Last week’s column spoke of arenas into which machines cannot transgress, this one explores where they do. Some of the slowing in white collar employment is attributed to the rise of automation, and while we speak of it as though it’s well understood, it strikes me that many lay observers might not understand the mechanics of exactly how such automation affects jobs.

Most of the replacement of humans in white collar jobs is accomplished with the use of bots (short for robots), and an entire sub-industry has been created for them. I predict that this sub-industry will last a while before being adopted as just another product from the stables of technology firms, but that’s a subject for another column. Meanwhile, it’s probably worthwhile understanding how a bot truly works.

The first step in setting up a bot is to create a “recorder" that cohabits a worker’s machine or computer. It then learns the basics of the job over a few days, by recording keystrokes as the worker does his or her job, and noting which parts of enterprise support applications such as billing, supply chain management and so on are accessed by the worker, and in what sequence, to complete a series of tasks. In a short while, the recorder learns a large proportion of the repetitive tasks a worker performs, and if allowed to stay on the worker’s machine long enough, can even pick up the nuances of what needs to be done when exceptions in process handling occur.

After the recording has been completed, a team of computer programmers converts the recorded actions into an automated script that can run on demand. Since the bot can now almost completely replicate what the human has so far been doing, its automated script can now replace the human.

Automation Anywhere, is a “growth stage" start-up firm that produces robotic process automation (RPA) software. I recently met Anubhav Saxena, executive vice-president of Automation Anywhere, and a member of its executive committee. Saxena, a former colleague, manages its global strategy, operations, and partnerships. It claims to be the world’s leader in the bot field and that it can automate any enterprise’s business processes by producing a “digital" workforce. That said, it has plenty of competition in firms such as Blue Prism and in the RPA divisions of technology services firms.

Saxena claims that each bot can now take the place of between 3-5 human workers. This has the follow-on effect of requiring less user licences from companies such as SAP and Oracle, on whose software platforms most enterprises run today. It also has the more sinister effect of converting human workers into “digital" ones, thereby directly contributing to job loss.

This is not the job loss we fear from Artificial Intelligence. It is simply the replacing of manual processes with mechanized ones, much as a tractor replaces the need for multiple farm hands. Many industries still require millions of person-hours to get repetitive tasks done: for instance, completing month-end Automated Teller Machine reconciliations across North American banks is an exercise that takes an enormous amount of time and effort. Little wonder then, that Automation Anywhere counts one of America’s largest banks as a large customer. This customer has already installed over 5,000 bots, as has a large Antipodean bank, which now counts over 3,000 bot “employees". These bots are the clerks of the future.

That said, Automation Anywhere markets itself as having the capability to automate any enterprise’s business processes, no matter how complex. According to Saxena, it aims to be the world’s largest “employer" without having any employees, simply by deploying thousands upon thousands of such bots, just as Airbnb is the world’s largest hotel chain without owning a single hotel room. Saxena claims the firm has created a variety of bots in a hierarchy, based upon the complexity of the work the bots do. At the lowest level are “task" bots, or those that perform repetitive tasks, like the ones we have already discussed. The next level up is a “meta" bot, who much like a human manager, has cross functional knowledge of many business processes, and whose job it is to orchestrate the task bots so that they do their jobs in the right sequence to effect a specific outcome. Another type of bot is the “IQ" bot—which has specialist knowledge; in the human world, these would be “subject matter experts". At the highest level are two other bots: the “vision" bot, who acts much like a CEO would, in setting vision for a business enterprise, and the “insight" bot, which with the use of sophisticated data analytics, arrives at nuanced judgments before taking decisions.

To support this phalanx of bots, Saxena is tying up with alliance partners who today are responsible for the business process work that has been outsourced to them by the world’s largest corporations. IBM has recently announced an alliance with Automation Anywhere, aimed mainly at deploying task bots within organizations that already run programmes like IBM’s Operational Decision Manager, which is used to manage complex business processes. This is not surprising, since the larger firm is hungry to add to digital data that its heavily marketed data-crunching Watson platform can use. I wrote recently in this column of how IBM has acquired a medical imaging company with access to billions of radiological images, just so that it can feed these data into Watson’s brute force pattern recognition engine, since it hopes to raise the accuracy of certain types of disease diagnostics.

What then, of today’s computer programmers who are in danger of losing out? Saxena sees no problem here. The production of bots still needs software tools and requires the ability to programme a script. He feels that retraining these individuals will allow them to stay employed as the world’s corporations shift to a digital workforce for their clerical back-office processes.

It is the clerks who will have to watch out.

Siddharth Pai is a world-renowned technology consultant who has personally led over $20 billion in complex, first-of-a-kind outsourcing transactions.

Close