Connect Quantify Optimize—The Model
The CQO Model
We saw in Part 2 how digital interfaces are reshaping our businesses. Be it mobile, Web, or sensors, we are increasingly opening up more and more interfaces with our customers, employees, and the environment. All of these connections then create gargantuan amounts of data that allow us to quantify our business in entirely new ways, which we covered in Part 3. And based on all of this data, we should be able to intelligently optimize our businesses as we discussed in Part 4. In this section, let’s look at the application of the connect, quantify, and optimize (CQO) model.
The start point for almost every good digital project has been the rethinking of a customer journey. Let’s say the check in process for an airline, or the mortgage application for a bank. Let’s consider the very hypothetical scenario of a near-perfect digital project where everything is done as it should be. In this world, the team studies the user (customer) behavior thoroughly, through observation and conversation, builds personas around the types of users, and understands their current user journey, and both stated and latent needs, and challenges with the current journey. They then run design sprints to recreate this customer journey. This effort typically culminates a new mobile app or a website redesign. It could end up improving an existing product or creating a new one. For example, a bank could redefine its mortgage processes across the board, or it could create a new 24-hour montage product for specific types of customers. In either case, it is likely that some existing processes will be modified, and some new processes will need to be created. The customer application form and the way it is submitted could be modified. But a new straight through process could be created, which brings the assessors, surveyors, and lawyers together in a much more efficient manner.
If all of this goes well, you now have a customer using and valuing a new digital interface. But additionally, the data output means that every step of your customer journey becomes visible. You might, for example, notice that most of your customers for your new product are using more expensive devices and are looking at high-end properties. Which in turn might lead you to focusing your sales and marketing efforts. But what if it led to your creating a two-tiered product structure for meeting the needs of a broader segment?
This is where it gets interesting because the data may throw up many more insights, which could reshape your business in big and small ways. In a 24-hour mortgage product, you might, for instance, realize notice from your data set that you have a set of customers who apply for a mortgage only after selecting a property. This may allow you to provide ancillary services to this set of people beyond the mortgage—such as bringing together an ecosystem of trusted providers aimed at helping customers to repair and redecorate, or even provide ancillary loans for specific house-related activities. This stage may actually end up changing your business model in some way—either by changing the commercial model or the way you generate business or fulfill orders. This is optimization, and very few companies are actually at this stage. Based on anecdotal evidence, I would suggest that 60 percent of all businesses are still getting the interface right, and about 30 percent have moved to the data stage. Only about 10 percent at best are at the optimization level. When people talk about digital transformation—this is the promised land.
Optimization can take many forms—in it turns out that Airbnb didn’t just disrupt hotels, it also disrupted the home rental market. In France, which is the second largest market for Airbnb, the data suggests that it pushes rents up in places such as Paris and Marseille. For Airbnb, renting entire homes for short leases has become a more powerful approach compared to just a spare room. Optimization may be just a minor tweak—such as a utility company changing a route for its service van, or it could end up as a significant rethink of the business model, such as a coffee shop charging for time spent in its Wi-Fi-enabled store, and giving the coffee away for free. Each business and industry will have its own patterns for optimization, based on context, current inefficiencies, and opportunities. Optimization is optional—you may just be happy with the first two stages, but it is highly improbable that the higher visibility and data offer no opportunities for improving your business.
The Rise of RPA
It would be fair to say that the past three centuries of industrialization have largely been about three layers of automation. The first was a shift from manual to steam power—of ploughs, looms, and locomotives. The second was a shift from steam to electricity. And the third, in the second half of the 20th century, was about a shift to computerization. While the first two were all about the source of energy, the third was about management. Not just of the machines, but also of the processes, both in the factories and in the offices created by the first two stages of industrialization.
Today, we take automation for granted. From the coffee machine in my kitchen, to the ATM where I get cash, and from the contactless credit card I use to pay for my train ticket, to the way my travel expenses get reimbursed. There’s automation all around us. Even while I type this, a copy of this document is being stored on a cloud using a behind-the-scenes automation. The role of automation in our lives is reflected in its prominence popular culture as well. Charles Chaplin gets entangled with the machine in Modern Times (1936). Ferris Bueller (1986) automates his dummy in the eponymous movie. And even in the animation film Sing (2016), Rosita automates her family’s morning routine from breakfast to school departure, with ultimately calamitous impact!
A dark side of process consistency is that it has reduced humans to robotic beings, devoid of any intellectual contribution to their job, and reduced to following a tightly defined script or instruction. In fact, that reason that so many jobs in factories and offices have already been automated is that they were jobs that had been broken down into a level that could be addressed by subhuman intelligence. From the number of minutes that potatoes must be fried, to the number of rings of a phone call before it needs to be answered. Every company depends on process adherence driven by consistency and repeatability. The industrial era had already thus reduced humans to machines, so replacing them by other machines was relatively simple, and had the added benefits that computers and machines don’t get tired or bored, and their output isn’t impacted by their emotional state. But this meant that only those processes where judgment was removed, that is, had been robotized, could be automated—such as fast food preparation or call center responses.
