This article is my commentary on an article published in Foresight, written by the authors of the book Humachine: Humankind, Machines, and the Future of Enterprise (Sanders & Wood, 2020). For the original Foresight article and other commentary, you can go here.
INTRODUCTION: A NEW ERA
The authors of The Humachine make a valid point that enterprises should not focus on artificial intelligence (AI) alone if they wish to leap forward in automation, productivity, and competitiveness while staying relevant for attracting talent. They elegantly position Kasparov’s Law (the combination of ordinary humans and ordinary machines using the right processes can lead to superior performance, even triumphing over human genius or powerful computers alone) as an addition to any future operating model in enterprises incorporating AI.
An operating model describes how an enterprise delivers value to its customers. It connects strategic directions with operational executing and stays with the vision, values, and behaviour framework for a business. The model contains, but is not limited to, organizational design, people and capabilities, roles and responsibilities, business processes and interactions, KPIs, reward and recognition, technology usage, governance and reporting. Many of these elements will be affected by AI and Kasparov’s Law.
We are entering an era where we change from people doing the work supported by machines to machines doing the work guided by people. The interaction between human and machine will become critical in any future operating model.
AI TO AUTOMATE DECISIONS OF THE KNOWLEDGE WORKER
The authors correctly consider AI to be the centre of value creation in the future enterprise. However, they don’t specify what AI will do. In my view, it is simple: AI is here to automate the knowledge worker.
Automation has been happening for hundreds of years. It has long been focused on improving productivity of boring, repetitive, or dangerous human activities. We now have automated production facilities, warehouses, and transport in our physical supply chain. With robotic process automation (RPA), it entered the back office to simplify repetitive tasks like invoice matching or purchase-order creation and acceptance.
The next step through AI is to automate the knowledge worker, either supporting or automating their cognitive tasks. It will help the knowledge worker make decisions and act if required. AI will enable cognitive automation. AI algorithms will gather, analyze, and interpret data, and make decisions and execute them. It will do so at a higher speed, larger scale, greater consistency, and precision, and with more endurance than any human is ever capable of. It might first augment decisions of a knowledge worker in medicine.
I know a surgeon who uses AI-driven pattern recognition as a second opinion before he decides to remove a polyp. Take this one step further and AI cuts out the polyp by itself. Similarly, AI-driven decision augmentation and automation is fast becoming a reality for the knowledge worker in the enterprise. However, there are limits to how far it can go.
FOCUS OF HUMAN AND AI INTERACTION IN DECISION MAKING
In my Foresight article “Technology Support in Decision Making” (2020), I described six drivers of whether AI will augment or automate decisions: data generation, decision granularity, frequency, complexity, impact, and human centricity. I argued that as the decision horizon lengthens from operational through planning to strategic, automation decreases, and human centricity increases.
This logic can be combined with the authors’ observation, through Moravec’s Paradox, that machines and humans have complementary strengths. These must be understood when deciding when machines should be the lead for decisions, when humans should be the lead, and when to require collaboration between them. I differentiate at least four types of decision making:
Operational During execution and the short-term operational horizon, business decisions are highly frequent, repetitive, and at a low granularity level with mostly small impact. These decisions can be highly automated. Think about a production line, stock movements in a supply chain, and the Amazon policy of changing prices of its millions of products automatically. In the humachine, AI is leading here and only guided by humans.
Planning Planning decisions beyond the operational horizon are less frequent, with higher granularity level, higher impact, and with often more complexity. There is decision time for the human to be augmented by AI while the cost of automating a decision might be excessive. The humachine needs to be highly collaborative in order to evaluate what-if scenarios, risk modeling, and probable outcomes suggested by AI to be decided by humans. This is where Kasparov’s Law is most impactful.
Strategic Strategic decision making is infrequent, at a high granularity level, with a high impact and complexity of relationships and interconnectivities. Examples would be the decision to enter a new market or engage in a merger. In the humachine, the human will lead and act while AI provides some augmentation, but no automation.
