Introduction

Consulting companies like McKinsey, Boston Consulting Group (BCG), and Deloitte occupy a unique and often controversial space in the business world. Hired as external advisors, their role is to solve complex problems, from crafting corporate strategy to managing massive technology implementations. While often maligned for their high fees and jargon-filled presentations, they serve an essential function: they are one of the primary engines of knowledge dissemination in the global economy. By taking best practices and innovative ideas learned from one client and cross-pollinating them to others, they help diffuse new ways of working that many companies would struggle to develop on their own.

A consulting firm's primary asset has always been its knowledge, and historically, this was managed in one of two ways. Some firms relied on the deep tacit knowledge of their senior partners—the "gurus" whose decades of experience and intuition were the main draw. In this model, knowledge is transferred through apprenticeship. Other firms focused on codifying their learnings into vast digital repositories of explicit knowledge—case studies, frameworks, and project documents that any consultant could access. This was a form of knowledge banking, creating a library of corporate wisdom.

Generative AI is a disruptive force that transforms both of these models. It acts as the ultimate knowledge bank, capable of ingesting a firm's entire digital repository and allowing any consultant to query it with natural language. This democratizes access to explicit knowledge on an unprecedented scale. More profoundly, AI can also begin to simulate tacit knowledge by synthesizing decades of project data to uncover patterns and generate novel strategic insights—ideas that no single human guru could ever see on their own.

Recognizing this pivotal shift, the world's top consulting firms are pursuing a dual AI strategy. First, they are building their own proprietary AI tools to augment their workforce and create a competitive advantage. McKinsey developed Lilli, an AI chat interface that serves as a research assistant for its consultants. Boston Consulting Group created Dexter, Accenture launched DiPa, and EY rolled out EYQ. Second, and more lucratively, they are all heavily investing in building out AI consulting practices to advise their clients on this transformation. This has created the largest new market for consulting services in a generation. AI is simultaneously threatening the traditional value of the human consultant while providing them with their most powerful tool and biggest business opportunity ever.

The following table summarizes some of the proprietary AI tools developed by leading consulting firms to manage their internal knowledge and augment their consultants. I encourage you to read more about them.

Startups explicitly targeting this industry using Gen AI have merged such as the product touted as McKinsey in a Box: https://share.google/nP2CvKAy95DZTe6u5

Discussion Questions

1. The introduction describes two traditional knowledge management models: one based on specialists with tacit knowledge and another on repositories of explicit knowledge. How does a generative AI tool like McKinsey's "Lilli" blur the lines between these two models?
2. Consulting firms are pursuing a dual strategy: using AI internally to improve their own efficiency and selling AI advisory services externally. Which of these two strategies do you believe represents the bigger long-term opportunity for them? Which carries more risk?
3. A traditional role of consultants was to "disseminate knowledge." If a client can now use a powerful public AI to access global best practices and analyze their own data, what is the new value proposition of a human consultant? What can they offer that the AI cannot?
4. The major consulting firms have all chosen to 'Build' their own proprietary AI tools (Lilli, Dexter, etc.) on top of foundational models, as detailed in the "Build, Buy, or Borrow" framework (Chapter 7). Why did they choose this path instead of simply having their consultants use off-the-shelf tools like ChatGPT? What competitive advantage does a custom-built tool provide?
5. Consider a consultant using a tool like BCG's "Dexter" to help them solve a client's problem. Which of the four 'Archetypes of AI-Enabled Change' (Chapter 10) does this relationship best represent? Is the AI a 'Co-Pilot,' an 'Automated Specialist,' or something else? Justify your answer.

<aside> 💾

Augmenting Expertise with NotebookLM

Google's NotebookLM is a prime example of a new class of AI tools designed not to replace expertise, but to amplify it. I experienced this firsthand while working on a recent policy initiative with my client, Nigeria's National Information Technology Development Agency (NITDA). The task required a deep understanding of all their past policy documents—a library spanning years of work.

Instead of spending weeks manually reading and cross-referencing hundreds of pages, I uploaded the entire corpus of documents into NotebookLM, creating a private, intelligent knowledge base grounded only in my provided sources. From there, I could simply ask it complex questions in natural language, such as, "What were the key arguments made for data localization in the 2023 framework draft?" or "Summarize all policies related to public-private partnerships in cloud infrastructure."

NotebookLM would instantly synthesize the information and provide answers with direct citations and links to the relevant source documents for verification. This didn't just speed up my review process from weeks to hours; it allowed for a deeper level of analysis, revealing connections between disparate documents that would have been nearly impossible to spot manually. This is the new reality of knowledge work: the expert's value is no longer just in finding the information, but in knowing what questions to ask of an AI that has already read everything.

You can check out this NotebookLM Notebook and ask it questions here (login to Google account required first): https://notebooklm.google.com/notebook/c22d4127-455a-4fa9-bd9c-23d03f19ec9b

</aside>