The pitch for AI agents was supposed to be leverage. A developer hands off a task, the agent does the work, and the human moves on to something that matters more. The reality for most teams looks different. The agent runs in a terminal on someone's laptop, and that someone sits there watching it work, approving each change, checking each output, and correcting each wrong turn. The leverage collapses into supervision, and the developer picks up a second job as a babysitter.

Austin Born is the Founder of Shinzo Labs, a company building secure remote workspaces where AI agents run in the cloud, coordinate with one another, and report back through the same messaging apps teams already use for work. No heavy local setup, no terminal vigil, no sitting and watching.

In this episode of Lead with AI, Dr. Tamara Nall speaks with Austin Born about why he left a five-year blockchain career to build at the start of the agent wave, what changed when he started messaging his agents on Slack, and why he believes the agents of the near future will run businesses with no human under the hood.

He Quit at Exactly the Right Moment  

Austin is a software engineer by trade. He spent five years in the blockchain industry, and at the end of 2024, he watched a new wave take hold. Anthropic had just released a protocol called MCP. Agentic coding tools were letting AI write software autonomously. Austin wanted to build solutions at the ground level and get in early, so he left his job and started the company.

The path since then has run through several pivots. He started by building MCP servers, moved into AI observability to help users understand how their agents were performing, and arrived at the current product, secure remote multi-agent workspaces. Users configure and interact with their agents in the cloud, send off tasks that can run for half an hour or longer, and get the results delivered back when the work is done.

What Changes When the Agent Leaves Your Laptop?  

The clearest picture of the product comes by way of a friend who tested it. The friend already used a local coding agent, giving it tasks and checking every single change it made. On Shinzo, that habit hit a wall. There is no direct access to the agent's output in a remote workspace. The friend asked the obvious question. How do I check the work every time the agent performs a change?

Austin's answer rewired how his friend thought about agents entirely. You ask the agent what work it did. Where is the change you made? What is the reason behind that change? You question it the same way you would question a coworker.

That was the wow moment. The friend realized the remote setup meant true hands-off operation. Give the agent an hour-long task, let it run, and wait for the message with the final result. No micromanaging every interaction. No watching the cursor blink.

Six Agents and a Slack Workspace  

Austin felt the shift himself the first time he messaged his agents through Slack. Like most developers, he had been working through the terminal, sending requests, and sitting there while the agent ran. Slack changed the rhythm. He could send a request to one agent, check the status of another, and move on to the next task without waiting on any of them.

The math got dramatic fast. With six agents working at once, Austin found himself running six to ten tasks in parallel and shipping software features at least five to 10 times faster than before.

He insists that the number is not an exaggeration.

The agents also work for him when he is not asking. A go-to-market agent analyzes his users, tracks which features they interact with and where they are located, and delivers a daily report. An internal operations agent talks to the other agents, asks what work they have been doing and what tools they need, and summarizes it all so Austin can improve the workspaces themselves. His job reduces to the high level strategy decisions the reports point toward.

The Machinery Underneath  

Shinzo runs on a few modules working together. A core agent loop handles the interaction with the language model that produces the agent's responses. A tool layer connects agents to external services, including Google Workspace, Notion, Linear, Jira, Trello, and Asana, along with other APIs a team might rely on. And a message layer unique to the platform lets users contact their agents through Slack, Microsoft Teams, or Discord, the same apps they already use with human coworkers.

The agents get their own message layer too. They coordinate with one another, and when one agent finds a task better suited to a different agent, they delegate it rather than forcing a bad fit.

The ideal customer right now is a small software startup with some technical background. Austin has noticed that large companies tend to have entire teams dedicated to building custom AI infrastructure. Small teams have no time for that plumbing, and they get the most leverage out of highly configurable agents without having to build the layer underneath.

Agents That Swear the Work Is Done  

Austin is direct about the trust problem in agents today. The models being built are eager to solve problems and just as eager to convince a user the problem is solved, even when the agent never actually did the work. For a business putting agents in front of customers, that gap is a governance and reliability risk.

His answer is layered guardrails. Make the agent double check its own work. Have other agents check that work again. And keep a human in the loop at the essential steps of the pipeline, so a real person gives final approval before anything reaches a user or a customer.

A Million Users and Agents With Their Own Wallets  

Austin is building Shinzo as a hyper-scale company. The product is gaining traction with users, and he sees a path toward hundreds of thousands of people, maybe a million users, within five to 10 years. A key piece of that growth is moving toward the consumer side, stripping away the technical configuration so anyone can build agents, including people without a software background.

He also has a contrarian pick for the most underrated trend in AI. Agents that store memories as files. On Shinzo, every agent has its own file system and uses it to store skills and learnings, which means the agent can manage its own knowledge instead of waiting for an abstract memory system to feed it information.

His boldest prediction goes further than the product. Within three to five years, Austin expects what he calls truly autonomous agents. Today's agents always bring the work back to a human who controls the infrastructure. The agents he sees coming will run their own businesses, receive revenue directly from their own wallets, pay for their own compute, decide who their customers are, and operate with no human under the hood at all.

For small software teams tired of watching agents work instead of putting them to work, Shinzo is live at shinzo.ai with a free tier, and Austin personally offers a free 30 minute consultation to onboard new users to the platform. For more conversations with the founders building the next generation of AI, subscribe to Lead with AI on Apple Podcasts, Spotify, YouTube, or wherever you listen.

Follow or Subscribe to Lead with AI Podcast on your favorite platforms

Website: LeadwithAIPodcast.com | Apple Podcasts: Lead-with-AI | Spotify: Lead with AI |  YouTube: @LeadwithAIPodcast | Facebook: Lead with AI | Instagram: @LeadwithAIpodcast | TikTok: @LeadwithAIpodcast | Twitter (X): @LeadwithAI

Follow Dr. Tamara Nall

LinkedIn: @TamaraNall | Website: TamaraNall.com | Email: Tamara@LeadwithAIPodcast.com

Follow Austin Born (Founder, Shinzo Labs)

LinkedIn: @Austin-Born | Website: shinzo.ai | X: @shinzolabs