Decades ago, artificial intelligence arrived with huge expectations for significant increases in efficiency and productivity. However, despite billions spent on technology, project after project stalled—mainly because challenges with company strategies, technical hurdles, and cultures kept the potential power of AI unrealized.
Over the last decade, enterprises have migrated en masse to online platforms and cloud providers. This evolution has paved the way for computing capabilities to handle much more data while simultaneously generating troves of new data that these systems can now analyze.
This migration has laid the foundation for a new generation of automation and analytics—the shift from enterprise AI 1.0 to 2.0. This created the capacity for more sophisticated insights. This includes end-to-end process intelligence powered by focused solutions and machine reasoning that drives exponential gains in operational efficiency and productivity. Enterprise AI 2.0 is overtaking the shallow learning approaches and simple task automation of enterprise AI 1.0.
The organizational shifts underway to embrace these changes from the top down—starting with leaders who understand that future growth is rooted in digital transformation—have driven this transition more than anything.
Let’s take a look at how companies move toward enterprise AI 2.0.
From Experiment to Mandate: Getting C-Level Support
Enterprise AI 1.0 was a crucial stepping stone to driving success in the new 2.0 phase. Small wins and incremental advances over the past two decades paved the way for the broader buy-in we see across organizations today.
However, enterprise AI 1.0 was hamstrung from the start by organizational structures. AI was being applied almost entirely by data scientists with speculative-use cases that often weren’t aligned to business objectives, processes, or budgets. That led to a certain amount of irrelevancy and a lack of buy-in, especially at senior management levels.
In one study conducted just before the pandemic hit, 93% of respondents—C-level technology and business executives representing Fortune 100 corporations—identified people and process issues as the key obstacle to implementing AI.
Bolstering that assessment, Gartner estimated in 2017 that up to 85% of big data projects fail—with other studies putting the failure rate in that range—due to a lack of buy-in among all levels of management. These failures often stem from data scientists driving AI investments that either don’t align with business objectives or aren’t accessible to frontline teams who could best leverage them.
A key difference in enterprise AI 2.0 is the greater ownership of the transformation at all organizational levels, including C-level sponsorship of AI applications that focus on strategic business impact.
McKinsey may have been one of the first to study this phenomenon. In 2019, the consultancy found that commitment from management was a significant factor in the success of AI projects. Experts and industry leaders have echoed this idea, including Chris Chapo, senior VP of data and analytics at The Gap, who spoke on the topic at Transform 2019 in San Francisco.
“Sometimes people think ‘all I need to do is throw money at a problem or put a technology in, and success comes out the other end,’ and that just doesn’t happen,” Chapo said, explaining that companies often “don’t have the right leadership support, to make sure we create the conditions for success.”
In sum, deep support from the C-suite is the foundation of AI success.
From Nascent Skills to Citizen Data Scientists
Enterprise AI 2.0 requires a team with an advanced mix of skills at the intersection of machine learning, software engineering, data pipeline engineering, governance and compliance, AIOps and CloudOps. These skill are needed to translate the initial work done by the data scientists within their sandbox environments to production-ready systems.
Enterprise AI 2.0 leverages sophisticated technology platforms and packaged solutions that streamline, simplify, and accelerate AI-driven innovation. Rather than cobbling together disparate tools and siloed environments, teams work with integrated approaches to manage data and machine learning pipelines from early development through production deployment and ongoing management. Purpose-built solutions abstract the underlying data and model development complexities while significantly hastening time to value.
Enterprise AI 2.0 will also see the growth of new platforms that unleash the power of AI for employees at all levels of training, throughout entire organizations – the democratization of technology. These business users will use next-gen tools that harmonize data and automatically build predictive models and intelligent applications.
These employees become citizen data scientists who can use AI, low-code/no-code platforms, and their deep domain expertise to overcome business challenges and exploit latent opportunities. They accomplish this in self-service mode, thus becoming critical enablers across the entire enterprise.
From Machine Learning to Machine Reasoning
The predominant predictive modeling approach used in enterprise AI 1.0 is based on supervised learning, leveraging shallow algorithms.
In contrast, enterprise 2.0 will usher in a wide variety of modeling approaches, including lightly-supervised, semi-supervised, self-supervised, low-shot, and unsupervised learning. In addition, we will build more intelligent systems that go beyond merely identifying patterns within data. We’ll create a more nuanced understanding by deriving meaning from enterprise data and user interactions, understanding reasons for a particular behavior or phenomenon.
These next-generation systems, based on domain-specific semantic intelligence, will leverage machine reasoning powered by propositional or probabilistic knowledge. This will work in tandem with machine learning to bring AI closer to human-level intelligence.
For example, consider an intelligent system that uses multimodal sensors to detect the operating state of a centrifugal pump in an industrial environment. The system can ingest sensor measurements, including pressure, temperature, flows, and vibration, to predict any upcoming performance degradation or equipment failure. By drawing upon a library of failure modes and effects analysis, the system can automatically act or propose mitigation advisories.
From Narrow Tasks to Intelligent Systems
Enterprise AI 1.0 machine learning has a narrow scope and simply added automation and intelligence to tactical capabilities. Enterprise AI 2.0 capabilities will broaden AI automation, so entire business processes and decisions can be more policy-driven and autonomous.
Imagine systems of intelligence that can help retailers understand each of their target markets to anticipate shopper demand. This allows sellers to execute personalized promotions, streamline supply chain logistics, ensure ideal inventory levels, and automatically set pricing to maximize quarterly business objectives.
The evolution of governance is also crucial to enabling enterprise AI 2.0. Companies deploying AI will need to make sure they self-impose system regulation to supervise AI-based decisions. This allows them to root out imprecisions, biases, non-compliance, or other problems as AI technology digests models.
Remember how exciting it once was when AI evolved to answer FAQs or score and sort a set of leads for the sales team? Yes, enterprise AI 1.0 solutions handled simple tasks well. This functionality isn’t going anywhere. But we can do so much more.
Companies are already changing their cultures, upgrading their data infrastructure, enhancing their systems and technology, and refining their processes to embrace enterprise AI 2.0 tools and solutions. These changes coupled with AI analytical advances can help companies exploit the full potential of enterprise AI 2.0.
About the Author:
Eshwar Belani is an operating partner at Symphony AI.