Get AI or Die Tryin’: Part 2
In Part One, we explored the possibility of the terminal fate of legacy businesses and jobs that are simply not moving fast enough (or in some cases, at all) to keep up with well-prepared data-driven competitors. Yet, with this disruption – or destruction – of some industry players and business models comes the bright side of a potential 4th Industrial Revolution. This revolution will result in the creation of amazing new products, jobs, and services once thought to exist only in science fiction movies: intelligent agriculture, personalized medicine, autonomous vehicles, decentralized power grids, space exploration, the list goes on.
Early adopters are moving fast to develop new products and AI-driven services in an effort to successfully compete against the looming tech giants who are well stocked with the resources and fuel (data) they need to prosper in new markets.
I will do my best to avoid the kind of hyperbolous claims that usually come out at this part of the conversation: that artificial intelligence (AI) and automation will either A) save the world or B) destroy it. Instead, I will try and go with something like C) AI and automation are decided competitive advantages that will have a significant impact on a majority of businesses and jobs, but only a small (but growing) minority are actually preparing for it.
However, it is that minority I will focus on in the conclusion of Part Two. It is the thousands of companies and their executives whom we’ve already seen begin to embrace a particular subset of AI called “automated machine learning” to try and “cheat death” and solidify, or even grow, their market share. These early adopters have picked up on the fact that Big Tech is starting to make their way into every industry vertical with a eye on legacy markets (banking, insurance, transportation & logistics, healthcare, etc.) and they can’t sit on their hands any longer. Early adopters are moving fast to develop new products and AI-driven services in an effort to successfully compete against the looming tech giants who are well stocked with the resources and fuel (data) they need to prosper in new markets.
I will highlight two key strategies we’ve seen these companies use to successfully manage the transition to becoming AI-driven enterprises, and identify some risks to look out for along the way.
The Industry Expansion and Cross-Pollination of Data-Driven Giants
Over the last twenty years, we have seen a handful of data-driven technology companies grow up to dominate their respective markets like search, media, mobile, and commerce. Now, as if they were each a single crop grown on the same patch of soil for many years, the growth of each incumbent’s ‘yield’ is starting to slow. As a result, these firms are all looking for new soil to till – business models and verticals to enter – and they’re bringing an army of data scientists and data engineers with them. This was the best analogy I could think of for the increasingly common trend of these industry giants cross-pollinating, both with each other and on their own, to enter new markets and continue to use AI and automation as their competitive advantage.
For starters, just this year alone we saw announcements of Amazon getting into delivery, grocery, insurance, and healthcare. Rumors had even floated of an Amazon-powered consumer checking account product with JP Morgan, after already being in the small business lending game, putting large parts of an entire domestic retail and commercial banking industry on watch. Tesla, Google, and Uber are now all gunning to lead the automated transportation market. Uber Freight and previously mentioned Amazon both have aspirations to upend the global shipping and logistics market. AirBnb and WeWork partnered to pilot a new bed-and-work program. Apple has ambitions to disrupt healthcare; so does Google through its Verily arm, and now even Uber is getting in on the action. Softbank is looking to take a rumored $10 billion stake in SwissRe while Amazon / JP Morgan / Berkshire Hathaway have teamed up to build a data-driven insurance company of their own.
But, I believe there is still time for nearly every organization and their leadership teams to realign their corporate priorities and data strategies to ensure their survival in the coming age of artificial intelligence (AI) and automation.
If this accelerating trend of monopolization and cross-pollination of industry isn’t a giant wake up call to every board of directors, shareholder, executive, and even employee (you still have time!) at every legacy insurance, retail, banking, real estate, etc. organization, then I don’t know what is. We will eventually find them somewhere left behind, stuck within the amber of old industry alongside Kodak and Sears.
It’s Not Too Late
As an executive on the board of directors of a Fortune 100 company recently told me after one of our presentations, “If you were thinking about AI in 2015, it felt like you were too early. Now it feels like if you’re just starting to think about AI in 2018, you’re too late.” That might be true, to an extent. When you see how fast the market is moving, it’s hard not to feel like you’ve been ‘left behind.’ But, I believe there is still time for nearly every organization and their leadership teams to realign their corporate priorities and data strategies to ensure their survival in the coming age of artificial intelligence (AI) and automation.
It’s worth repeating: we have seen $10m to $100m+ returns on single use cases deployed in an organization thanks to automation and accuracy gains with AI applications.
