How business leaders upgrade their strategic planning on a multi-agent platform



Even at its best, strategic planning is a challenge for business leaders. Global uncertainty and volatility risetasks become even more difficult and are essential for business success. When the world is fluid, it’s the organization’s option It is as important as the organization’s resilience. In fact, as research has repeatedly shown, Resilience It is currently a substantial factor in the outperformance of a company.

However, building options for resilience and strategic planning challenges human cognitive (and financial) bandwidth. The seemingly endless future scenario, coupled with our own human prejudices, is colluding to fix our understanding of the future with what we saw in the past. Generating AI (genai) helps overcome this common organizational tendency of entrenched thinking and alleviate the challenges of being human.

One powerful approach to incorporating Genai into strategic planning is to combine it with agent-based modeling (ABM), which is already used to simulate complex and dynamic scenarios. Rather than relying on ABM’s deterministically encoded agents, old LLMSs, constrained by the boundaries of human imagination, can make simulations much more flexible, human-like, (fruitful) unpredictable. Additionally, they can achieve this at part and cost of in-person planning workshops, providing powerful tools to explore a wider range of futures and prepare for the unexpected with more agility.

Promise of LLM-driven behavioral simulations

Genai could potentially charge strategic weapons. Strategic arsenals are (or at least should be) common among large corporations, such as war games and scenario planning. The key is to use this technology to simulate interpersonal or interorganisational dynamics, from boardroom discussions to international competition and engagement with regulators.

Software has been deployed in a similar way for a long time Social and Natural Sciences Use agent-based modeling in which heterogeneous independent agents interact with time. ABM is used to simulate the spread of Infectious diseases or The emergence of codes of conduct In human society. Use of ABMs usually involves time-collecting designs for each heterogeneous agent. This was traditionally composed of a set of hardcoded rules that determine how to respond to input and interact with other agents. Currently, the ability to use dynamic LLM as an agent makes ABM within reach for most companies. (This use of LLMS should not be confused with the 2025 buzzword, Agent AI.)

Helping for this kind of flexible, inexpensive and scalable strategy makes it much easier for businesses of all sizes to actually place. OODA Loop It is often used in military contexts:Observe To adjust the reference scenario based on the cockpit indicator. Orient By identifying strategic options according to the company’s starting point. decide About the most effective options. and Activities Quick and accordingly. “The OODA loop makes organizations better at faster learners.

In fact, our discussion reflects our own experience using the multi-agent LLM simulation platform built by the BCG Henderson Institute. I used this platform to mirror real war games and scenario planning sessions that I’ve led with my clients in the past. As we first saw firsthand, what makes LLM multi-agent simulations so powerful is the two unique features of genai: personification, or the ability to mimic human behavior, its stochasticity, or creativity.

LLMS can role-play in a very human-like way: Research with Stanford Google The release earlier this year suggests that LLM can do it I’ll simulate it The individual personality is closely closely enough to accommodate a particular type of survey with 85% accuracy as an individual themselves. Other studies have shown that LLM can be done when prompted appropriately I’ll replicate Human decision-making patterns in economic experiments, or accurately possible I’ll reproduce it Linguistic patterns and political tendencies observed in the behavior of actual social media users.

Furthermore, LLMS is not a deterministic model. This means that the same inputs do not always yield the same output. Instead, they are probabilistic, blessed with a degree of randomness inherent in the way the output is produced. Probability is the underlying cause of Genai’s so-called “hastisation” or false answers to factual questions, but it also gives them the potential to improve such creativity. These “hatography” can be a positive when it comes to expanding creative simulations and potential future concepts.

Using a modeling platform that incorporates multiple LLMs allows you to have a variety of agents, such as regulators, customers, and competitors. Columbia University researchers discovered Using a multi-agent approach with a single agent approach increases the accuracy of human behavior simulation by approximately 75%.

New fixes for old issues

So how can LLM-driven ABMs help leaders develop and strengthen their strategic planning efforts?

