ChatGPT and other artificial intelligence tools have dominated the conversation lately. Their power to imitate human writing and art is raising concerns that machines could start replacing white-collar workers, the way they took over many blue-collar jobs in the 19th century. We at Digitate are thinking about machines’ role at work too, as we develop software tools to make the autonomous enterprise real.
In our vision of the “autonomous enterprise,” machines (or rather, AI algorithms) fulfill highly repetitive or defined tasks, while strategic, decision-making tasks are driven by humans.
You may think that rule means it’s easy to decide which tasks can be assigned to machines. But as AI and machine learning continue to become more sophisticated and powerful, the dividing line keeps moving. However, the distinguishing factor remains the same: Whether the task under consideration handles data in a defined or undefined way.
Defined: Activities in the defined cluster offer all the information (data) and instructions that you need to perform them. No information is hidden, and the specific instruction to use can be calculated using the data available. Defined data activities are ripe to be machine-managed.Undefined: Activities in this cluster don’t offer all the necessary information to perform. Intuition, interpretation, analysis, deduction, and guessing are required. Undefined data activities do not adapt well to machine management.
Games can help to understand how to deploy AI
Games that are prime examples of these two clusters are chess and poker, respectively. These categories were first defined by pioneering mathematician and computer scientist John von Neumann (who created a whole field of study with his 1944 book, Theory of Games and Economic Behavior).
I was reminded about von Neumann’s distinction when I attended a speech last fall by scholar and poker champion Maria Konnikova that covered some of the points below.
First, think about a chess game. It has a defined set of pieces with specific roles, a clear set of rules, and a defined space (the chess board). All data is on display for both players, with no hidden information (and no ambiguity about whether a move is legal or not). The total number of all possible moves is very high, but not infinite. This means a machine equipped with a good set of algorithms and enough computing power can beat any chess champion. (In fact, this first happened a quarter-century ago.)
Now think about poker. It also has a defined set of pieces (a deck of cards), a set of rules, and a defined space (the card table). However, not all information is on display; in fact, the central mechanism of poker is to guess which cards your opponents secretly hold, and then successfully predict how they will bet. The game must be played by assumptions, clues, and intuitions about both the cards available and human behavior under specific emotional pressures.
Know when to fold ‘em? That does not compute
Here is the major difference: Machines don’t do well when not all the necessary information is available.
While I realize people might object that AI is progressing and it is mimicking human intelligence, there is no enterprise-grade application of such solutions yet. At least for the next few years, machines still can’t beat us at poker.
End-to-end enterprise operations are closer to poker than chess because often not all data are available. Decision-making is often driven by limited data, information interpretation, and intuition.
Machines are very effective and efficient at managing tasks with a clear set of data and a well-defined set of rules, also known as Standard Operating Procedures. In many enterprises, a wide range of operations from sales to HR can be described with SOPs, and therefore automated. (In IT operations, where I’ve spent my career, 80% of tasks can be machine-managed.)
The typical journey towards an autonomous enterprise usually moves through these phases:
Manual: There is no support by machines; all tasks are executed by humans.Augmented: There are specific routines that alleviate repetitive tasks, but these routines are triggered by humans. (The most common phase nowadays.)Automated: The machine reacts to a ticket (human’s request), triggering a specific routine to solve the problem.Autonomous: Machines suggest and execute actions to prevent incidents or improve overall performance. Usually there is a supervisory period where the human is “teaching” or modeling actions for the machine to take, which later it will execute without supervision.
At Digitate, we built “ignio™,” our flagship AIOps platform for IT and business operations, to become fully autonomous. After its “learning” period ignio’s proprietary machine learning algorithm can filter out excess information generated by the production environment, focusing only on the activities needed to improve or rectify the situation.
Staying one move ahead with autonomous operations
Like any good chess computer program, ignio has a library of over 10,000 customizable moves (use cases) to apply when a situation occurs. Of course, at the beginning ignio will seek human approval before executing the use case. But when the machine learning period is over, ignio is ready to not just self-heal IT problems but optimize all kinds of business processes.
The bottom line: ignio is designed to be an autonomous enterprise solution for IT operations. ignio focuses on the whole landscape, not just single aspects such as data flow, ticket management, or monitoring. ignio is not merely a tool for a specific need, but rather a solution to make the IT autonomous enterprise a reality.
And you can bet your whole stake on that deal.
To see ignio in action, click here to request a demo.
Artificial Intelligence, Machine Learning