Is My Organization Ready for AI?

AI could help move your business forward in ways you haven’t even thought of yet. The good news is that taking that first step isn’t nearly as hard as you might imagine.

Artificial intelligence (AI) has been described as the “new electricity” and is the cornerstone of the Fourth Industrial Revolution. You’ve likely heard a lot about what AI can do, but despite the hype there remains a significant gap between the promise of AI and the reality of implementation for most businesses.

AI is being adopted faster than experts predicted and the success stories are becoming more numerous and diverse. According to data released by IBM in 2020, 45% of large companies and 29% of small and medium-sized businesses said they had adopted it. 

Even so, many businesses have no idea where to start with AI adoption. Many aren’t even convinced AI can help them at all! Change can be a scary thing, and jumping into this area can feel like jumping into the unknown. But we’ve also seen what AI solutions can accomplish, and those who do it well reap massive rewards. 

Is your organization ready for AI? Here are a few key questions that might help you sort it out.

Do you know what problems you want to solve?

OK, that sounds kind of trite, doesn’t it? Every company has problems it’s trying to solve. But it’s also something a lot of people don’t understand about AI. Unlike the academic, boundary-pushing approach to AI that often gets press, there’s another type of AI called applied AI. This involves using existing AI techniques, often, but now always, employing more classical (or shallow) technologies. 

In business, AI is not something to build for its own sake; it should be applied to something. When you are planning to implement AI, it should be to solve a problem within your company. And it doesn’t have to be a moonshot, either. Even simple increases in optimization or process automation can make a huge impact on your bottom line. This isn’t to say that the research happening in academia isn’t important. Rather, if you are asking the question of how to take the first step into AI, then the most successful application is unlikely to be found in bleeding edge research.

Some other common examples where applied AI can help include optimization, detection, prediction and process automation. The impacts depend on the problem area and size, and could include improving profitability, mitigating risk, driving revenue and finding new competitive advantages. Of course, where AI holds value for your company depends on your business - and your data.

Many organizations think the first step toward applying AI in their business is purely a technical endeavour. In reality, it’s based on what businesses know best: the problems and the processes they need to improve. If you forget all the buzzwords and hype, AI is simply about using data to improve (or reinvent) business processes. 

So, the first step of a company’s applied AI journey should be laying out clearly defined business problems.

Does your company have data? (Or, rather, how bad is our data?)

Data scientists and ML engineers can’t do a whole lot without, well, data. The larger and richer the datasets, the higher the quality and variety of signals that can be generated to make predictions and generate value.

Unfortunately, many organizations get hung up on this step and spend ages collecting more data, building out a huge data lake, and basically pushing any AI work down the road for quarters - or even years! In our experience, this is nearly always a mistake. The truth is, you’ll likely never get your data perfect. Read that part again; it’s important. 

In fact, it’s probably better to get those business problems together and bring the data you do have to a qualified data science team. The key is to have data that is aligned with the problems you want to solve. For example, one client that engaged our services was a chain retailer looking to optimize per-store performance. An analysis of their available data found that each store’s performance depended on a number of factors they weren’t even aware of. By using a clustering algorithm, our team was able to determine the positive and negative factors driving performance at individual stores. For the business, that represented an opportunity to take action to boost store performance.

Getting the right data is about knowing what to collect, but also how to collect that data. Here’s an example from a project we worked on involving some downhole data from oil wells. Our team was confused because they had really high-quality data for a subset of wells at one-second increments. Then, a few years back, the time increments increased to every few minutes. The longer delay in data meant that the AI model was basically useless on the longer time increments. When we asked their engineers what had changed, the response was that the files were getting too big to open in Excel. It was a manual configuration change done on purpose. If, in that instance, the company had spent a year building out the ultimate data lake, it wouldn’t have helped them in the slightest.

We’ve seen this over and over again. Having an ML engineer do exploratory data analysis should always help inform the business processes around collecting data. For this client, collecting more data wouldn’t necessarily be useful because they didn’t yet have a use case or understand what that data could tell them. By starting the conversation earlier between the data science and operations teams, we were able to level set on what might be possible. At that point, had they needed additional data, they would know what kinds of data would be helpful to collect.

Data scope or quality was identified by Gartner as a barrier to AI/ML adoption for 34% of surveyed organizations. The good news is that internal data can also be augmented with public datasets and other resources such as synthetic data. In other words, even if you don’t have all the data you need, there might be other ways to gather it outside of your organization. And, of course, you can start collecting specific data points directly relevant to what you are looking to accomplish.

As an applied AI company, AltaML brings business and technical AI experts to the table and works closely with companies’ domain experts to ensure that data and potential use cases are aligned. The goal? Developing AI that is laser-focused on solving key business problems. Maybe your data is fine as is or maybe you need to be collecting different data altogether. You won’t know until the business problem is defined and the data is analyzed in relation to that problem.

Is Experimentation in Your DNA? 

Many companies struggle with AI implementation because goals are set too high on some grand, all-encompassing application, locking teams into long-range project plans that might be built on flawed assumptions. Unfortunately, such a waterfall approach to applied AI is likely to fail given all the uncertainty at the outset. A better approach is iterative. We refer to it as making “little bets.”

So, what does that mean? In practice, it involves starting small, assessing where our work is gaining traction and growing from there. When we work with companies, we focus on narrowly defined business problems and then rapidly iterate through AI experimentation to cut weaker opportunities and succeed quickly on the stronger ones. We can then build on those quick wins to take on larger challenges and accelerate overall AI adoption. Accumulating wins with this approach can amount to significant changes in productivity, savings or profit margins, and helps build momentum around AI adoption within an organization. 

Does your company have the capacity and desire to grow into an AI journey?

Successful applied AI relies on bringing together AI experts and domain experts and taking an agile, little bets approach to AI experimentation and solution development. In other words, successfully leveraging AI within an organization is a journey - and it isn’t one that needs to be fully mapped out in advance. 

In fact, it shouldn’t be. 

In my experience, the businesses that succeed with AI start with business problems. They experiment and develop proof-of-concept models. Finally, they pilot and deploy potential solutions. As your business becomes increasingly data-driven, you become better and better equipped for your AI journey. Your first step sets the stage for the future - but you won’t know what’s possible until you take the first step. 

Applied AI implementation is an ongoing process, requiring frequent monitoring and adjustment. If you look at businesses with varying AI success rates, the most successful companies are those that adapt to early results. These technologies can lead to a considerable return on investment, but only if you can cultivate a culture of ongoing improvement.    

The Next Step?

AI presents a huge opportunity for forward-thinking companies to gain new advantages - especially if they adopt these solutions ahead of their competitors. We’ve seen so many businesses get stuck thinking about adopting AI, or kicking the opportunity down the road as they conduct further internal analyses. In reality, all you need to start your AI journey is some business problems, some data and a willingness to jump into the unknown. Like so many other parts of succeeding in business, AI is about experimentation. It’s about getting a set of results and iterating to get even better results the next time around. Above all, it’s about taking a risk on something new to see where it takes you. 

Is AI right for your company? An experienced data science team can definitely help sort that out and get you pointed in the right direction. Maybe, though, the first question you should be asking is whether you’re ready to make the leap.