We Have Been Automating with Digital for Some Time

The way in which Narrative Science or Automated Insights is automating journalism and other writing-based activities such as stock-performance reporting is very sophisticated and represents a new frontier of robotics. However, again, let's not forget we've been consuming automation for a long time, and much as with AI, once used it's not even noticed.

Consider the last time you went to the airport. During your drive, you likely passed through an E-ZPass tollbooth on the highway. Pulling into the airport parking lot, a machine dispensed your parking pass. In the terminal, you received your ticket and checked your bags at a kiosk. Then walking to your gate, you stopped at the ATM to get cash for your trip.

Some two decades ago, people were working at each of those spots: the tollbooth attendant, the parking garage attendant, the airline clerk, and the bank teller. Yet, many (but not all) of those jobs have been automated away (others have been enhanced with technology, which we'll cover later), and as a consumer you are probably grateful.

From your house to the airport gate, your trip was at least a half-hour faster than in pre-automation days. And is anybody really nostalgic about the tollbooth attendant, that poor soul who had to stand in that cold metal box for eight hours, making change, breathing in exhaust, all the while creating long lines for the rest of us? Not even the ex-tollbooth attendant wants that job to come back.

With the new machine, we are now crossing a line to making “intelligent” processes truly intelligent. The new waves of automation are popping up all around us, yet unlike the EZPass scanners over the highway, they are both so subtle and powerful (as in robo-journalism) that they may even go unnoticed.

Set Your 25%–25% Automation Imperative

When you walk into work, (nearly) everyone has a purpose. Things hum along pretty much the same way, every day. On any given Monday morning, it's almost impossible to imagine work being done differently. But this is exactly what you need to do first. To begin any meaningful change using automation, you must suspend disbelief and imagine that things can be done in an entirely new way.

We think your baseline expectation should be cost reductions of 25%, with an associated productivity increase of 25%. Based on where the current average is today (around 15%), and the productivity improvement seen by some solutions (up to 90%), this should be your achievable near-term rule of thumb for initial robotic process automation efforts.

The reason the 25%–25% rule is effective is that it forces your team to think differently about “the way things are done around here.” When the goals are, say, 5% cost savings or 5% productivity gains, your team will continue to think in traditional ways. By raising the bar to 25%–25%, it will quickly become clear that the traditional cocktail of reorganization, outsourcing, and/or enterprise software will not be good enough. Only the digital automation of the process will deliver such results. Put simply, if you cannot gain at least 25% cost savings, you don't have a real “automation” solution. And if you cannot find 25% productivity increases, then the AI platform is not performing as it should.

In addition, you need to create a “no-excuses” environment. There will be those who push back, saying—for various reasons—“that can't be done.” Yet it's becoming clear there are no true hurdles with technology, cost, or scale. Once you set that goal, it's time to pick your spots.

Find Your Process-Automation Targets

When looking for your initial automation opportunities, which should be no more than two or three processes, you should “pick your spots” on the human-machine work continuum (see Figure 7.5). The best automation starting points are to the left of the continuum; in these processes, the machine can rapidly take over the majority of the work effort. As we described in Chapter 3, these should be areas with a high percentage of work tasks that can be automated.

You need to be thoughtful when selecting your first automation targets, because the success, or lack thereof, of these initial implementations will dictate your subsequent enterprise AI initiatives. In our work with dozens of intelligent automation initiatives, we've seen that the successful ones all meet these simple criteria:

• Highly repetitive tasks: Find tasks that are highly repetitive but that occur at great scale within your organization. In short, look for activities that lots of people do every day. It could be processing an invoice, proofreading documentation, “picking and packing,” reconciling orders, answering the same questions over and over on the phone, and so on. Many of these tasks are already heavily automated; but many, in our observation, are not. Give yourself an honest answer: Are these areas—and many more—as fully automated as they could be? And is there intelligence in the current process? Any high-volume, highly repetitive task is a prime candidate for automation.
• Tasks with low demand for human judgment: Robots are great at calculation; humans less so. Conversely, humans are better at complex judgments, while AI and algorithms are not. Thus, jobs that lean heavily on decision trees (as opposed to ambiguity, intuition, insight, complex judgment) are strong candidates for automation. Machines may be able to inform decisions, but making more nuanced judgment calls will for the foreseeable future require a human touch. Conversely, any job that's mostly a series of “if-this-then-that” steps should be automated.
• Tasks requiring low levels of empathy: Order entry, claims adjudication, and invoice reconciliation are processes that require accuracy, consistency, and speed…but rarely empathy. If a bot can materially boost accuracy, consistency, and speed, many of us will settle for dumping the empathy factor. There are many areas within your organization right now to which that logic can be applied. It won't be hard to find them.
• Tasks generating and handling high volumes of data: Any process with the potential for generating lots of data, particularly customer data, should be automated for the sole purpose of collecting that data—your proprietary raw material. Humans, even if you can afford enough of them, will not be able to deal with the volumes of data when more and more “things” become smart and connected. As one simple example, imagine when all the street lights in a medium-size city are “smart” and able to report on not only their own “health” (i.e., whether they need replacing) but can also report on vehicle traffic flows. The volumes of data generated by this type of city-wide information “mesh” will be far beyond the human capacity to manage and make sense of it. In scenarios like this, automation won't replace people but will be a fundamental building block of creating new services and new insights that in turn will generate new value. In your organization, you should look for areas in which processes and workflows can be sensor-enabled and from which transaction data, previously unavailable, can be generated and managed.

Identifying your automation targets will give your teams a clear path to success, but there still remains a significant hurdle in managing change within your organization. That leads to our third rule of the road.