A lot of processes in the previous generation of automation involved getting people to enter data into systems. But this also resulted in multiple and often unconnected systems and data islands. A single process could span these systems and involve a manual data input from one system to the other. When businesses look at their financial cycles such as order to cash, or procure to pay, these manual points represent inefficiencies and sources of errors. Robotic process automation (RPA) systems create robots that can automate this by mimicking the steps followed by the human process. These narrow, software robots are designed to only perform one task. For example, get your accounts receivable information into your CRM system.
Doing this under the hood, or via an API, may be expensive and time-consuming. The category of tools broadly known as RPA looks to solve this process-level automation and optimization problem. They still use programmatic tools, but they mimic the human act of extracting from one system and entering into another. This is done through software robots, which are designed for individual processes and keep working away to run the process as a human should. Although the business case for RPA is often built around reduction of people, the reality is that it enables people to be freed from rote and repeated data entry tasks for more rewarding work. But it is also true that over time, this makes the overall business process more flexible. RPA tools have attracted huge investments of late. UI Path, one of the three major providers, had an IPO in 2021, which valued the company at $29bn. A number of new players are now in the fray, coming in from different software segments such as CRM, middleware, and AI, attracted by the high growth in this segment. According to some reports, automation was the fastest growing area of enterprise software in 2020, which is not surprising, given that a typical large organization deploys upward of 150 to 175 applications, and many processes require interactions with a significant cross-section of these apps.
Tip: Think of any job that requires the reduction of human judgment in favor of a predefined set of steps and ask yourself why we need humans to do this job instead of machines.
There are also processes that have never been automated because they have worked off physical documents or handwritten forms. These may have originated outside your organization—for example, bills of loading for cargo, or diagnostic lab reports for health insurance companies. Increasingly, RPA products are able to offer digitizing processes involving these kinds of documents on the back of character recognition, or even handwriting recognition. Additionally, there are libraries of industry terms that are recognized by the software platforms, so if you’re looking at medical compliance information, the software will be able to recognize abbreviations, terms, or even logos related to compliance bodies. Today’s technology can therefore do more, and automate more complex tasks—such as grammar and spell-checking documents, recognizing a song and matching the lyrics to the music, or checking expense claims, where the instructions can’t be codified into a set of sequential tasks. This is often referred to as intelligent process automation (IPA) and is a natural evolution of the RPA model.
The Future of Automation
One of the fascinating aspects of robotic automation is the implication for the workplace and HR. As more rote jobs are picked up by software, systems, and robots, humans are no longer performing those jobs and are instead presumably doing more creative work, or work involving more judgment or higher order skills. But that also requires rethinking of metrics and goals. HR organizations may need to rethink job descriptions, training and reskilling, and measurement—how roles and performance are measured in this world of co-bots (colleagues who are also robots). The optimization of processes may therefore also lead to an optimization of HR.
A useful way of thinking about automation over the next few years is that previous generations of automation involved systems that required humans to be trained to operate them. In the emergent digital world, the systems use machines that are trained to work with humans.
There is a strong correlation between the size of organizations and the amount of time wasted within its walls. Try this exercise. When you go for lunch with another person rather than just by yourself, you will invariably add a few minutes agreeing when and where to go, and just when you’re ready, the other person will be finishing a call. Now try with three, four, and five people, and watch the coordination time shoot up exponentially, till you simply have to name the place and time and ask whoever wants in to be there.
It’s not just at lunchtime though. In large companies, time is wasted in a vast variety of ways and at an industrial scale. E-mail is one of the big culprits, but so are internal admin and the never-ending drive toward faux efficiency. Many organizations will save headcount and the cost footprint of the IT department, the travel desk, and other back office operations, but nobody counts the hours wasted by employees on systems that don’t work. Perhaps because decision makers in large traditional environments often have assistants doing much of their administrative work, and don’t feel the pain themselves!
Being insensitive to employee time makes your business less efficient, it impacts culture and retention of talent, and it promotes shadow IT—where your people use their own digital tools, leading to compromises in your information security. Companies, such as KDS (now the Neo Technology Group), provide systems through which you can book travel and complete expenses in minutes, with minimal manual work. I have seen this solution elicit an almost emotional response from people in large companies, who have to spend hours over these administrative tasks.
Remember that, thanks to our personal digital tools, we’ve learned to value every five minutes because it can be productively used. Most people are trying to fit in as much as they can into one life. Why wouldn’t you give them the tools to do as much as they can for their job?
