September 23, 2022
5 Steps to Analytics and AI Success
Before businesses can design and deploy an effective artificial intelligence solution, they must first identify the data that will make it work.
We see many companies today rushing headlong into the world of artificial intelligence and data analytics. But too often, business and IT leaders start with an imagined solution, without first considering the data and processes that will support it.
Often, the data that will be most valuable for analytics and AI efforts is siloed, locked away in specific systems (or even in spreadsheets). When building an AI practice, leaders should work their way through the following steps with a trusted partner to achieve optimal results.
Begin by Identifying Business Challenges That Need Solutions
We’re still beginning with the end goal in mind here: a business problem that the organization wants to solve. However, rather than rushing to adopt an off-the-shelf AI product, it’s important to spend some time early in the process assembling business stakeholders and outlining the project’s mission, vision and scope.
We recently worked with a silicon manufacturer that wanted to improve speed and accuracy in identifying defects. The company’s existing processes took five and a half minutes, and resulted in 67 percent accuracy. Those baseline data points were invaluable in helping to later measure the success of the company’s AI efforts.
Partner with Stakeholders to Discover AI Needs
At this stage of the process, we work with business stakeholders to identify the data that will help the organization meet its AI goals. We also help to identify the skills, policies, procedures and cultural assets that will best support the AI effort.
This step is important not only to assist with the overall design of an AI solution, but also to ensure that the solution is eventually implemented effectively. We’ve all seen tech initiatives that have flopped simply because users did not adopt them, and this is especially common for emerging technologies that internal users don’t fully understand.
Dig Deeper to Fine-Tune Solutions
This is where most organizations (erroneously) start their process. But by engaging in the previous two steps, we can ensure that the design and modeling phase is much more productive.
During this step, we provide a second set of hands, recommend industry-specific use cases and software development kits, provide ongoing consultation and make recommendations for automated machine learning. There are often a number of spots where the rubber meets the road at this stage, with questions that stakeholders haven’t yet considered. For instance: Where will the AI solution reside? In an app? On an endpoint? And how will it be disseminated to users?
Choose the Right Hardware for a Solid Foundation
Hardware can be an important — and overlooked — consideration in the design of new AI systems. Remember the silicon manufacturer we mentioned earlier?
We discovered that the company could dramatically lower latency (and, as a result, improve the speed of their defect detection process) by utilizing GPUs rather than CPUs. This was simply a blind spot on the part of the organization, and one that they never would have uncovered without another set of eyes.
Reap the Payoff with an AI Implementation and Deployment
At this final stage of the process, we assist with all aspects of AI and machine learning operations, including product management, user enablement, feedback mechanisms, product rollout and scale-up. After working carefully through each step, organizations will often achieve results that would have been unthinkable at the beginning of the process.
The silicon manufacturer we worked with improved the accuracy of its defect detection from 67 percent to 93 percent and, incredibly, shrunk the length of the process from five and a half minutes to ten seconds. If you’d like to achieve similar results with your next AI initiative, it’s worth taking the time to work through these steps and do the process right.
Jeff Myers, a CDW Principal Field Solution Architect. He is a hands-on technology strategist with 25-plus years of experience as an enterprise solution architect with artificial intelligence and machine learning, enterprise platform integration, semantic web, database design and directing large scale technology teams. Currently, he is enabling clients with the power of AI/ML and deep learning on a hybrid infrastructure.