Presented by Ople
AI has the thrilling ability to transform a range of businesses. But let’s be frank: it’s also a beautiful, massive disappointment for many companies.
Here’s a common trajectory for many AI and data science projects in an enterprise: A company decides to incorporate AI into their business. They spend one to two years searching for AI experts to build a team of solid data scientists, but not necessarily industry experts. The team works for a year or so on a project, only for the company to discover that the project is irrelevant and they need very different people. So they restructure the team, winding up back at square one, four years later.
If you’ve been in this situation, you know how hard it is to find data scientists who can do all the tasks required, which range from soft communication skills to hard statistics. In fact, many companies look for data scientists who can comprehend and engineer the data, build and tweak models, communicate with the necessary teams to understand the business, and deliver applicable solutions. This isn’t a run-of-the-mill candidate search. This is a grand quest for AI unicorns.
The other path companies take is to buy a vertical-specific AI solution from a third-party vendor. The software platform looks sleek and pretty and promises to deliver exactly the kind of predictive power the company needs. But looks can be deceiving, and few products have proven their worth at this point. While vendors claim to be vertical experts, these one-size-fits-all types of products often require significant investment to match the companies’ needs. Many of them do not deliver at all. We are simply too early in the AI transformation.
So if building a team is futile and buying your AI is a real gamble, what’s a company to do? Both, but in the right balance. You can build an in-house team based on the right third-party AI platform. One that will empower existing employee experts or incoming hires to maximize their efficiency and actually solve what matters to your business.
Companies who ask the right questions to enable this hybrid approach save themselves grief and resources. They are also more likely to end up with a viable, business-impacting AI solution.
What data and resources do you already have?
Lots of companies are already formulating problems that can utilize machine learning. Actuaries, sales teams, and a variety of professionals in diverse industries are predicting things and observing patterns via data, addressing issues like risk, resource allocation, or price optimization. They have been doing this since they started, sometimes for centuries, using other techniques. In other words, the knowledge is already there, whether they have AI or not.
If you’re already asking this kind of question, you can better see where AI fits in and what it should deliver for the investment. And If you’re innovative, you can start asking questions in your industry that no one’s asked before, simply because the data analysis techniques were too cumbersome, and create more impact on your business.
What problems will have the biggest impact on your business that also have elements of predictability?
You know your business better than anyone else, but you also know you’re not a data scientist. So how do you formulate your problem for a potential AI project? In machine learning, it’s less about what you ask as about how you ask it. The question should be answerable by someone with infinite time and copious data, and a clear idea of what to predict. By applying machine learning to this type of question, you will get more accurate predictions in less time than before.
Let’s assume you are in ecommerce, and you’re interested in predicting the likelihood of a customer checking out her shopping cart. If you are making the prediction based on factors like the previous purchase history and time spent on site, you will get a reasonable outcome. But if you are looking at factors like the number of red cars on the street at the time or the number of windows in your office, you will never get a reliable answer.
This example is somewhat extreme, but the point is that AI is not magic. Moreover, it requires your expertise because the answers to what matters are often industry- and even company-specific. In other words, you need to rally your teams and figure out the questions and answers you seek before embarking on any AI venture.
What answers and level of accuracy will actually move the needle?
You’ve formulated a problem, you have the data, you have your approach. Now try it out and see how it impacts the business. AI models should not exist simply because they are cool. They need to move the needle for the company.
In many cases, moving the needle means testing a hypothesis. What good is a model with 99 percent accuracy if the prediction doesn’t matter? The biggest project killer does not come from teams failing to hit a certain accuracy. It’s from failing to define the right problem and not knowing how better predictions will impact the company’s business. If it’s a data science success, that’s lovely, but it has to matter to your bottom line, not to scientific research circles. The goal is to change your business. AI can do that, but only if the right tools are given to a team with clear objectives. That’s what a “both” approach can do.
At Ople, we use AI to build AI. We have developed an artificial intelligence platform that acts, thinks, and learns like a data scientist, providing our customers elite-quality deep-learning models deployed instantly and ready to make predictions in minutes, not months. If you would like to learn more about what you can achieve with Ople, please visit our website to download our one-pager and set up time to meet.
Sponsored posts are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact firstname.lastname@example.org.