It’s easier said than done, but from working with cutting-edge enterprises and public agencies across a wide-range of industries, here are two key trends related to automated machine learning that we’ve seen at DataRobot and have been correlated with organizations that are successfully transitioning towards becoming ‘AI-driven’ and remaining competitive against the ‘data giants’ eager to steal market share from new and existing industries:
AI starts (and ends) at the top – There is simply not enough time to allow AI and machine learning (ML) to remain as a ‘pet project’ within an organization. Those days are long gone. There are enough concrete examples of low-hanging fruit use cases across all industries to know there is significant ROI to be gained, if you know where to find it. That means executive teams should first and foremost prioritize their own education on AI/machine learning applications within their respective industry. Then, more importantly, they must work on educating their employees and training them to develop the skills necessary to be AI operators and architects of their own.
A ‘bottom up’ adoption strategy might seem nice as the BoD and C-Suite can stay out of the weeds and let things develop organically within their organization. This might have worked well for Business Intelligence (BI) and Big Data, but AI and process automation is simply too radical and disruptive to individual and inter-team workflows to gain the kind of widespread buy-in and garner enough resources to even get projects off the ground, let alone drive the enterprise-wide adoption that is needed to remain competitive with the data giants. At DataRobot, we’ve developed executive training curriculum called Machine Learning and AI for Executives to help fill this gap.
Prioritize winners and then invest in success – With so much hype around AI and machine learning, internal expectations for success and ROI can range from SUPER LOW to SUPER HIGH depending on who you ask and when you ask them. The problem is that both sides have a chance of being right initially depending on the early results and time to value. This is why it is extremely important to prioritize ‘low-hanging fruit’ use cases for AI and automation vs. trying to tackle more complex use cases that might appear to be ‘sexy’ in a board deck but will almost certainly fall well short of expectations when reality sinks in.
This means you might initially focus on building a lead scoring, churn, or consumer propensity model instead of that deep learning image recognition recommendation system (“But it worked for Netflix!”). Avoid the trap. By achieving quick and tangible ROI with data science and automation, organizations are able to gain momentum and confidence in both the C-Suite (“This stuff works!”) as well as with end-users (“What I’m building is actually being used in the organization and making us money!”) to align an organization and set them on the right direction towards a more data-driven culture, top to bottom.
The good news is that by starting with #1 (the right executive sponsorship), we have seen many organizations go from a very nascent or even non-existent AI-driven corporate strategies to fully-deployed machine learning applications in a matter of just a few months, some of which have generated eight and even nine figure returns within a single year (for a single organization).
It’s worth repeating: we have seen $10m to $100m+ returns on single use cases deployed in an organization thanks to automation and accuracy gains with AI applications. Executing on AI strategy and monetizing data is not limited to the top 10 tech companies. And, do you want to know the kicker? Thanks to advancements in automation within machine learning, many of these AI apps, including some of the most lucrative ones that we have seen, were designed by individuals with no prior data science experience.
With the right combination of tools, training, and support from their leaders, these ‘lower-skilled’ data workers (when compared to classically trained data scientists) were able to leverage individual creativity and domain experience to design automated algorithmic processes that generated significant, tangible value for the enterprise.
As with many new technologies, this does not come without disruption or resistance by some data scientists, who are unable to comprehend how this kind of automation is impacting their (previously very exclusive) world and bringing significant value to their organization. This is definitely a risk to watch out for. The dismissive tone expressed by these individuals bodes a striking resemblance to the ‘higher-skilled’ artisans resisting automation and machinery back in the 19th century, as laid out in the 2016 research report by Obama Administration on AI & Automation previously cited in Part One:
For example, the 19th century was characterized by technological change that raised the productivity of lower-skilled workers relative to that of higher-skilled workers. Highly-skilled artisans who controlled and executed full production processes saw their livelihoods threatened by the rise of mass production technologies. Ultimately, many skilled crafts were replaced by the combination of machines and lower-skilled labor. Output per hour rose while inequality declined, driving up average living standards, but the labor of some high-skill workers was no longer as valuable in the market.
With the right combination of tools, training, and support from their leaders, these ‘lower-skilled’ data workers (when compared to classically trained data scientists) were able to leverage individual creativity and domain experience to design automated algorithmic processes that generated significant, tangible value for the enterprise. While that might be bad news for the armies of expensive data scientists employed by the technology giants, it’s great news for the rest of these early-adopters who are embracing automation and best-in-class technology to maintain or even grow their market share against the looming giants. They know that their Life’s on the Line*.
*Life’s on the Line is the last song on 50 Cent’s debut album Get Rich or Die Tryin’.
About the Author:
Ben Solari is the Director of Inside Sales at DataRobot, where he now leads a team of 15 Corporate Account Executives spanning from Boston to Singapore. He started as the first Inside Sales hire at DataRobot in March 2016. His background is primarily in finance and analytics, with previous data and analysis roles at organizations like ESPN, UBS, Trillium Trading, and InsightSquared.