Skip to main content

​
The Power of Parallelization

In a traditional workflow, tasks are handled sequentially. An analyst can only review one valuation report at a time, or run one forecast scenario at a time. This creates a linear, and often slow, path to completing complex projects. Agent Swarms fundamentally change this paradigm by introducing work parallelization. Instead of being limited to a single instance of Kelso, you can deploy a “swarm” of multiple agents to execute tasks concurrently. This capability transforms your potential output, allowing you to accomplish in minutes what might have previously taken hours or days.

​
How Agent Swarms Work

There are two primary ways to leverage Agent Swarms, both designed to dramatically increase your efficiency.

​
1. Parallelizing One Workflow

This is the most common use case: applying a single, complex workflow to many different inputs at the same time. Instead of a one-to-one relationship between you and an agent, you become an overseer of a team of agents all performing the same task in parallel. Example: You have ten valuation reports from ten different clients that all need a standard mathematical audit. Instead of opening each one and prompting Kelso ten separate times, you can simply instruct the swarm:
“Deploying a swarm of 10 agents. Each agent is to perform a full mathematical audit on one of the valuation reports in the [Client Reports Q3] folder and notify me upon completion.”

​
2. Parallelizing Many Unrelated Tasks

You can also deploy a swarm to handle multiple, distinct tasks simultaneously, allowing you to gather diverse information and accelerate your research process. Example: You are preparing for a client strategy meeting and need to assemble information from several sources. You can deploy a swarm to work on these unrelated tasks concurrently:
  • Agent 1: “Summarize the section on [Distribution Policy] from the client’s Plan Document.”
  • Agent 2: “Pull the historical account balances for all participants who terminated in the last three years.”
  • Agent 3: “Query the 5500 database for the three largest ESOPs in the same industry and state as our client.”
While each task is different, the swarm executes them in parallel, gathering all the necessary information for your meeting in a fraction of the time it would take to do it sequentially. This allows you, the human operator, to focus on the high-level strategic thinking the meeting requires, rather than the low-level data gathering.