Organizations require right technology solutions to pivot organizations
As recently as 6 months ago, AI spending was projected to exceed $50B worldwide by 2021. Meanwhile, an August 2019 HBR article reported on the survey responses from thousands of executives regarding AI/Advanced analytic usage at their firms. Only 8% said their firms engaged in core practices that support widespread adoption of AI/Advanced analytics.
A few months later the Coronavirus crisis struck across the world and may have created a whole new reason for organizing and employing AI/ML. Organizations large and small are seeking solutions to understand where the new normal may be going when we emerge from the Coronavirus crisis.
I recently spoke with Jamal Robinson who has published multiple articles on artificial intelligence, has worked at Intel as Director of Cloud & AI, is currently at Amazon Web Services as a Global Lead for Machine Learning and was a professor teaching data science at the University of California, Berkeley.
Gary Drenik: Why should an organization develop a strategic plan for Artificial Intelligence & Machine Learning (AI/ML)?
Jamal Robinson: Great question Gary, the primary answer from my experience is AI/ML represents a unique ability for startups, SMBs and large enterprises to accelerate digital transformation across their key business initiatives. This is not limited to a certain industry as AI/ML has positively impacted many industries like healthcare, financial services, retail, robotics, oil and gas, automotive, manufacturing and as you know, forecasting. We are currently seeing many industries gravitate to AI/ML as the potential to realize meaningful change and innovation is strong regardless of what your business’ focus is. This is why research firms like IDC have predicted AI and cognitive technology spending will exceed $50B worldwide by 2021 and Gartner’s 2019 CIO survey discovered Enterprise use of AI has grown 270% over the past 4 years.
Drenik: Well that all sounds great, but what are some of the challenges leaders and organizations face when implementing an AI/ML strategy?
Robinson: When answering this common question, I like to break down my response into two areas: challenges implementing AI/ML technologies and challenges with adoption across the organization. At the organization level, executives need to democratize AI/ML which requires a good understanding of how to structure their organization to accommodate successful implementation of the technology. Rapid adoption of new technology, regardless of whether it is AI/ML, quantum computing or migration to a new CRM system, is not an easy process. At this point, executives who have started their digital transformation journey with AI/ML also know that it isn’t easy to find and retain top talent. This is an emerging area where expertise is limited and requires organizations to compete for scarce, top talent which is something they may not have had to do before. For the technology-related challenges, while there has been a push by many software and cloud vendors to democratize AI/ML and make it approachable to all, there are a lot of options available. Unlike selecting a database vendor, many of the players are new to this space without an extended history of producing and supporting enterprise grade software for ten or more years. This can make vendor selection and implementation decisions, like self-hosted or managed services, more difficult for enterprises. Executives also need to forecast their journey with AI/ML to make sure they are adopting a platform that is simple enough to realize value from immediately, yet mature enough to grow with the enterprise as their needs for AI/ML become more complex.
Drenik: How do you determine the goals for the initiative?
Robinson: For executives, I think this is similar to any other digital transformation initiative. To quote Covey’s 7 Habits of Highly Effective People, they should “start with the end in mind.” This means to clearly identify the reason(s) they’ve undertaken this effort which could be to make a business process more efficient, create new product functionality and expand their addressable market, improve customer satisfaction, increase employee performance and satisfaction or all of the above. At that point, which hopefully happens before the business undertakes their AI/ML initiative, they’ll want to create metrics to quantify how well they’ve achieved their desired goals. This can range from the simple, like putting out a satisfaction survey on a new process or feature, to the more advanced, like use of advanced AI/ML backed analytics to quantify employee or customer behavior and responses to the implemented changes.
Drenik: What types of data should be considered and where can they be found?
Robinson: This question primarily depends on the what the initiative is. Companies, whether they realize it or not, have a wealth of data choices available to them. Internally they can audit business operations to discover existing systems with relevant data that help them gain business insights or add metrics to those systems to start generation of data that helps the company make data-backed decisions with AI/ML. Externally there are open-source data sets provided that are popular with data science teams. Some people think open-source is free and you get what you pay for, but these open-source data sets when used correctly have fueled technologies that have touched millions and generated billions. There are also enterprise-grade curated marketplaces and vendors for data which may be more apt to enterprises with specific use cases or data requirements. A benefit of the curated enterprise-grade dataset is they are offered by someone with an understanding of enterprise needs. Prosper’s datasets are a good example of this as they cover consumer behaviors across retail, financial services, auto, mobile, dining out, digital commerce, health care and media while also keeping user data anonymous and are GDPR, HIPAA and PII compliant.
