Why AI Agents Need Human Harnesses | Robert Vis artwork

Why AI Agents Need Human Harnesses | Robert Vis

MTS

October 11, 2026

Bird CEO Robert Vis discusses expanding communication infrastructure to AI agents using MCP and CLI harnesses, cutting headcount from 1,000 to 120 via automation, and executing a $450M debt financing round for dividend recapitalization. Turn ideas into software people love.

0:00

Robert Vis

I would say the new language is MCP or CLI, basically as a way for agents to communicate with each other and for developers to communicate with the agents. And what we've basically done is created a harness that allows for developers and agents to securely be able to send these emails and send these text messages, but keep the ultimate control back to the human. So by basically providing both the full flexibility of agents across all of our comms channels, but still keeping humans in the loop in a safe way.

0:29

Sophia

Hello, everyone, and welcome back to MTS. Today, I am joined by Robert Vis, who is the founder and CEO of Bird. This is formerly Message Bird, and they're building the communication infrastructure for email, SMS, WhatsApp, voice, and now they're opening the infrastructure to AI agents. And last week, their company just secured a $450 million in debt financing. So quite exciting. Robert, welcome to MTS.

0:56

Robert Vis

Hi, Sophia. Nice to meet you.

0:57

Sophia

Welcome. Welcome. Great to meet you as well. I would love to hear a little bit more about the origin story behind Bird. How did you identify that this was a problem that needed solving?

1:07

Robert Vis

So yeah, I founded Bird in 2011, after selling my previous company called Zepe. And basically what happened was, back then, it was very hard to send a message.

Every time you sent your friends a message, Bird always arrived, but then as a business, if you would do that, these messages somehow got lost in translation. And so I started a business around that. And then I think we got very lucky because as apps proliferated on your phone, every app needed a message to be sent to basically verify the identity of a user. And if you think back about our product today with agents, verification is such a large part of it. So it's kind of like full circle 15 years later.

1:46

Sophia

So one thing I want to talk about is we're seeing a rise of agents. I mean, obviously with Muse and Instinct and GrockBot, and now we have Dots in the mix. Everyone's just using these personalized agents and they're throwing a bunch of tasks like book me a new flight or cancel a restaurant reservation. And they have to message with other agents. They have to communicate, they're calling, they're sending emails. But what's broken right now in how they're communicating with each other?

2:12

Robert Vis

Well, look, first of all, I think the assistance are amazing. And I think that's a very good way for consumers to use some of this technology. I think where we focus much more is on the business side. So how do you securely, as a business, make use of AI and make use of all the possibilities to send an email, send a WhatsApp, send a text message, do all these things, and give that full empowerment to agents, but at the same time hold human controls so that they don't go off and send the bad text or anything.

2:41

Sophia

So where would that actually happen? Like how might they go off and send a bad text or what needs to actually be inside the product so that they don't?

2:51

Robert Vis

Yeah. So what we've basically done is we've created. So if you think about like APIs was kind of like for software systems to communicate with each other. I would say if you call that a language, let's call it English. Then I would say the new language is MCP or CLI, basically as a way for agents to communicate with each other and for developers to communicate with agents. And what we've basically done is created a harness that allows for developers and agents to securely be able to send these emails and send these text messages, and give control back to, but keep the ultimate control back to the human. So we verify someone's identity as a business or as a person tied to the agent, and then you can give the agent as much or as little controls as you would like. For example, probably if you were sending a bunch of emails and a email box all of a sudden these messages land in spam, you probably want your agent to recognize this behavior and give you a notification. Whereas if you want to send production credentials like a secret key, you probably don't want to give that to an agent and keep that under human control. So by basically providing both the full flexibility of agents across all of our CommitComms channels, but still keeping humans in the loop in a safe way.

Read the other 14 minutes of this episode here.

$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.

Not what you expected? Full refund within 14 days.

Thousands of transcripts fetched by people building searchable podcast archives

Fetch transcripts with code

The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:

request
curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.

Using your own key:

request
curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000795272869