Blind angle detection

Experiences that make high-level executives valuable – industry knowledge, pattern recognition from past crises, relationships built over the years – can also limit the ability to imagine truly disruptive or merely unconventional movements by others in the market. This actually creates blind spots for executives and their organizations, making them vulnerable to confusion and shuts them down to creative solutions. For example, Amazon’s surprising entry into grocery retailing through the 2017 acquisition of Whole Foods must have been incredible and not considered a predicted scenario for current retailers at the time. Such unlikely, but highly consequential possibilities are the most difficult to predict due to human bias, but addressing them can expand your strategic imagination and promote a plan rich with more adaptive options.

These biases and constraints become apparent at all business levels, not just individuals. Similar constraints can affect groups and teams as well. In groups and teams, innate collective drift towards group thinking limits the diversity of ideas. At the organizational level, general cultures can establish a specific path, stance new ideas by the sidelines, leading to institutional inertia. Using AI can help overcome these constraints by allowing organizations to better identify Unknown not clear. If converted to It’s known Unknown, by building flexibility in planning, it makes it easier to preempt such possibilities.

Expand the scope of strategic planning tactics

The costs associated with strategic planning can make it impractical to gather all the stakeholders needed to game out dozens of possible scenarios. Such costs can affect how often businesses gain brain trust, or may reduce the likelihood that these exercises will be carried out across the organization. But the more often an organization can engage in scenario planning and war games, the better they will be prepared to navigate unexpected exogenous shocks into the business.

To complement live strategy sessions, LLM-driven ABMs are easy to use, cheap to deploy, and scalable. As a result, they can promote a broader spread of successful strategic planning tactics (and beyond new organizations) to a wider team of businesses (and beyond new organizations), as well as a more frequent reassessment of strategic planning and decisions.

Promote convergence

ABMs with LLM cannot replace classic, face-to-face strategic sessions, as these face-to-face interactions help leaders unite around a shared strategic orientation. However, these new tools can help you achieve this by building confidence in your organization’s strategic decisions. The simulation can be repeated and brings out a pattern in which a Venn diagram of the commonality of the entire scenario appears. Frequency does not necessarily correspond to probability, but it allows us to pay attention to previously overlooked paths, which helps us to align new strategies.

For example, when comparing the output of a multi-agent ABM with LLM-powered, strategic workshops held with a top life insurance company, the simulations reach three of the same strategic recommendations created by a human-led workshop, allowing company leaders to move forward with more confidence in these decisions. LLM-driven ABM simulation also pointed to two new Options that were not featured during face-to-face workshops, including strengthening workforce training in emotional intelligence and AI literacy.

How to get started and why do it now?

For organizations that want to get started, the first step is to enhance their strategic decision-making process with a multi-agent Genai platform. This doesn’t require you to start from scratch. Leaders should help establish goals, contexts, and dynamics between agents using existing frameworks. It is important to use the platform before and In addition to existing strategic planning sessions, run on scale to find the Venn diagram of the most resilient strategic steps. These results can be used to build consensus on strategic decisions and to harness and maximize future possibilities rather than fearing the uncertainty that comes with them.

Starting now will help make AI a routine input and help organizations adapt to the higher frequency change and adaptation environments informed by real-time learning. Evidence has been shown to suggest that resilience and optionability are more important than in decades. The sooner the company can upgrade its strategic planning and visionary capabilities, it is likely to thrive in a world of increasing uncertainty.

***

read other luck Column by François Canderon.

François Canderon is a partner at private equity firm Seven2 and is a former global director of BCG Henderson Institute.

Leonid Zhukov is the director of BCG x It is based in the AI ​​Science Institute and BCG’s Dubai offices.

Max Struver He is a leading engineer at BCG X and ambassador to the BCG Henderson Institute.

Alan I’m a partner of Boston Consulting Group Director of creativity and scenarios, and Thinking with a new box.

Elton Parker He is a partner in the Boston Consulting Group and is the Associate Director of its Uncertainty Advantage Team.

The author would like to thank Nick D’Intino for his contributions to this article.

Some of the companies mentioned in this column are past or present clients of the author’s employer.

Leave a Reply

Your email address will not be published. Required fields are marked *