This is a guest blog post from Positronic, a Premier sponsor of the StampedeCon Artificial Intelligence Conference 2017 in St. Louis on October 17.

Gaining an Edge with a Cogent AI Strategy

The Oscar W. Larson Company is a third-generation family-owned business that provides maintenance and repair for point-of-sales systems and fuel dispensers at gas stations. Truck re-rolls are big problem for Larson. A truck re-roll is the term for when a truck must make multiple trips to a gas station because the technician didn’t have the right tools for the job. It’s same for me every time I tackle a simple home plumbing repair. I can’t make just one trip to the hardware store. Larson worked with external partners, an external consultant and Google, to build a machine learning system to manage the inventory that technicians take with them. With this new AI-powered system in place, truck re-rolls went down by 20 percent while simultaneously reducing the inventory levels on trucks by up to 35 percent.

Artificial Intelligence (AI) is fundamentally changing how businesses of all sizes operate across all sectors. As Andrew Ng famously said [1] at the Stanford MSx Future Forum, “Artificial Intelligence is the New Electricity”. Every business, no matter the size or industry, should have a strategy for embracing AI to help guide their participation in this revolution. Even small businesses with limited resources can operationalize AI-enhanced software. Developing that strategy requires participation from both technology and business leaders. Let’s get started.

Using AI to Enhance Your Existing Business

There’s no better way to get your company on board with implementing AI than to make a quick impact to the bottom line. Netflix did just that with their inaugural machine learning project, a recommendation engine. The entertainment giant has access to a broad set of data; i.e. what each person watches, when they watch it, where the title was placed on the screen. To generate suggestions that are on-target, Netflix applies that data to classification, regression, and clustering machine learning models to personalize the service and provide recommendations for new things to watch. According to the Netflix executives who authored an academic paper [2] describing the work, “the combined effect of personalization and recommendations save us more than $1B per year.”

The First Step in Evaluating the Potential of AI for your Business

Start with an analysis of your company’s existing value chain, the high-level model that describes the end-to-end processes that go into delivering products and services to your company’s customers. Identify the touch points within that value chain that have the highest costs or losses. What if you could programmatically identify, classify, and group into cluster the inputs and outputs to that touchpoint. Would that expose an opportunity to improve process? Look for teams that are bonused on key performance indicators (KPI) at that spot in the value chain. They will be your cohorts. It’s likely that one of them is already leveraging data analytics give them insight into how to impact those numbers.

Leveraging AI & Automation to Create New Products

Time and again, we’ve seen that companies very quickly embrace AI and automation after seeing the positive impact of a single AI project on an existing line of business. However, as their trusted artificial intelligence consultants, we are careful to give our clients this advice:

Improving or replacing elements of your existing product line is a safer path to near-term success.

Your company may adopt the identity of being a technology company and grow an appetite to create new value propositions. Ensure that these efforts align with existing value chains and business models. Don’t let your team throw away a business model that works to follow the latest startup fad like “Uber for X”. There’s plenty of opportunity to leverage AI with existing business models and existing value chains.

PayPal recently built a new fraud detection mechanism using deep-learning techniques to replace the existing linear models. “There’s a magnitude of difference,” Hui Wang, senior director of global risk and data sciences at PayPal, said [3]. If the traditional analytics software sees a pattern of the same account being accessed by five IP addresses in five days, it flags the activity as suspicious. PayPal’s deep-learning system can look at each more closely and see, for instance, that the user is a pilot buying gifts for his family while on the job. The new deep-learning techniques have cut the false-alarm rate in half, Wang said. PayPal provides a great example of embracing AI to augment or replace elements in an existing product line rather than creating an entirely new line of business.

 Institutionalize New Capabilities

Now that you’ve completed a project and proven the value of AI it’s important to ingrain this organizational change into culture so that it continues to drive the company forward and keep your company competitive into the future. Lasting changes only are anchored by demonstrating results; both short and long term, and having company leadership communicate those results loudly and often. Having company leaders that communicate the importance of AI, both internally and externally, is critical to institutionalizing the new capabilities.

Consider the email that Microsoft CEO Satya Nadella sent to his executive team. It starts, “I know that this is going to ruin a number of your weekends”. In it were links to several AI resources and the missive: “if you want to be an exec at this company – you need to be competent at AI”. Jeffrey Snover, a Microsoft Technical Fellow, shared this with me over a recent lunch after I demanded more of Azure ML and interrogated him about Microsoft’s ML strategy, “How serious is Microsoft taking AI at the executive level?”. Satya’s message is clear; the winners tomorrow will have a strategy for AI today.

Listen to the way that Jeff Bezos talked about AI in his most recent letter to shareholders [4]: “big trends are not that hard to spot (they get talked and written about a lot), but they can be strangely hard for large organizations to embrace. We’re in the middle of an obvious one right now: machine learning and artificial intelligence. Over the past decades, computers have broadly automated tasks that programmers could describe with clear rules and algorithms. Modern machine learning techniques now allow us to do the same for tasks where describing the precise rules is much harder.” Bezos understands the impact that AI is having to his industry. If yours doesn’t, it’s time for a lunch and learn.

The biggest challenge you’ll face institutionalizing AI is that the demand for data science talent is significantly higher than the supply. Large companies are aggressively hiring and acquiring machine learning talent to build internal capabilities. According to a recent McKinsey report [5], 80-90% of all today’s AI talent are working at the largest technology companies in the world. In the current market medium-sized businesses are finding more success partnering with AI technology firms than trying to recruit and retain a force of their own. Small businesses with limited budgets will find their best ROI buying off-the-shelf software enhanced by AI.

 Get Started with Business AI

The best source of inspiration are success stories. Seek out and read as many case studies as you can get your hands on. Attend meetups to connect with peers to hear their stories. The more success stories you consume, the more you training data you provide to the neural network in your brain to help classify and predict the sorts of problems that make good candidates for AI. Once you’ve trained your internal neural network you’re ready to apply it to your own business.

At Positronic we take a four-stage approach to implementing a new data science project.

First, we go through a DISCOVER process where we visually navigate the available data looking for patterns and correlations and we apply advanced analytics and visualizations as a guide to building hypotheses for what sorts of predictive models may fall out of the data. Second, we test chosen hypothesis through EXPERIMENT. An experiment is a unitary data research project over a big data collection. Often data cleanup and augmentation is required to piece together a viable dataset for learning. The experiment process explores all relevant machine learning models to generate mathematical proof of efficacy. Third, we choose a model with proven efficacy and refine it with further training until the model achieves the desired level of accuracy and performance on new unseen data. The product of this phase is an api that can PREDICT. Finally, our cloud apps team partners with your engineering and IT teams to DEPLOY the new interfaces into an existing line of business app or a new cloud app.

Looking for technical experts to serve as your AI sherpas as you guide your organization down this path? Contact our team to learn about process – and to start with an audit of your current data and automation potential. Positronic.AI



[2] https:/