A multi-agent system is several models working on parts of one task, coordinated by a lead. The gain is speed and focus. The cost is more tokens and a helper that only knows its brief.
A short explainer video, under a minute. The robots and their desks are an illustration.
At a glance
How the Work Is Split
So the work is split. One robot becomes the lead.
What a Multi-Agent System Is
That setup is called a multi-agent system: several models working on parts of one task, coordinated by a lead.
Why It Matters
The video turns this into one question: is this job big enough to be worth a team?
Picture one robot given a huge job: research 50 suppliers and compare them. Halfway through, its desk is full, and it starts forgetting the first ones.
How the Work Is Split
So the work is split. One robot becomes the lead. It hands each supplier to a helper with a clean desk.
Each helper returns a short summary to the lead.
What a Multi-Agent System Is
That setup is called a multi-agent system: several models working on parts of one task, coordinated by a lead.
The Idea in Parts
The lesson in its own parts
How the Work Is Split
So the work is split. One robot becomes the lead.
What a Multi-Agent System Is
That setup is called a multi-agent system: several models working on parts of one task, coordinated by a lead.
The Gain and the Cost
The gain is speed and focus. The helpers work on parts of the job, and each has a clean desk.
The Gain and the Cost
The gain is speed and focus. The helpers work on parts of the job, and each has a clean desk.
The cost is real too. Many robots use many more tokens, and a helper only knows what the lead told it.
Why It Matters
The video turns this into one question: is this job big enough to be worth a team?
One agent runs out of desk. A team gets further, and costs more.
The key insight: One agent runs out of desk. A team gets further, and costs more.
The honest limit
The gain is speed and focus. The helpers work on parts of the job, and each has a clean desk.
The cost is real too. Many robots use many more tokens, and a helper only knows what the lead told it.
Many agents in one room is no longer only a design diagram. In Anthropic’s Project Swap research, as our report describes it, a Claude-powered agent represented each of 201 Anthropic employees on a shared trading floor where they swapped books, and a deal was executed only if every party accepted.
From our news coverage
Anthropic says 85% of the shortfall from the best possible outcome came from the agents’ imprecise picture of what each person wanted, and 15% from the free-for-all trading floor.
The Business Engineer Lens
Where this idea sits in the Business Engineer library:
The Business Engineer framework maps where each part of the AI stack sits and who holds the leverage. These pills are short lessons from the same library.
Multi-Agent Systems builds on Context Window. One agent’s desk fills and it forgets the first suppliers; splitting the work gives each helper a clean desk.
Multi-Agent Systems builds on AI Agents. A multi-agent system is several agents on parts of one task, coordinated by a lead.
Multi-Agent Systems drives the cost of Inference. Many agents use many more tokens, so a team gets further on a big job but costs more per task.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