The last point I’ll make is consideration of the “right data” may not always be immediately clear. For example, an enterprise whose product is food and beverage, may be able to improve prediction of customer likeliness to consume a product if they also factor in weather data as better weather conditions can correlate to higher consumption of certain products. Same is true with financial services firms looking to assess risk on car loans, data on whether a potential customer has a gym membership and how often they wash their car can give better predictive insight on how the customer may value their car and likeliness of making consistent loan payments. These are two examples of the data sets that may not immediately jump out to executives thinking through their data strategy. My recommendation is to look at the selection of data as an art and not a science, be creative in what data you pull together when trying to find correlations to make predictions and reexamine the data in use often.
Drenik: After an executive has taken their organization through this process, can you provide examples of what success will look like for them?
Robinson: Thanks for asking, as I find this is another area organizations need clarity on. I’m going to do what I feel like is going against the grain a bit here and tell your readers to not be paralyzed to inaction by looking for the home run use case for AI/ML. Marketing for AI/ML has gotten a little out of hand and execs are thinking anything short of 10-20x gains aren’t worth considering. Uber, known for thought leadership across autonomous driving, provided a great example of how modest gains from AI/ML digital transformation can be impactful and did so in an area applicable to almost every business regardless of industry, customer support. Uber’s customer support team responds to thousands of tickets from the 15 million trips that happen every day. They used AI/ML to speed up the processing and resolution of support tickets. The first roll out of AI/ML for customer support sped up ticket handling time by 10 percent with similar to better customer satisfaction ratings. The second version drove an additional 6 percent speedup in ticket resolution, which is 12.7% cumulatively. I know speeding up ticket resolution by 12.7% may not seem very sexy when compared to advertised promises of 800% gains or more with AI/ML. But consider this, for an enterprise with call centers processing 2,000 tickets daily, with an average resolution time of 20 mins and average employee salary of $50,000, this new efficiency means less staff is needed to resolve the same amount of tickets. In this scenario, the call center saves an OPEX of $3.2M annually in average salary across 65 employees while delivering the same levels of ticket resolution. To me this is the power of AI/ML in digital transformation, approachable and reasonable gains that significantly impact the business positively.
Drenik: Great insight Jamal. Prosper’s April survey uncovered a dramatically higher level of concern across all market segments, with roughly 50% saying their purchasing behaviors will change in the future as consumers are reordering their priorities. We also uncovered the majority expect recovery to take up to 18 months. For the last question, I’d love to get your perspective on the impact of Coronavirus and how it may affect AL/ML initiatives.
Robinson: Yes, the elephant in the room. Or I should say the elephant on the video conference as we are practicing safe social distancing. I think people can look at this a few ways. One perspective is COVID-19 has disrupted the economy in a way that has caused executives to de-invest in any new initiatives, including AI/ML, and this will be the case for the foreseeable future until society and their balance sheets return to normal. You can also look at COVID-19 as having one of the most significant impacts we’ve seen on digital transformation across startups, Fortune 500 and Global 2000 enterprises. A simple example of this is the ability for employees to work from home. Before, many companies thought this wasn’t possible but in coordination with government recommended non-essential work shutdown, companies rolled out comprehensive work from home strategies almost overnight. Note that this is not limited to private enterprises as public sectors like education have also rapidly adopted digital solutions more broadly which has enabled teachers to educate students remotely. I’m aligned with the latter perspective and think COVID-19 has shown us the power of IT and how having the right technology solutions available allows all enterprises to more easily pivot their business operations with market changes, even changes as unexpected and disruptive as a global pandemic. In March, IDC conducted a survey in China with CXOs across 10 industries which found corporations noted COVID-19 had a positive effect on remote collaboration, their company’s ability with digital marketing and business development and with creating company-wide recognition of the value of technology and digital transformation. I believe now more than ever, leaders can leverage the created momentum to delivery digital transformations, across AI/ML and other technologies, that allow leaders to gain better customer insights, deliver better customer experiences while also empowering their workforce.
Drenik: Well said. Jamal, thanks for your insights on AI/ML and especially for the importance during this time of disruption. In addition, Prosper Insights & Analytics has teamed up with AWS Data Exchange to provide free access to our Coronavirus data. To learn more, click here: Strategic Insights: Coronavirus Covid-19 Consumer