The concept of “self-driving” capabilities in any realm of business means they’re continually optimized through data-driven guidance. In other words, people and machines can, under a growing range of circumstances, be trusted to take actions by themselves, because they’re guided in real time by the best data and best predictive and prescriptive models available.
Self-driving business ecosystems tap into the best expertise available for the challenge at hand. More decisions are too complex for any one person or even team to have all the answers. Likewise, many challenges are growing too dynamic and multifaceted for any one software algorithm, artificial intelligence model, or rules-driven program to address definitively.
Automation is a big piece of the self-driving vision, but it’s far from the sum total of this paradigm. With advances in AI and the “internet of things,” people are starting to realize that, within specific scenarios, enterprises realize that they can trust computers to automate a growing range of problems that fly by too fast or involve far more factors than human experts could ever keep up with. Considering the complexity of the algorithms involved, plus AI’s ability to learn from fresh data and dynamic circumstances, automated systems can drive unprecedented results that might not have occurred to any human expert beforehand.
When Oracle Corp. executives discuss self-driving as a unifying theme for their cloud solutions portfolio, as they did Monday at its OpenWorld conference in San Francisco, they often use the term “autonomous” as a near-synonym for the paradigm, but with a clear emphasis on machine learning-driven automation at the platform level. For example, in Monday’s afternoon keynote, Oracle Executive Chairman and Chief Technology Officer Larry Ellison spoke of his vision for a second-generation cloud built to run Oracle Autonomous Database, which he claimed is “the industry’s first and only self-driving database.” Within Oracle’s infrastructure-as-a-service, it supports “new Oracle Cloud Infrastructure security services that are highly automated, detective and predictive to help remediate threats.”
As Ellison went deeper into the vision, he described how AI enables the database — and by implication the other “autonomous” platforms in its cloud portfolio — to deploy, scale, manage, tune and patch themselves automatically with no downtime. “With Oracle Autonomous Database, there is nothing to learn and nothing to do, which makes it really easy to use,” he said. “Developers are more productive, they bring up new applications, they do a better job of analyzing data. Your system is more reliable. It never goes down.”
Ellison was most compelling when he discussed AI-driven automation’s indispensable role in cybersecurity. He discussed how Oracle has architected its cloud infrastructure from the ground up, using AI-driven automation to erect “impenetrable barriers that can block threats from getting into cloud, and autonomous robots that can find those threats and kill them,” while also preventing threats that pop up in one’s customer’s cloud zone from invading other customers’ zones.
“This is all completely automated,” said Ellison. “The threat is discovered automatically with no human beings involved and the security path is applied immediately while the platform still runs.”
Automation and jobs
Where he was less convincing was in addressing the legitimate concern that all this platform-level automation might put many information technology professionals out of a job. He addressed the issue by vaguely predicting that somehow all of these skilled technical personnel — who are Oracle’s core user base, after all — will rapidly shift toward business analytics, data science and other specialties where their data chops come in handy. However, as a longtime industry observer, I doubt that many database administrators, for example, have the subject matter expertise, business skills and other aptitudes that would suit them for these professions.
Fortunately, the morning’s keynote by Steve Miranda, Oracle’s senior vice president, of applications development, delved into the other side of the self-driving equation: data-driven “next best action” and “intelligent process automation.” That refers to data-driven guidance for human decision makers as enabled by AI-infused digital assistants.
These enhanced capabilities were amply evident in the rollout of Oracle Digital Assistant and of the new business apps that incorporate this AI-powered chatbot and its data-driven next-best-action recommendations. For example, enhancements to Oracle’s Human Capital Management cloud apps use the digital assistant to deliver a dynamically AI-tailored employee user experience; support automated sourcing, candidate search, referral recommendations and recruiter system connections; identify best-fit candidates; and detect access anomalies that might compromise data security and privacy.
In a similar vein, the latest enhancements to Oracle ERP Cloud support intelligent process automation. It blends intelligent rules-based processing with the digital assistant’s AI-functionality chatbot interface. It simplifies and secures expense processing and audit compliance, generates vendor-specific offers in exchange for early payment of outstanding payables, accelerates multifactor categorization and ranking of suppliers in the procure-to-pay process, and constantly examines all users, roles and privileges against a library of active security rules.
Oracle amply demonstrated the breadth and depth of its investment in self-driving business infrastructure in both of these senses: automated platform optimization and augmented decision optimization. But what was missing was any clear vision for combining these two “next best action” spheres of operation: the bot-centric and human-centric, respectively. Ellison’s and Miranda’s discussions of these distinct paradigms seemed to coexist very loosely within Oracle’s go-to-market strategy.
In the comprehensively self-driving enterprise of the future, people and computers collaborate seamlessly. As Walter Frick stated several years ago in this Harvard Business Review article, “in many areas, the combination of human and machine intelligence will outperform either on its own.” In the era of AI, the IoT and cloud computing, we need to stretch our notion of productivity to reflect the increasing coequal and codependent collaboration of humans and machines.
The human touch
Automated systems can make amazingly expert decisions, but only humans can truly judge whether the criteria of compliance, appropriateness and value are met by computationally generated outputs. To power next best actions, a self-driving enterprise needs to have infrastructure that can automatically locate the next best expert — humans, machines or some partnership among any and all — depending on the problem at hand.
To power this vision, there needs to be a shift in how we think of “autonomous.” We should move away from autonomy as an absolute — in other words, completely unassisted, self-sufficient bot-driven automation — toward more of a spectrum. Many real-world operating scenarios require varying degrees of semiautonomous operation by bots that are guided, supervised and controlled in part by human experts in various scenarios.
Indeed, building failsafes into AI-powered security automation scenarios absolutely requires the option of manual override when the underlying models reach the limits of their competency or when their predictive accuracy in a particular circumstance declines below acceptable confidence intervals.
I hope Oracle will have a more comprehensive vision of human-machine collaboration, grounded in the concept of next best expertise, incorporated into its self-driving solution strategy by the time of next year’s OpenWorld.