The New Productivity
When we talk about productivity in industrial and economic terms, we usually implicitly mean labor productivity. Typically measured in terms of the output per individual. And there has been a concern that over the past decades, despite all the influx of technology and computing, labor productivity, or output per employee hasn’t really shifted much. From where we stand, this may actually be asking the wrong question. Ford produced over six million cars, in 2007, and a similar number in 2020, but it has done so with 40 percent less people, from 300,000 to 185,000 in that time. Note that technology that makes people more efficient is different from technology that replaces people. In the case of the latter, the simplified measure of labor productivity is no longer valid. Capital now constitutes an increasingly larger part of the value of a typical product. Often therefore, investment decisions are made on the basis of how to make the machines more productive, rather than the people. The battleground has shifted to the return on capital. While labor productivity remains an important metric for the government, the landscape for investment may have shifted.
Robots
When Isaac Asimov was born in Petrovichi, in Russia, the Czech author Karel Capek was publishing his science fiction play Rossumovi Univerzální Roboti (R.U.R.)—the play that would give the world the concept of a “Robot.” Ironically enough, when Asimov’s family made the journey of almost 7,500 km to Brooklyn New York, three years later, Capek’s play was also crossing the Atlantic to be screened in the United States as Rossum’s Universal Robots. Asimov went on to become one of the great thinkers about robots—constructing the three famous laws.
In a way, this all seems very ordained. But let’s remind ourselves that robots were intended as workers who would perform tasks. They would be powered by computers, so their capability was circumscribed by the processing power of the computers of the day. As we move into the world of hyper-intelligent computers, including all forms of AI, robots will continue to grow in capability. As such, robots fall into the realm of automation. They are designed for performing physical tasks that humans can’t or don’t want to do. Carry things, fix things, lift things. Any number of manufacturing processes are not run by robots–a significant part of car is built by them.
The robots conceived by Asimov were actually an amalgam of a number of distinct ideas. The first is a physical, humanoid form, which would enable the anthropomorphism of robots. The second is an AI-like brain, which could power the robot and enable it to perform tasks. The third is the range of tasks the robot can perform—starting with specific narrow physical tasks, to rote processing work, through to more sophisticated intelligent work. In parallel, the interface or communication could evolve from very basic and specific commands to natural language and voice-based commands. The path to this humanoid version of robots is circuitous, and the progress is, to quote William Gibson, “unevenly distributed.” We have made a lot of progress in narrow tasks performed by robots. These range from physical tasks performed by physical robots—such as automobile manufacturing, to processing tasks performed by software robots, which have no physical presence. Both however are examples of automation—which is essentially the act of replacing a human performed process with a software and/or hardware. Usually resulting in less errors, more precision, efficient scaling, and crucially more transparency and data.
As each task is picked up by a robot, or a machine, that is electronic rather than mechanical, it creates a new source of data. The benefit of automation is immediate, but the benefit of this new interface and data is arguably larger over the long term. As with any interface, it starts to give us an ability to improve the performance of the machine and therefore of the process itself. The interface is between the robot and its environment, or its task. But it is equally an interface between the organization and its environment.
The idea of robots as quasi-sentient beings is covered extensively in science fiction. Two that deserve mention are: 2001 A Space Odyssey (cocreated by Arthur C Clarke and Stanley Kubric in 1968) and I, Robot (written in the 1940s by Asimov, and turned into a film by Alex Proyas in 2004). Both explore the notion of a powerful and thinking computer that goes beyond its brief to take decisions that are contradictory to its stated purpose and with potentially deleterious and far-reaching effects. But the very idea of a thinking robot or machine is the realm of artificial intelligence, which we shouldn’t confuse with the notion of automation. Think of it as the brain versus the body. In the world of automation—be it a robot, a piece of software, or a machine, we are simply building a tool to perform tasks. The task robot can do nothing more than the tasks it is explicitly designed for. And as such, it is simply another interface that will generate data for the brain. To understand how we are looking to supplement the brain, read the chapter on AI.
The word robot is used loosely and to mean a number of different things today. One way of classifying robots is as hardware (having some physical form) and software robots. A robot may be an arm connected to a machine that can lift pallets and stack them. Or something that automatically fixes caps on bottles. The question is, what makes these different from traditional machines? Why is something a robot and not just an extractor, or a press? Likewise, software robots automate tasks such as completing workflows or processing documents or images. The same question may be asked—why is it a robot and not just software? The answer to both of these questions lies in how we classify automation. Task automation is done by traditional machines or software. They are narrow in their focus and can do one thing with little or no freedom. Usually, when we mean process automation, we are referring to a series of one or more tasks that cumulatively achieve a goal. The focus is very much on the goal rather than the task. While any machine or software can focus on a task it’s given and perform that task repetitively (like fixing hinges to doors or vacuum sealing bottles), when we refer to robots, we are usually referring to the achievement of goals that may involve more than one task. Often there is some basic decision making involved, and elements of higher-level processing and even bits of AI may be involved. Any process automation involves some level of basic decision making. For example, the Roomba automated vacuum cleaner, or Gita the carrier robot. Both of these can scan the environment using a range of technologies, take a number of decisions but ultimately are focused on completing a job that they are assigned.
Tip: See the YouTube videos from Boston Dynamics (now a part of Hyundai) to understand the power of what robots can do, and also how they could be nonhumanoid.