As technological change keeps moving faster, markets fragment into ever more segments, and mistaken corporate initiatives continue to hurt shareholder value, executives need better decision making tools. For several years now, we’ve been hearing about agent-based simulations as an advanced, high-powered method for conducting experiments, exploring options, reducing risk, and optimizing both internal and customer-facing processes. In this article, Mark Buchanan brings us up to date on how this technology is being put to work by organizations including Nasdaq, Post Danmark, and major pharmaceutical companies.

Business leaders have always needed skills in seeing through the thicket of complexity in order to grasp the core issues of business decisions. But the degree of complexity and uncertainty facing today’s executives often exceeds the capacity of the unaided human mind. Driving the growth in confusing choices and unclear options are factors that include large, multifaceted organizations, global markets and workforces, intricate supply chains, extensive partnership networks, shifting consumer demands, more customized products and services and technologies that evolve and cross-fertilize. The limits to cognition push some leaders to fall back on intuition, or “gut feel”.

Agent-based simulations, according to Robert Axelrod, represent a third way of doing science. Neither inductive nor deductive, science through simulation explores logic experimentally. These technology-enhanced thought experiments show which outcomes are likely to emerge given certain starting points. Cisco Systems, Nokia, Capital One, and Boeing have used simulations to model and prototype product and process innovations without incurring the costs and risks of real-world experiments. Buchanan explains that simulations can also be used to resolve conflicts, and to boost innovation and creativity.

Although Buchanan’s piece does a good job of telling us about some of the latest uses of agent-based simulation, he leaves aside their value in management training and other human capital development processes. We would also benefit from a clearer idea of when simulations are appropriate for a problem, and when they are too costly or poorly-suited. Buchanan does make some comments in this area, noting that simulations typically cost at least $100,000, going up to $500,000 for Nasdaq’s. He also explains, though quite abstractly, how a threshold known as “high computational complexity” can help us decide when simulation should be used.