Bits on Bots: Is a Bot Going to Steal My Job?

Our Perspective on Robotic Process Automation


Raunak Kasera

3 years ago | 5 min read

Ever since I started investing in early stage AI and automation startups, I’ve regularly found myself explaining what robotic process automation (RPA) is, and repeating my “schtick” over and over again.

This is understandable. It’s easier to think about a robot if it has some kind of tangible plastic or metal “body.” RPA however, is not your regular physical bot. It’s a piece of software code that performs repetitive tasks for you. To be fair, the complexity of software like RPA demands a better name than just “computer” or “program.”

With true AI still a long way off and this software designed to act like a human, we settled on calling them robots (thanks to sci-fi movies of the late 1990s) and the term robotic process automation stuck (there were many moniker variations before it did).

On this blog and elsewhere in the RPA world, we often end up referring to RPA robots as little minions with anthropomorphic qualities. But that is only half accurate. They are just very sophisticated computer programs. 🤷‍♂️

With the evolution of screen scraping in the early 2000s, automation joined hands with the then prodigal science product, artificial intelligence, to form RPA―a fresh kind of wine packed in an age-old conceptual bottle.

I would argue this original definition is a misnomer today in that production grade AI-driven RPA is a few years away. We still see most RPA vendors trying to add AI capabilities via partnerships or third-party add-ons. Notwithstanding, I know it. You know it. It’s the question on the tip of everyone’s tongue the minute someone mentions automation.

Is a bot going to steal my job?

It’s natural to ask this question, but like so much else in life, the truth is in the details. When you dig a little deeper, you find three core truths about RPA that really limit its ability to displace the human workforce:

  • With the current technology, McKinsey predicts less than 5% of all occupations can be automated entirely
  • Today, RPA bots cannot deal with unstructured data or execute decision-oriented tasks. They work best only on simple business rules that govern low-skill, repetitive tasks
  • The reality is that bots are being hired to do jobs that are hard to find human talent for. Think advanced analysis, help desk, and data correlation

To appease our hypothetical colleague worried about losing their job, even when machines do take over some human activities in an occupation, this doesn’t necessarily spell the end of jobs in that line of work.

For instance, we didn’t have data science as a profession just 10 years ago. On the contrary, the number of jobs at times increases in occupations that have been partly automated, because overall demand for their remaining activities continues to grow.

For example, the large-scale deployment of bar-code scanners and associated point-of-sale systems in the United States in the 1980s reduced labor costs per store and the cost of the groceries for consumers. But cashiers were still needed; in fact, their employment grew at an average rate of more than 2 percent between 1980 and 2013.

Startups have historically found it challenging to target the horizontal use cases of RPA. Focus has always been on building a sophisticated product that targets a single vertical with niche use cases.

We have a good factory but we need a fancy showroom.

Recently, a director of digital engagement at a mid-sized RPA provider admitted to me that although they have the brains in the firm to bring research around intelligent automation to life, they are incentivized to present niche RPA solutions customized for a client i.e. have a fancy showroom that appeals to the customer and sells.

This sentiment added fuel to my hitherto untested hypothesis that companies are only beginning to abstract and template a vertical-focused model to leverage across industries―an aggregator platform model where code and semantic models are reusable across industries and regions.

While there are many horizontal use cases―most of them use legacy technologies implemented in the 1990s and 2000s. Bots are more likely to be adopted in niche use cases, where there is a clear repeatable task, but broad basis bots are unlikely to take over many jobs.

I was curious to investigate the intersection of the broad use cases and innovative technologies.

With that, I looked at the top RPA vendors―Automation Anywhere, Blue Prism, and UiPath―to assess how they stacked up against the most successful applications of generic RPA tools.

Consolidating data into standardized formats

RPA is ideal for routine work of collating, rekeying, and posting data between systems. It will likely need to be used together with other tools to structure unstructured and semi-structured data or paper such as invoices and freight bills or voice calls.

Exhibit 1 shows a typical problem in which RPA could be used in finance and accounting to enter structured, semi-structured, and unstructured data into systems. Considering the critical capabilities of the overall development environment and feature integration, I believe UiPath = Automation Anywhere > Blue Prism.

Exhibit 1. Getting Data into Systems. Illustration by Hayon Thapaliya (@hayonnah on Instagram)

Moving data from A to B to C

RPA can be leveraged in many organizations where we have people rekeying data between finance & accounting and supply chain management systems and/or entering data into the two systems from digital images or paper, as shown in Exhibit 2. When we consider support for wider business processes vs. task-level movement, I hold Blue Prism > UiPath > Automation Anywhere.

Exhibit 2. Moving Data between Systems via People, RPA, BPaaS. Illustration by Hayon Thapaliya (@hayonnah on Instagram)

Automating or building a structured workflow

At its most basic level, the RPA tool handles single transactions.

At its most sophisticated, a pool of robots is capable of being deployed as required to follow process maps, move structured data, run straight-through processes in a “lights-out data center,” and be allocated to different processes in real time controlled via operational dashboards.

An RPA tool can be triggered manually or automatically, move or populate data between prescribed locations, conduct calculations, and trigger downstream activities. When it comes to Optical Character Recognition (OCR) or script library support including version control and reusing sub-elements, Automation Anywhere > UiPath > Blue Prism.

I want to nuance these generic use cases with the reality that there are various architectural structures of RPA tools, ranging from ones that operate on individual desktops with limited ability to take different data feeds to ones that operate on enterprise servers and are able to perform multiple scheduled tasks while meeting enterprise security criteria.

As investors, we look for the white spaces or areas where a critical mass of startups has not entered―could be a new market or gaps in the existing market or product lines.

Now for an analysis of the white spaces in the RPA domain for physical-digital interfaces? Well, that’s another blog post for next time. 👨‍💻

Originally published on Medium


Created by

Raunak Kasera







Related Articles