https://xitter.gq/t.co/7zWVK4UG6T CEO and Broadway producer
New York City
Joined July 2006
- Tweets3.3K
- Following1.1K
- Followers4.9K
- Likes707
Brooks’s Law says adding people to a late software project makes it later. Jan van den Berg asks whether it still holds with AI agents: j11g.com/the-mythical-man-mo…
Broadway could have told him.
You don’t open sooner by adding actors in tech week
Rehearsal is sequential. Nine directors won’t get you there in a ninth of the time
The director defines done. Without that, you just reach a bad first preview faster
Nine choreographers will happily restage each other’s numbers all night
Agents are the same show. Know what you’re opening before you add to the cast.
Oren Michels retweeted
AI agent adoption is still very bimodal right now. You have coding and coding adjacent tasks which have taken off, and then everything else.
And even within coding, you have a very wide continuum of adoption patterns, with some likely a small percentage of developers deploying background agents working on projects in parallel, and the rest of everyone else still working with agents 1:1.
Then, there’s the entire rest of knowledge work where real agentic adoption is still very early. The reason for this is that most of the workflows still need to reengineered to work with agents.
This isn’t the same as deploying a chat system for better research efficiency, but instead requires workflows to be rebuilt, data to be wired up in new ways, new practices for accountability and liability of decisions when agents are in the mix, governance and compliance paradigm changes, security upgrades, and more.
We’re still so unbelievably early in what this is going to look like outside of a few categories of work right now. This is why you can basically expect 100X more agent adoption from what we’ve seen so far.
Seems that every day brings another ungoverned-ai-runs-amok story. For people to really use AI, you need connectivity and trust. Trust is more than security - it's the belief that the AI is not going to create a mess that the humans need to clean up. That's why @BarndoorAI exists - come check us out!
Dude, I truly hope you can recover the data.
The shortfalls you called out w/Railway tokens/scopes for CLI and MCP -- this is true across many platforms (god-level tokens).
barndoor.ai has granular access controls that can be applied to any remote MCP server.
Oren Michels retweeted
We are so unbelievably early with agents right now. The majority of companies aren’t even using coding agents at scale, let alone for the rest of knowledge work. We’re still mostly in the chatbot era of work for most of AI right now.
Diffusion of tech takes time, even in the most breakneck of markets, because there are major workflows that need to be reinvented, any regulated or large business has huge governance processes for deploying new tech or agents, data needs to get into well-organized environments, and there’s technical literacy that needs to be established. All things that get solved, but takes time nonetheless.
A point of comparison for technology diffusion: in 2010, a time by which every person in silicon valley knew that cloud was the future, AWS revenue was $500 million, Azure had only launched that year, and GCP was called Google App Engine. By 2025, these 3 platforms generated around $225 billion in revenue. And that’s only about 60% of the cloud market.
So from the moment the tech industry saw the future of cloud to today, the market is nearly 1,000 times bigger. And it’s still growing at an insane rate.
The same will happen for agents. Coding agents are like the early days of cloud computing when developers got on board for initial use cases. Then came the bigger workloads. This gives you a sense for how early we actually are in this transformation.
A couple of times a week, I find myself convincing a CTO to try coding agents. We’re very early.
Someone was telling me that it took over 10 years for most enterprises to adopt the cloud, and there are still holdouts.
We’re only 1 year into AI coding agents.
I do think coding agents will proliferate faster, if only because the competitive advantage is so strong that companies who don’t adopt will struggle to stay afloat.
This is exactly why we are building @BarndoorAI
“We’ve also been moving off legacy systems with poor, slow, outdated, and inconsistent APIs.”
If you’re building software that can’t work fully headlessly in a way that agents want to use, you’re not prepared for what the future of software is going to look like.
Agents will use software 100X more than people, and people will more and more interact with their data and workflows via agents across many different platforms.
This is the real risk but also opportunity for platforms right now. Software doesn’t go away, but it becomes the guardrails and business logic for what agents are able to operate on. But if you can’t connect to wherever the agents want to do that work, you’re DOA.
Absolutely true. And the way you “give it access to your systems” safely and securely is through @BarndoorAI
I've spent the last 3 months building my chief of staff.
Not hiring one. Building one, with AI.
And i think this is going to BREAK corporate America if they can't figure out their BYOA policy... fast.
BYOD (bring your own device) was about hardware. Your phone. Your laptop. But nobody really cared.
