Onslaught of the smart machines has been saved
How should organizations capitalize on and cope with cognitive technologies in the workplace? Should they simply designate the last person to leave the office, schedule their departure date, and point out the location of the light switch?
If you believe, as I do, that cognitive technologies—those systems that can perform typical knowledge-oriented tasks like digesting and understanding text, answering questions in human languages, and making informed decisions—will soon be coming in force to many different organizations, what should be done about it now?
There are at least three different levels on which action is needed. I’ve focused much of my recent writing1 on what individual knowledge workers should do about this, and that will always be an important component of this problem. One could also focus on getting policymakers to understand this problem and do something about it at the national or societal level. One doesn’t have to be a political scientist, however, to realize the challenges at this second level with regard to our current political institutions.
The third level of action, of course, is organizations. What can they do to better prepare for the onslaught of smart machines? What will it do to their strategies? How will their people react and cope? Should we simply designate the last person to leave the office or the plant, schedule their departure date, and point out the location of the light switch?
I believe there are four key steps that organizations need to take with regard to these technologies.
Know what can be done. As with big data or e-commerce, the first thing a management team should do about a new technology is to learn what it is capable of. What are the key elements of cognitive technologies? How might they affect marketing, customer service, and R&D? What kinds of applications exist to support these different business functions? And just how good are the decisions made in these areas?
New vendors are popping up every day in the automation space. There may well be some that target some aspect of your industry. If you’re in financial services, for example, your management team should be familiar with the cognitive technology options in consumer financial advising (sometimes called “robo-advisors”), portfolio optimization, trading, and fund accounting. Companies can either buy or develop automated or semi-automated solutions in each of these areas.
If you’re a large organization, you probably already have some kind of technology monitoring and assessment capability. Make sure that cognitive systems are on the list of technologies that are regularly followed. You may even want to commission a special project to assess opportunities in this area.
Decide what should be done. The next step is to develop a strategy about the specific fit of smart machines with your organization. These systems have long-term impacts on employee capabilities, competitive strengths and weaknesses, and even customer relationships. Think about a situation, for example, where everyone in your industry had the same level of automated expertise about a key business domain, or one in which everyone offered the same advice to customers.
In this process your organization can identify the most pressing business domains for an automation initiative, and those that could be explored further down the road. This discussion should almost certainly go beyond an IT organization to involve key business leaders. And if the size of the opportunity—and the anticipated costs—are substantial, it may be advisable to involve the board of directors.
Assess how to do it. At this stage an organization needs to select the specific technologies and vendors it will work with on its first project or proof of concept. If your company is truly committed to cognitive technologies, you can even begin to build a broader platform than can support multiple applications.
But be aware that there are many different types of cognitive technology, and some work better than others for specific applications. If you have quantitative data to support a decision, that will involve different technologies—perhaps neural networks or machine learning—than if you have textual information. If the knowledge you are embedding in a project is relatively static, rule-based systems are probably the best approach. If it changes rapidly as new data comes in, you may need technologies that continuously learn and improve over time.
Determine the impact on people. Obviously these technologies will have an impact on knowledge workers in your organization. Instead of simply automating them out of a job, I would encourage you to think about how smart machines and smart people can augment each other. Some people may work closely with the technology, monitoring its performance and dealing with exceptions it can’t handle. Others may “step up” above the technology to examine its broader impact on the business—portfolio management in asset management firms is an example of this. Computers handle the day-to-day trading, but humans develop and monitor the portfolio strategy. And, of course, humans need to build and configure the systems when they are initially implemented.
It’s only fair to your employees to give them some sense of when they might be threatened by automation, and how they might begin to acquire the skills to coexist with it. One insurance company, for example, found that a third of its existing underwriters could work directly with the automated underwriting system in the jobs I’ve described above. Another third could move into communications roles with agents and customers. Unfortunately, a final third had neither of these skills and had to leave the company. But at least the majority kept their jobs, and those who didn’t immediately have the needed skills had some time to acquire them.
The time is right to think seriously about these issues. However, be careful about announcing your automation projects to the world at large. The development of automated systems takes years in many cases, and not all projects work out. And why would you want to tell your competitors what you are doing with cognitive technologies? Your plans and deals with vendors should be in internal strategy documents and contracts, not in press releases.