Deep AI Verticalization: No Silver Bullets

Around since the 1950s, artificial intelligence has taken on a new valence over the past decade. In both reality and rhetoric, AI has emerged as a leading subject in business technology discussions and is viewed by many as both a silver bullet and an existential risk to business organizations and society as a whole. Whatever side of the AI debate you are on, no amount of analysis and scrutiny is too much. The worst thing that can emerge from this critical focus is better products, services, and processes. Either way, technology and business leadership must come together to let AI yield the sorts of outcomes that justify the investments made.

The jury is still out regarding the application of AI to business. While there have been profound advances, there have also been a host of false promises and hyperbolic predictions that never materialized. As one particularly famous investor put it, "You promised us flying cars, but you gave us 140 characters."

Indeed, the distinction between “real” and “rhetorical” AI has become a sine qua non of success not only in Silicon Valley but in all of industry. Mere claims can get a company funded- perhaps but cannot deliver real value. Ask any CIO about the historical divergence between promise and reality in technology and you’ll see that it’s no different with AI.

Irrespective of whether one takes a vertical or horizontal approach, deficiencies can be found quickly. Take, for instance, the horizontal area of sales. Here, there are countless claims from AI for better prospecting to AI to “know your customer.” Firms that could scarcely afford even one solid engineer, scientist, or data maven put out marketing slicks about AI-driven sales. Such notions are too far-fetched to be given any credence in sensible circles. With regard to verticals, take the area of investing. One can count the number of companies that claim to use AI to create trading signals in the hundreds.

With regard to deep verticalization, the sort of context-driven approach that successful organizations require, the problems get hairier by the day. Solutions require a clear connection between technology and business leadership teams but also must be democratized within the ranks of the organization. Here are some areas to consider:

1.    Business Context— AI cannot be “delivered” in a vacuum. In the absence of deep vertical knowledge, it is scantly possible to train computers to understand the relevant patterns or to process the relevant data in the context necessary. Pulak Sinha, CEO of asset management data platform Pepper, makes the point clearly: “In the asset management and investment industry, if you don’t understand the regulatory constraints on actions, then you can’t ‘innovate’ your way to success. That parameter is simply too big an elephant in the room to gloss over.”

2.    Data, Data, Data— AI comes about when computers have large training sets of data from which to divine patterns to understand and then extrapolate from. Such data sets might not exist yet in particularly complex verticals. In the fertile area of computer vision as it applies to real estate, these data sets are being formed as we speak. As Malcolm Cannon, COO of Quantarium, says, “Wonders can be done with computer vision with just enough data. But the key word is enough. Below a certain threshold, companies can claim all they want but won’t be able to deliver value.”

3.    Timing— Business decisions are time-bound; they must be made within a certain window to be effective. For AI to deliver results in the long-term, it must offer insights that can be acted upon in that window rather than continually ingest new data sources to offer even basic decision-support guidance. A noted stock trader makes the point with no small hint of irony: “Anyone can be a great trader in hindsight. I know exactly what to do yesterday. How can you help me figure out what to do today and tomorrow?”

4.    Culture— The decision-making and investment culture in an organization will determine whether AI can help organizations push ahead. In some verticals, the culture of risk-mitigation militates against quick-decisions, which in turn suggests that AI engines must be trained on different questions than in agile industries. In high transaction sectors, AI can be focused on short-term decision making while in long-lead and low transaction sectors, AI should be focused on larger, macro-questions that might yield answers in years. As Jeremy McCarty, CEO of Valligent Technologies, opined, “The right tech with the wrong culture might as well not exist.”

There are countless other areas upon which AI verticalization depends, but these four represent the major ones.

All this said, AI verticalization has profoundly affected financial services, healthcare, manufacturing, and a host of other verticals. Still, the most fertile times are ahead. To realize success, organizations must take into the account the admonitions we’ve offered here. They must also abandon the notion that AI or any other technology or process constitutes a silver bullet.

Context matters. Data matters. Timing and culture matter.

When this is understood, huge progress can be made. It requires clear collaboration between technology and business leaders and, further, the banishment of any notion of “silver bullets.”

Related Content:

Enterprise IT Leaders Face Two Paths to AI

AI Adoption Spectrum: Are You Ahead of the Pack?

Vertical AI is the New Black