Brad CostanzoEntrepreneur, AI & Innovation Strategiest

Why I use FOUR AI Agents to Write Like Me (and how they work behind my back)

Name: Why I use FOUR AI Agents to Write Like Me (and how they work behind my back)
Uploaded: 2026-10-19T02:23:18+00:00
Duration: 2 min 7 s
Channel: Brad Costanzo
Description: In this video Brad Costanzo of Accelerated Intelligence explains why using a single AI to write in his voice leads to self‑grading problems and factual…

19 Oct 2026 · 2 min

Watch on YouTube

2:07

In this video Brad Costanzo of Accelerated Intelligence explains why using a single AI to write in his voice leads to self‑grading problems and factual errors. He introduces a simple, code‑free multi‑agent system built from four text‑file agents: Otto (orchestrator), Bradley (writer that mimics Brad’s voice), Echo (voice‑tone checker), and Ray (fact‑checker). Brad walks through the workflow: Otto receives an assignment, passes it to Bradley, who may request quick facts directly or send deeper research requests through Otto to Ray. Bradley drafts the content, then Echo reviews it for tone and style, sending back notes for revision. Once Echo approves, Otto delivers the final draft to Brad for his final input. The separation of duties prevents the writer from grading its own work, reduces editing load, and ensures factual accuracy. Brad emphasizes that each agent is defined by a few paragraphs in a text file—no programming required—and invites viewers to contact him to set up the system in their own businesses.

Chapters

Most people ask one AI to write like them. I get it. I did it too. You build a voice prompt, maybe a whole skill. You paste in your best posts. And what comes back sounds— almost like you. Close enough to bug you. I'm Brad Costanzo, from Accelerated Intelligence. Let me show you a simple way to get AI agents working together. Here's the problem: when one agent writes and checks its own work, it's grading its own homework. It likes what it wrote. So the robot phrases slip through. The made-up facts slip through. And you end up as the editor, every single time. So I stopped asking one agent to do everything. I gave the job to a small team. Otto runs the job. Bradley writes. I named him Bradley on purpose, as close to my name as I could get, because his job is to write like me. Echo checks the voice. Ray digs up the facts. Here's how it works. I give Otto an assignment. Otto hands it to Bradley. If Bradley needs a quick fact, he looks it up himself. If he needs real research, it goes through Otto to Ray. Then Bradley writes the draft and hands it straight to Echo. Echo has one job: find anything that doesn't sound like me. A sentence that runs too long, a line that reads like a press release. She sends it back with notes. Bradley fixes it. Back and forth until Echo signs off. Then it goes to Otto, and Otto brings it to me. Why does this work better? Each agent holds one job, so it does that job well. The writer never grades his own draft. The checker only flags, she never rewrites. And by the time a draft lands on my desk, the easy fixes are done. I spend my time on the part only I can do. And this is simpler than you think. Each agent is just a text file. A few paragraphs of instructions, links to the skills and tools it can use, and a short set of guidelines for how work moves from one agent to the next. That's it! No code, just text files in a folder. If you like this and you want to set it up in your own business, reach out. Email me at brad@acceleratedintelligence.ai. I'll show you where to start.