, you need to ask a number of important questions.
Questions like what is the business opportunity? And do you have the resources you need to implement process transformation? Are there security implications?
What data do you need to solve the problem and what will you need to acquire it?
And maybe most important, are there any ethical implications for implementing an AI solution?
To help you get clear on these questions and more, here are a few things you must consider before seeking out an AI solution or hiring a team of machine learning engineers to build something in-house.
Understand what artificial intelligence is good at, and what it isn’t.
The question may seem trivial, but a lot of organizations we talk to don’t understand what problems are good and not good machine learning problems. Artificial intelligence is not a solve-all so make sure the problem you’re trying to find a solution for is appropriate.
Some common tasks AI is great for includes forecasting, anomaly detection, object detection, pattern detection, auto-generation, enhancement and reconstruction.
Have a well-defined problem
You need to consider what is the problem and why you are trying to solve it. If the scope is too broad, your initiative will quickly fail. For example, pathology of a whole-body offers too many variables but focusing within one body part is much better and will warrant better results.
Keep your scope narrow and build from there.
Identify the performance criteria for AI
Like any well-defined business initiative, before you begin, you need to identify what success looks like. Are you hoping to achieve greater accuracy than a human could achieve? Are you hoping to simply automate a task to save time?
Good performance criteria for an AI initiative will define performance on a narrow criterion with a given percent accuracy rate.
Determine the team and technology capability
Does your organization have the technical ability to work with AI? Currently, there are 300,000 machine learning engineers available and several million open positions.
Machine learning experts can earn as much as football players. Working with AI often requires understanding arcane mathematical and computer science concepts that most software engineers simply don’t have.
Finally, do you have the right tools to create and support artificial intelligence and machine learning processes?
Understand the long-term impacts
As I mentioned, the challenge with bottom up projects is that they often fail because of a lack of political will in organizations.
AI is simply not understood by most people in the organization and even framing a business argument for deploying AI is not always clear.
Obviously, a clear understanding of ROI will help but even this isn’t enough because in the end, like any other technology deployment, the ROI has to be compared to other non-AI alternatives.
Lastly, it is likely that AI will displace individuals. In one of the companies I worked for, we developed an AI solution that resulted in a 60% reduction in engineering issues for a very expensive manufacturing process.
Obviously, this would have had a significant impact on the business but in the end, after two years, the solution still did not gain as much traction as we would have desired because it would have entailed the elimination of an entire team.
Training data for machine learning
Do you have the data you need to effectively train a model? Plus, is that data accessible?
Artificial intelligence governance
Developing AI is only part of the process. Can you deploy and support the AI in production, deprecate it, or determine if the AI is performing to specs? Do you have a mechanism to enable broad deployment and management or the people to perform the work needed?
Few organizations have a complete strategy for how the AI is to be used or managed by their business. For example, a simple question of whether to deploy the AI into the cloud, on-premise, or deploy to the edge is not always clear.
Finally, is your AI solution “future-proofed.” If changes in technology or capability occur – how easily can the organization adapt?
Once you’ve gone through these set of questions and considerations, you’ll be ready to take on an AI solution (AI Dynamics, Inc, Bellevue, WA) or kick off an AI initiative within your organization. And that’s when the fun really begins.
Image Credit: Michael Dziedzic; Unsplash