Cultural Any business decision that involves values, behaviours, ethics, or virtues needs to be human centric. Defining who you want to be as a business, or understanding the social or cultural aspects of decisions, can only be done when the human is leading in the humachine.
KEY ELEMENTS FOR INTEGRATION OF AI INTO AN OPERATING MODEL
To start the journey towards the humachine and integrate AI in the operating model, a company will have to consider at least the following four elements, in a seamlessly integrated fashion:
Dynamic Data It all starts with data— the petroleum of the 21st century—and AI algorithms are data hungry. Without data you cannot extract value from AI. If you do not own the data in your value chain, it must be purchased or otherwise acquired. Digital natives, companies that have built their enterprise around data, have an early advantage; think Netflix and Amazon. The nondigital native enterprise will have to catch up, requiring connections to dozens or hundreds of internal legacy systems as well as dozens of external sources, maybe even thousands in the case of IoT connections. The data model needs a read capability to feed up-to-date information to the algorithms and writeback capability to underlying legacy systems. A static data lake will not suffice anymore. As I heard a Chief Technology Officer mention: “Data lakes that are not dynamic and can’t operationalize in day to-day business become data graveyards.”
Transparent Science Companies with advanced analytics and AI capabilities tend to employ groups of data scientists working apart from the rest of the organization, possibly for months at a time, before presenting results to stakeholders. Their algorithms need to be transparent, a glass box. Subject-matter experts in any function need to be able to interact with algorithms, change them, tune them, and operationalize them in a user-friendly way. Most companies do not need to own algorithms as intellectual property or trade secrets. There are open-source AI algorithm libraries and languages like R and Python.
Digitised Processes Kasparov’s Law requires a process that works effectively between human and machine. In future operating models, the process will be mostly digitized. We already see the rise of the “digital twin,” the automated process that exactly copies the steps from a human to analyze, decide, and act, whatever the functional area is. Companies must be able to digitize their processes, a vital prerequisite for any operating model that has AI at its core.
In order to enable the humachine and nurture Kasparov’s Law, an enterprise will have to rethink several change elements.
1. Machine-to-human interaction. How will the machine present advice to the human? Will there be an easy to-understand user interface? And how does the machine inform the human of decisions that have been automated without human involvement? These policies need to be transparent, not a black box.
2. Human-to-machine interaction. How will the human provide input and guidance to the machine? People will have to instruct the machine during setup and periodically examine the machine’s decision logs to check and possibly correct alignment with corporate values and strategic goals.
3. Organizational change. How will the enterprise support the humachine? Will the organizational design adapt as fewer humans are charged with increased scope of responsibility? How will recognition and rewards to employees be adjusted to promote collaboration with machines and excellence at human-centric capabilities?
Sanders and Wood are right that a shared vision and purpose is prerequisite to embrace the change needed to transform to the humachine. Digital natives, whose culture is already data centric and AI collaborative, again have a natural advantage. Nondigital natives must drive this new culture from the top down. This is already happening. The CEO of Unilever, a 90-year-old CPG giant with 2.5 billion daily customers, 161,000 employees, and 300 production facilities across 190 countries, has put digital transformation at the heart of its strategy. Marc Engel, its Chief Supply-Chain Officer, has publicly declared that investing in agility, which he defines as quickly sensing change and responding to it, gives in his opinion a 10X return versus investing in forecasting and scenario planning. As a nondigital native, Unilever is becoming a global leader in cognitive automation.
While the right corporate culture can help, much is also dependent on the individual mindset of the employee. One can look at Garry Kasparov for inspiration: he is a highly motivated and extremely competitive individual. Although the world’s best at what he does, he lost to a machine while the world watched and then proclaimed to be the first knowledge worker put out of a job by a machine. But, with some hesitance, he took an interest in AI and the value it can bring when the machine is working together with the human. He opined that the humachine is the best way forward for increased performance, first for chess, then for the wider world and business.
We need individual mindsets like this if the transition to the humachine is to happen. Knowledge workers should take notice.
Photo credit: https://thelogisticsworld.com/