Don't onboard your team into AI. Onboard AI into your team.
Your team already knows the goals, the work, and what good looks like. Your job is to bring your agents up to speed in the same way.
Doing this well leans on two muscles most managers never had to build.
1. Map your business: most operators work on instinct and never stop to make the process explicit.
2. System design: given a job to be done, what are the steps, and who or what handles each one.
You can't hand a process to an agent until you can describe it.
What turns a generic chatbot into a teammate that actually knows your project?
In this clip, we build it live, starting from a blank agent and adding one layer at a time.
1. A bare agent knows the world, nothing about you. It's a new hire on day one. Smart, but zero context on your project, so a status update request gets you nothing.
2. What you give it access to shapes what it thinks it knows. That's on you, not the model. The real risk is the old habit of trusting a confident answer no matter what.
3. Tools give it reach. Turn on Slack and it goes and fetches real context, walking into your actual channel to pull what's there.
4. Fetching isn't remembering. Every new chat is a fresh intern walking through the door, reconstructing everything from scratch with no memory of yesterday.
5. A brain fixes that. A readme, status updates, research, the decisions you want to carry forward. Your project's notebook, made persistent & readable.
6. The real level-up is shared context. Add a teammate to the workspace and what one person did in Claude yesterday shows up in another person's today.
Stack the layers and a generic chatbot turns into a teammate that grows.
Most AI QA is theater. A dashboard, a green checkmark, a feeling that it's working.
Before you can grade an agent, you write the syllabus: the real journeys people put it through.
Go outside of the happy path.
The request that times out.
The user who flips their preferences halfway through.
The request that breaks a rule you've set.
We started with 35 journeys built by hand, in a spreadsheet. It turned out to be the most important thing we did.
Most teams are grading a test without writing the questions.
If the agent's answer looks right and the grade says pass, why would you ever look under the hood?
What the user saw:
"Done. I've saved your window seat preference."
What the trace showed:
→ save_preference() called
→ response: 200 OK
→ database write: none
The agent fired the tool, got a clean response, and reported success. But the preference was never saved. No error fired because, as far as the agent could tell, nothing went wrong.
Only the trace tells the full story.
How can you trust an agent when the same question gets a different answer on a different day?
AI doesn't hold still. One prompt change breaks something you signed off on three weeks ago. Nothing else touched. Same question. Different answer.
4 questions for building AI you can actually trust:
1. What's on the test? You can't grade a test you haven't written. Start with the real user journeys, edge cases that actually break things.
2. Did they get the right answer? These aren't multiple choice answers. They're essays. Build a rubric. Read the transcripts. Use your judgment.
3. Did they show their work? The grade tells you it failed. The trace tells you why. An agent can say it did the job without ever finishing it. Open the hood.
4. Can they pass again tomorrow? Passing once doesn't mean the same grade the second time. The model updates. Variance kicks in.
Trust in an AI product is earned, interaction by interaction, and lost the moment you stop checking.
What does product work look like when user feedback automatically flows into the next build?
Prototype a feature, get real reactions, build the next version. No more clunky handoffs.
@TheoTabah & @gregisenberg, on the Startup Ideas Pod.
Using AI isn't the same as being AI-native. Most companies haven't felt the difference yet.
The way we work is changing shape. Agents do the execution. People move to the edges. The org runs on a shared brain that gets smarter every day.
That's a different kind of company. One built on three layers:
People who direct.
Agents who execute.
Context that remembers.
The gap between "uses AI" and "AI-native" is widening fast.
👇(1/2)
Want to help lead the design movement for AI products? Apply to join our team.
We're hiring:
→ Sr. Product Designers
→ Full Stack Software Developers
Apply here: https://t.co/VPDYUPPGC0
Visit our website to learn more about about us: https://t.co/zr8rSw5DBY
Your company knows more than you think. The problem is none of it is readable.
It's buried in Slack threads, old meeting notes, emails nobody saved, decisions nobody documented.
Making your company readable to agents and to yourself is the foundation of an AI-native org.
@TheoTabah takes on the Startup Ideas Pod with @gregisenberg
Watch the full pod 📷
https://t.co/FG8awkscej
What does it actually mean to be AI native?
There was no clear guide on the internet for how to become AI native so we built the definitive one (60 min masterclass):
1. An AI native org has 3 layers: people for strategy and taste, agents for execution, and a shared context layer that makes the entire company readable to agents.
2. AI eats the middle of your work. You used to spend 80% of your day on execution. Now agents do that. Your job is the bookends: deciding what to do and judging whether it's good enough.
3. Everyone is a manager now. Your output is the output of your agents. If your agents produce garbage, that's on you. You set them up wrong.
4. Using ChatGPT doesn't make you AI native. That's like having a website and calling yourself a tech company lol.
5. No AI native org without AI native people. Most companies skip straight to the tools. That's why it fails. If your people don't understand how to manage agents, the tech doesn't matter.
6. Making your company "readable" to agents is the real work. Every process, every decision, every piece of knowledge needs to exist in a format an agent can consume. Most companies are nowhere close.
7. Speed without signal is just expensive chaos. You need the system to move fast AND know if you're moving in the right direction.
8. The skill chain is how agents get good at your specific workflows. Skills build on skills. The more you invest in them, the more your company compounds.
9. The moat is the system. People managing agents, agents reading from rich context, the whole thing getting smarter every week. That compounds. Your competitor can copy your tools. They can't copy your system.
Full episode with @TheoTabah from @meetLCA on @startupideaspod. This is the stuff we normally keep internal but all the sauce is yours.
@TheoTabah is the brains behind advising the world's biggest companies on AI and building AI products. Your fav CEO's first call for figuring out AI.
You are in for a treat
Become AI native in under 60 minutes
https://t.co/EzreBHFyIJ
Watch