Break Through the “Brass Wall”

OK, you may have identified your initial automation targets, which are being pursued across your industry and clearly pass the 25%–25% threshold. Everyone in your company will see the wisdom of your decision and will be fully supportive, right? Wrong.

We learned this lesson 20 years ago during the aforementioned business process reengineering (BPR) wave. Back in the 1990s, acolytes of Michael Hammer proclaimed, “Don't automate. Obliterate!” The mantra was to find ways to deliver more with less—way less.

BPR, in theory, was very sound. Yet, in most cases, the practice failed (widely soiling the name of the movement in the process). Why? Quite often, the very middle managers who were responsible for architecting and implementing BPR initiatives soon figured out they were reengineering themselves out of a job. Politics and sabotage ensued.

In hindsight, this was the “brass wall” phenomenon, which was first seen with reengineering efforts within police forces of major cities in the United States during the 1980s. During the explosion of drug-trade-related violence, police chiefs in New York, Boston, Miami, and Los Angeles all looked to establish new ways of policing (e.g., more cops walking the beat, following the “broken windows” philosophy of addressing even the smallest of crimes, leveraging data for smarter policing, and flattening the organization model to remove bureaucracy and cost). The people at the top of the policing pyramid (e.g., the police chief and his or her team, as well as municipal leaders) were hugely supportive, looking to drive the change. Young police officers, living with the frustration of the existing system, were largely supportive of these new policing techniques, as well. Yet, such change efforts were initially halted by the “brass wall”—the officers sitting in the middle of the pyramid (those with brass on their shoulders), who were at least a couple of decades into their careers, had earned their way into seniority and the power structure, and could begin to smell their pensions. To them, the perceived personal threat of such changes greatly outweighed any potential community benefits.

This brass-wall phenomenon is reappearing in today's corporate environment with digital automation initiatives. Senior employees may recognize, “Yes, automation is good for our customers and investors, but is it good for me?” Rarely are their comments so direct. Instead, you will be hit with issues, such as “That's impossible.” “It's too expensive.” “We'll get in trouble with regulators.” “We've done it this way for 50 years, and it's working fine.” “Nobody can prove the ROI today.” “We should focus our efforts elsewhere.”

This is a tough problem to solve, but solve it you must. We recommend identifying the most vocal proponents of “business as usual,” talking to them, and explaining to them the reality of what automation means for the organization. Invite them to “get with the program” and help their colleagues get with it too. If they can't, avoid them and have them focus their energies elsewhere. And if there are no other opportunities available in other parts of your organization, let them go (of course following all the proper HR protocols). Achieving the goals of automation is too important to allow failure to happen due to the intransigence of employees whose skills and experience are no longer a good fit for work infused with systems of intelligence.

As we stated earlier, there will be blood; automation will hurt. Nobody said this would be easy. But automation is a fact of life, and facts of life need to be faced.

Build a Repeatable Process to Obliterate Work

Once you've set bold goals for making automation happen and have chosen solid targets that will lead to business impact, the next step is to convert work from “Then Margaret in procurement does this” to “Beep. Done.” This is how you go about doing that.

In every case of successful automation we've seen, a creative business leader looked at how a process was being executed; concluded “We really stink at this”; and found a better way of using code, artificial intelligence, and algorithms to improve the speed, accuracy, and cost profile of the process.

Additionally, these leaders realized they need to apply the new tools, which may be fairly simple or incredibly sophisticated, to automate very specific process steps. Regardless of whether you are re-creating a business process from scratch or injecting a tried-and-true automation technology into middle- or back-office processes, there is a path you should follow.

You will need to adapt the path depending on the complexity of what you're doing; what follows is a high-level walkthrough, but the seven steps for automating any process or task are basically the same:

1. Set your automation strategy. As discussed previously, the first step is setting a bold vision for applying the new machines to reduce costs. You simply can't skip this one.
2. Start small. In spite of the large potential benefits of automation, it's a best practice to start small. Begin by finding a few pressure points in key processes. General proclamations of “do automation” are a waste of time. Every successful automation example we've seen has been focused on a specific task or business process. What is a “process pressure point?” Typically, it's a bottleneck that frustrates everybody today (e.g., payments, mortgage loan applications, medical records management, travel planning, etc.).
3. Apply the new machines. This is where “market sensing” will pay off. (We have more on this in Chapter 10 on abundance.) Identify the relevant automation tools that already exist and test them out in a limited trial. Of course, if there are no commercially available tools for the process you are looking to automate, you may need to build your own AI platform.¹² This will certainly be more complicated; look to buy before you build.
4. Develop a prototype. You have a strategy, some process targets, and a few potential automation machines. It's time now to put the pieces together, but you need to start with a prototype. This phase includes blueprinting and the first rough build of your automation engine applied to your specific context. You'll need the data, the machine, and a clear understanding of what happens before, during, and after the automation activity. When an automation tool already exists, this may require only minor configuration. If, on the other hand, you're building a new engine, this can take some time. The output of this step is a prototype that more or less works in a controlled environment. There will be bugs, and there'll be plenty of people saying it doesn't work, but you need to exit this step with enough of a working system to warrant moving to a production environment.
5. Pilot and scale. Now it's time for real work to be affected. As with any big change, discretion is the better part of valor, so start small—for example, a subset of the work, a single store, some customer interactions. All the questions around security, privacy, and compliance must be resolved, and people's experiences should be assessed to avoid heading into a knowledge-process industrial accident.
6. Analyze the results. What has the pilot told you? What have you learned? What went well? What didn't go so well? Be honest; be critical. Look for reasons to say, “No, this isn't working.” Pivot; adjust; recalibrate. Stay resolute.
7. Repeat! The requirement to automate is a constant, so no single project will ensure success in the digital economy.

These seven steps should be the foundation for reaching the next level of automation in your organization.¹³