A handful of AI products have burst on the scene over the past decade. From IBM Watson to Salesforce Einstein, some have heralded these tools as the arrival of foolproof, accessible AI for all.

But many companies still find themselves struggling to wrangle AI technology to fit their own purposes — and in some instances, decision-makers are surprised to learn they can’t access a ready-made solution for their companies’ unique needs.

For example, suppose a manufacturing company wants to use AI to understand when its custom-built machinery requires preventive maintenance. To make the initiative a reality, the company would potentially have to perform a multi-year process of gathering and cleaning data, and then using it to train an AI platform to function effectively as a custom tool.

Despite the media and vendor hype that AI has finally arrived, some companies find out the hard way that a one-size-fits-all, ready-to-go AI solution does not exist. Instead, companies must do the legwork to tailor AI tools to their requirements.

Navigating the early days of AI access

While experts have theorized about AI’s capabilities for decades, the business world has only recently acquired the computing power and data capabilities to take advantage of it on a wide scale. We’ve entered the age of AI, but we’re only in the early stages of discovering its potential.

Google Cloud’s VP of AI, Andrew Moore, put it well when calling out the hype : “AI is currently very, very stupid. It is really good at doing certain things which our brains can't handle, but it's not something we could press to do general-purpose reasoning involving things like analogies or creative thinking or jumping outside the box.”

AI has endless use cases, but AI utilization is still in its infancy, and the road to building a useful AI tool isn’t straightforward. In fact, half of AI projects fail at one in four companies, according to IDC. But if you have the necessary resources, the payoff could be worth the effort.

A formula for AI customization success

Across industries, organizations can start with a generic AI tool and form it into something niche and relevant for their own unique challenges. For example, Chase Bank could take Oracle’s AI platform and turn it into a credit card fraud prevention tool, while Nordstrom can take the same platform and develop a product recommendation engine for shoppers.

While the two enterprises may start with the same blank slate and tools, they can create solutions that deliver very different end results. But no matter what kind of solution you want to develop, you ultimately get there via the same process. Before you dive into an AI customization initiative, you absolutely need these three elements:

A clear problem to solve – It can be easy to get caught up in the AI craze, and some organizations may adopt the technology just for the sake of saying they’re powered by AI. But you can’t successfully train AI into a useful tool unless you have a specific and targeted problem to solve. Before investing in AI, make sure you have a legitimate objective nailed down, and that you have the necessary buy-in from others in your organization to move the initiative along.

Clean data – AI is only as good as the quantity and quality of your data. If you don’t have enough data or if that data isn’t scrubbed free of incorrect, incomplete or duplicated information, the outputs of your AI tool won’t offer any value. For many organizations, this is the most prohibitive part of taking advantage of AI. Unstructured data makes up 80% of businesses’ data, and it will take significant maintenance to make it usable for your AI.

The right talent – The tech skills gap is real. Emerging tech like AI and IoT has left companies with thin talent pools as professionals rush to become experts in these new technologies. Thanks to high demand and low supply, the price for high-quality, full-time tech expertise is unaffordable for some organizations. But if an AI initiative is right for your team, you can access expert talent through outsourcing and working with contract IT professionals or consulting firms that can help tackle the execution.

We haven’t yet reached the point where AI like IBM’s Watson or Google Cloud’s AI Hub is packaged and ready for use in any custom scenario. It’s true that AI can transform a business of any size — but only if you know what you’re in for and have the necessary resources at your disposal.

Kevin McMahon is Executive Director, Mobile & Emerging Technologies at SPR

Artificial intelligence