From manufacturing to banking, and everything in between, artificial intelligence is transforming businesses and industries. The potential gains are exciting, but implementing AI solutions can be daunting for many companies. It’s hard to know where to start.
We’ve put together a 5-step process to help businesses identify a suitable use case and understand why making the right choice is important.
Step 1: Identify a business problem
Any major undertaking, whether it involves AI or not, should address a specific business problem. After all, there’s no sense in changing things that work well and still meet your needs.
Ask your business units what’s affecting the company’s competitiveness and bottom line. From maintenance inefficiencies to distribution bottlenecks, most businesses can come up with a long list of issues they’d like to resolve. At this stage, it’s useful to jot them all down.
Step 2: Assess your existing data
Artificial intelligence only works if you have large amounts of information—i.e., big data—to work with. Specifically, you need historical data about your operations (e.g., equipment, order volumes, customer needs, etc.) as well as complementary data that could provide insights into trends (e.g., weather, natural disasters, political events, etc.).
Looking at your list of potential use cases, shortlist those for which you already have extensive data. Remember, the more information you have, the more value you’ll be able to derive from your AI system. This is an important criteria to keep in mind, since it can be tempting to tackle your most pressing business threat, even if it isn’t backed by years of data. That would be a costly mistake on many levels.
There’s no way around it: you need data for your first AI use case. Remove any business problems from your list that aren’t supported by accessible data. It’s important to understand that most of time associated with an AI project will be spent around cleansing the data and getting it into a digestible format. An AI project will be that “light bulb” moment on understanding the importance of a good data strategy on which businesses will be building their future. So, start now.
Step 3: Choose a problem that can be scaled down
Now you should only be left with a handful of viable AI projects. Again, resist the temptation to put your biggest business concern at the top of the list. A high-stakes issue should not be considered for a first AI use case.
Since this is your company’s first AI project, select a use case that can be broken down into smaller pieces. For example, if you want to boost sales, start with a single product line or customer. Even though increasing volumes with a single customer won’t do much for your bottom line, working toward this goal can help you build AI models that you’ll be able to apply to other customers later.
Step 4: Set a goal that’s achievable yet meaningful
How will you measure success? This step involves defining the minimum improvement you want to see within a specific time frame. Your goal should be realistic, measurable and worthwhile in terms of ROI. While everyone hopes for a 50% increase in sales, most businesses would agree that a 5% or 10% gain is substantial. Find the line between nominally better and significantly better, and make that your goal.
This step needs to be supported with a plan for tracking key performance indicators. The plan should clearly outline what you’ll measure and how, as well as who will compile the information and how often they’ll report back to your management team. Typically, a project that’s moving in the right direction should start showing value within 3 months.
Step 5: Build your MVP and fail fast
Once you’ve clearly defined your goal and established the minimum criteria for success, it’s time to start work on your minimum viable product (MVP) and brace yourself for failure. Failure?
Since each AI model is highly customized and specific to your business conditions, the development process necessarily involves a series of trials and errors. The key, however, is to quickly figure out which avenues are dead-ends. That’s where the widely used expression “fail fast” comes from. But what it really means is that you’ll learn fast and achieve your goal sooner.
Whenever we meet with companies that want to extract insights from their data, there are a few things we want them to understand. First, AI projects are only possible if you have data that machines can learn from. If you don’t have enough data, it’s time to start collecting it. Also, despite the revolutionary potential of AI, each initiative starts small and builds incrementally on its successes. As a result, it’s important to manage expectations within your organization. And finally, few businesses have the in-house expertise needed to implement AI solutions efficiently. Given how quickly the field of machine learning is evolving, it makes sense to partner with firms that specialize in artificial intelligence and related technologies.
Vice-Presidente , Innovation
R2i is a Montreal-based firm that specializes in IT hardware and cloud solutions. We provide technical and strategic support to companies of all sizes starting on the journey to artificial intelligence. If you’d like to know more about how to put AI to work for you, contact Tamara Vandersluis, Vice-President, Innovation, at 514-312-3007, firstname.lastname@example.org.