The era of using a single, generic AI chat for all your work is officially dead.
We often think of AI as a single assistant we chat with, but the biggest players in the business world are rewriting the rules entirely. I recently found a compelling breakdown by a Reddit contributor who analyzed Boston Consulting Group’s (BCG) decision to deploy 36,000 custom GPTs for their 32,000 consultants. That is more than one AI model per human employee! The original poster explains that this isn’t just a tech flex; it is a fundamental shift in how businesses operate. McKinsey is doing the exact same thing, aiming for 45,000 agents by the end of the year.
The Logic Behind the Numbers
At first glance, building tens of thousands of bots sounds excessive. However, the author makes a great point about why this massive number makes perfect sense. If you run a global consulting firm, you handle thousands of different projects every year. Each project has unique data, unique goals, and unique constraints.
Instead of forcing a single, generic AI to understand every client context from scratch every time, they build a dedicated AI for each specific engagement. If a complex project needs a pricing expert, a market researcher, and a strategy writer, that is three distinct bots right there. Multiply that by thousands of projects, and suddenly, 36,000 bots seems like a conservative number.
The expert emphasizes that this signals a change in mindset: AI is no longer just a tool you open in a browser tab. It is becoming the infrastructure of knowledge work itself. The goal is to stop treating AI as a calculator and start treating it as a specialized colleague.
💡 Specialization is King
The post’s author highlights that BCG isn’t just giving people access to a standard chat interface. They are building role-specific tools. Think about the difference between a general handyman and a master electrician. One is good for quick fixes; the other is necessary for complex systems. BCG is building the master electricians of AI.
They create GPTs specifically trained on internal frameworks for strategy, operations, or marketing. This means the AI already “knows” how the company thinks. It doesn’t need to be taught the basics of the company’s methodology every time a new consultant logs in. By grounding these bots in real frameworks, which is essentially the company’s intellectual property, they reduce errors and ensure that the output actually sounds like it came from the company, not a generic robot.
✅ The “Must Have” Fundamentals
There is a lot of noise right now about “autonomous agents” that can do everything without human help. However, this industry pro points out that BCG focused on the practical “must haves” first. They didn’t wait for sci-fi technology; they built what works today.
The core value lies in memory and reusability. In most companies, if you have a great conversation with an AI, that value is trapped in your chat history. No one else learns from it. The BCG model changes this. By building a custom GPT, the instructions and “memory” of that project are saved.
A team member can leave the project, and a new one can join, and the AI, with all its context, is still there, ready to assist. It turns fleeting chats into permanent assets. This prevents knowledge silos where one person has a good prompt and no one else sees it.
📌 Scaling is the Real Hurdle
Once you accept that you need thousands of bots, you run into a new problem: How do you actually make them all? The Reddit contributor notes that the challenge shifts from “how do I prompt?” to “how do I manage this fleet?”
Creating, updating, and assigning 36,000 distinct AI models is a logistical nightmare if you do it manually. The author suggests that tools like GPT Generator Premium are becoming necessary simply to handle the volume. You need a system that allows you to clone a successful bot, tweak it slightly for a new client, and deploy it instantly.
This is the “operating system” approach. You aren’t just experimenting; you are building a factory that produces intelligence on demand. The bottleneck isn’t the intelligence of the model anymore; it’s the management of the assets.
How to Replicate the BCG Model
You probably don’t need 36,000 bots tomorrow, but you can apply this logic immediately. The original poster offers a smart “entry point” for smaller teams to start building this muscle.
Start Small: Don’t try to overhaul your whole company at once. Pick one critical role or one repetitive project type.
Build One Bot: Create a robust custom GPT for that specific task. Give it your best instructions and frameworks.
Test and Clone: Run it for a few weeks. If it works, clone it. Then, refine it for a different team or project.
The goal isn’t to hit a high number; it is to build organizational capability. The winners won’t be the ones with the best tech in five years; they will be the ones who started organizing their knowledge into AI today.
If you want to understand the future of work, you need to read the full breakdown from this insightful creator.
💡 FAQ & Troubleshooting
Why does the number of GPTs (36,000) exceed the number of consultants?
The high volume of GPTs is driven by project requirements rather than a simple 1-to-1 ratio per employee. In this operational model, every client project requires at least one dedicated GPT, while complex engagements often utilize 3–5 specialized agents (e.g., specific agents for strategy, pricing, or research). When a firm runs thousands of projects annually, the number of required custom instances naturally exceeds the headcount.
Are these “Custom GPTs” just scoped versions of standard models?
Yes. Technically, these are base models (like OpenAI’s) that have been scoped and constrained using specific parameters. Unlike a general chatbot, these custom GPTs are trained on internal frameworks, proprietary methodologies, and specific roles. This transforms them from a general tool into a specialized asset with persistent instructions and shared project memory.
How does this approach change employee workflow and training?
Deploying role-specific GPTs serves to standardize output across the organization. It allows employees to be flexibly placed into various positions with minimal adjustment time. By consulting the specific GPT assigned to a task, a consultant can apply “learning by doing,” using the AI’s embedded framework to guide their execution immediately, rather than relying solely on prior memorization.
What is the recommended strategy for scaling GPTs in a smaller organization?
Do not attempt to deploy thousands of agents immediately. The effective strategy is to “clone, refine, and scale.” Start with one critical role and one well-defined framework for a single pilot project. Once the value is proven, use tools capable of managing large numbers of custom GPTs to replicate that success across other teams.
Boston Consulting Group (BCG) has announced the internal deployment of more than 36,000 custom GPTs for its 32,000 consultants worldwide.
byu/abdehakim02 in