BYOA (bring your own agent) is about intelligence. How you think. How you decide. All facilitated by your agent.
The most talented people I know are focused on LEVERAGE and obsessively working on this.
They're not just "using AI". They're building personal agents trained on:
- how THEY think
- how THEY make decisions
- how THEY work
These agents draft their comms. Flag risks. Manage contracts. Analyze data. And they do it all in their voice.
Someday soon, someone's going to walk into a job interview and say:
"This is my agent. Give it access to your systems and I'm productive on day one."
But its going to get incredibly messy if corporate America cant figure out HOW to manage it.
Why?
Because every BYOD policy had to answer: where does the data go? Who owns it? What happens when they leave?
BYOA has to answer all of that PLUS:
→ What does the agent know from their last employer?
→ If the agent makes a bad call, who's liable?
→ When the employee quits, what institutional knowledge walks out with them?
→ How do you "wipe" learned behaviors (not files, behaviors)?
With BYOD, the tool was passive.
With BYOA, the tool reasons. It acts. It carries knowledge between jobs.
That's a completely different problem.
The companies that build BYOA frameworks first will get the best people. Period. But those that just "wing it" will potentially lose trade secrets.
Its a freaking tricky balance.
Because soon, telling a top operator "leave your agent at the door" is like telling a surgeon "you can't bring your own instruments."
Good luck with that.
Enterprises want AI to write to the systems of record their employees use every day. That is where the real value is. Until that happens, AI is largely limited to suggesting actions to humans who still have to do the work.
Once AI moves from advising to acting, the governance model changes. We cannot govern AI the same way we govern humans.
Identity and Access Management (IAM) assumes judgment. Humans have it. AI doesn't.
When an agent updates Salesforce or touches a financial system, it inherits human permissions without human context. If it takes a destructive action, IAM sees a valid identity making an allowed request and permits it. It is understandable that enterprises are wary of trusting AIs to govern themselves.
I wrote an op-ed in @TheAIJournal1 on why identity alone is not sufficient for agentic AI. What is missing is judgment enforced at runtime.
At @BarndoorAI, we focus on the dimensions of control that allow teams to use AI safely and with confidence: defining which actions an agent can take, which tools it can use, what data it can access, how context is scoped for each request, and maintaining a complete audit trail of every action.
AIs are not people. Governing them with tools designed for humans will inevitably create the failures enterprises fear. But keeping AI read-only will not deliver the productivity gains everyone expects. Governance designed specifically for the behavior and risks of AI is the only path to real, sustainable value.
Full piece in AI Journal ↓
Full article here: barndoorai.co/4bJYIoh
Most leaders view AI guardrails as a set of handcuffs.
They’ve got it backward.
In the enterprise, guardrails aren’t there to stop people from doing things—they are there to give them the confidence to start.
Innovation doesn't happen when you give people unlimited freedom. It happens when you make them feel safe failing.
If your team is terrified of breaking a database or hallucinating a legal contract, they won't use the tools.
Build the fences, then let the horses run.
Great talking this through on the Call Me [Data] podcast - link below
#AgenticAI #Governance
The @washingtonpost is worried about AI agents booking the wrong flight.
I’m worried about the "100,000 Agent Problem" inside your enterprise—and the ROI you're leaving on the table. 🧵
5/ The WaPo is right about audit trails and consent. But you need the infrastructure to enforce them in real-time.
The goal isn't just to stop "silent errors"—it's to give agents the green light to actually do the work.
#AgenticAI #EnterpriseAI #BarndoorAI #Governance #AISecurity #AIGovernance #AI
If MCP is getting so much attention, why aren’t more companies actually using it?
The idea is right: connect AI to the tools employees use every day.
The problem shows up at scale. Performance drops, costs rise fast, and teams lose confidence. Too many tools get pulled into context. Tokens are burned before any real work happens. The AI can’t reliably execute.
That’s what keeps AI from being usable across a business.
We built ToolIQ by @Barndoor to make MCP work in real organizations—by controlling which tools are surfaced, when they’re used, and why. The result is AI teams can deploy broadly without exploding costs or operational risk.
More about why this layer is missing from MCP and how ToolIQ unlocks usable, reliable, and cost-controlled AI.
#MCP #EnterpriseAI #AgenticAI
More in the article below... linkedin.com/pulse/barndoor-…
