Talking with a friend and former colleague the other day, we wandered into a discussion of just how amazingly successful the IT industry is in selling a never-ending parade of Must-Have investment, without which – if you believe the script – your business is doomed to fail and will only be remembered as an example of a once fine company that faltered due to the limited imagination and lack of intelligence of the luddite management team.
Deep Learning, Neural Networks, Computer Vision and Natural Language Processing are just a few of the domains under the umbrella of today’s hottest area for IT – Artificial Intelligence. AI, if you believe the press, is a capability that forward-thinking companies are already rebuilding their strategies around and for those that don’t, swift start-ups who have embraced this disruptive asset will be stealing share from more established players.
For the record, there is some truth in the hyperbole. Companies like Alphabet, Facebook, Microsoft, and IBM have seen the writing on the wall and are dedicating much of their R&D budgets to advancing and embedding AI into all of their offerings. While start-ups with AI-driven business models attract strong interest from the VC community.
As a non-tech company, three years ago, answering the question of should we are investing in AI was fairly easy. It came down to the BIG 3 – BIG Data, BIG Problems, and BIG Budgets. If you had already made the investment in capturing detailed data that flowed through the business AND had a complex, well-identified problem that couldn’t be solved by conventional means AND had the resources to make sustained investments in highly skilled staff and specialized systems, then you were good to go. However if one or more of these elements were missing, you were wiser to hold off.
Today the landscape has shifted. Fueled by the sustained investment of the tech giants, we are experiencing a massive wave of AI democratization. Where once, terabytes of proprietary and processed data were required to train a neural net, we now have off-the-shelf software that is able to make high-quality predictions in novel domains when trained with a tiny fraction of the data historically required. Similar advancements in natural language processing have brought us to the point where customers transparently engaging with a machine is not just technically feasible but wholly satisfying and economically compelling. Betting on the potential returns from broad adoption, AI tool providers have priced their offering at levels that create attractive ROIs for even smaller tactical and operational problems. The impact of these converging factors has been a morphing of the question around AI from “Should we invest?” to “Which of the potential projects should be first?”
Maximizing risk-adjusted return is the prime directive with financial investments and not a bad frame for considering opportunities with AI. Simply put, where can we find big bang, with small bucks and a high likelihood of success? Judging from the successes and failures I see across industries, the AI projects with the most positive impact currently orbit a constellation of People, Patterns, and Probabilities.
Where the People are in an organization is a strong proxy for both costs and level of effort. It follows that if we are going to have an outsized impact from a project it almost certainly arises from improving the productivity or capacity of staff. So we should start our evaluation of by answering the question of “If successful, will this project fundamentally shift the productivity or capability of a sizeable group of employees?”
Having identified a target group, it is then helpful to focus on Patterns. The human brain is masterful at pattern matching. Even at a subconscious level we cannot help but spin a tale about the world around us and how it works based on tiny and often inconsistent snippets of incoming data. Because finding and acting on patterns is so fundamental to what we think of as intelligence, it shouldn’t be a surprise that much of the AI research over the past several decades has been directed to recreating this ability in silicon. Even the most advanced research centers still have a long ways to go before replicating the capabilities of a human brain, but what has been achieved is the creation of very impressive pattern matching tools that are available at an extremely reasonable cost. Perhaps even more exciting is that the pattern matching in silicon works in subtly different ways than it does in our own brains and because of that divergence it turns out that it is possible to identify and act on patterns which previously were opaque.
The final factor to look for in an attractive AI project is that of Probabilities. As digital engines, computers have historically excelled in the domain of the discrete and we have looked to them to provide concrete answers from factual inputs. The nature of intelligence, however, rests on a foundation of uncertainty, both in inputs and outputs. As humans, we believe or disbelieve things with varying levels of intensity and the vast majority of our decisions and actions are based on our perceptions of the probabilities related to those beliefs. On this dimension, machine intelligence and human intelligence are rather similar. In their current form, AI systems tend to be relatively poor performers at making deterministic, 100% certainty decisions. Instead, they excel at estimating the probability that particular alternatives are likely. While it is admittedly a subtle distinction, AI systems excel at identifying the great answer rather than correct answer and the most successful projects embrace this distinction.
Circling back around to the recent conversation with my friend, we can’t help but acknowledge that a fair amount of marketing for technology rests on an implicit Fear of Missing Out and that the pitches we hear today for AI products and services are no less guilty than others of playing to that emotion. With that said, the strong trend of falling costs and rising capability of AI tools has made selectively embedding AI an attractive option for many businesses as long as those investments are targeted at areas that are leveraged for the business and play to the strengths of the current generation of AI tools and that means focusing on People, Patterns, and Probabilities.
About Mike Bracco:
Mike is a general management executive, investor and was the former CAO of Bank of the West, where he led the technology, operations, marketing, and human resource functions for the award-winning regional bank. An alumnus of Bain & Company and Bain Capital, he is currently on staff at Stanford University where he was appointed a Fellow with the Distinguished Careers Institute and is engaged with the Knight-Hennessy Scholars graduate fellowship program.
An active member of the Board of Governors for the Commonwealth Club of California. Mike holds a Bachelor’s degree in Computer Science from Rensselaer, an MBA with distinction from Harvard Business School and has been awarded a professional certificate in Genetics and Genomics from Stanford University’s Center for Professional Development.