I build AI products at scale. I share what it takes to be an AI PM. Only practical, real-world advice. No-fluff, no theory.
Start here ➡
Joined November 2014
- Tweets69.7K
- Following3.8K
- Followers29.7K
- Likes38.2K
AI product managers, please learn how AI agents work:
Most teams pick the wrong architecture. Here is how to pick the right one.
(Save this, you will need it to build your own agents.)
■ 𝗦𝗼𝗹𝗼. 𝗢𝗻𝗲 𝗮𝗴𝗲𝗻𝘁 𝗴𝗼𝗶𝗻𝗴 𝗿𝗼𝘂𝗻𝗱 𝗶𝗻 𝗮 𝗹𝗼𝗼𝗽.
Start here. The agent 𝘁𝗵𝗶𝗻𝗸𝘀, 𝗮𝗰𝘁𝘀, 𝗼𝗯𝘀𝗲𝗿𝘃𝗲𝘀 at what happened, then thinks again. It keeps repeating this loop until the job is done. So the important thing is to give it the right goal. Finally, connect tools and give it a memory. And it handles way more than you would expect. 𝗠𝗼𝘀𝘁 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 𝗻𝗲𝘃𝗲𝗿 𝗻𝗲𝗲𝗱 𝗮 𝘀𝗲𝗰𝗼𝗻𝗱 𝗮𝗴𝗲𝗻𝘁.
■ 𝗔𝗴𝗲𝗻𝘁 𝘁𝗼 𝗔𝗴𝗲𝗻𝘁 (𝗔𝟮𝗔). 𝗢𝗻𝗲 𝗵𝗮𝗻𝗱𝘀 𝗼𝗳𝗳 𝘁𝗼 𝘁𝗵𝗲 𝗻𝗲𝘅𝘁.
Use this when the work splits into two or three clear jobs (or goals) that always run in the same order. The refund agent processes the return, then hands over to the stock agent. Cheap to build. The catch is that nobody is supervising. So, if one of the agent makes a mistake, the other one never finds out.
■ 𝗢𝗿𝗰𝗵𝗲𝘀𝘁𝗿𝗮𝘁𝗼𝗿. 𝗢𝗻𝗲 𝗯𝗿𝗮𝗶𝗻 𝗵𝗮𝗻𝗱𝘀 𝘄𝗼𝗿𝗸 𝘁𝗼 𝗺𝗮𝗻𝘆.
Use this when the sequence is dynamic and several agents need to run at once. One agent understands the situation and the goal, then delegates the work to other agents. It then waits for each one to report back with a status. This central agent is the orchestrator and controls the success of the entire system. That is the advantage, but also the bottleneck. If the orchestrator breaks, everything breaks.
I am build these agentic systems live in my upcoming masterclasses.
You can join them for free.
Next one is on Satruday (9am PT, 4pm UK, 9:30pm IST.)
Link below
Join for free
Limited seats.
First come first serve
justanotherpm.com/ai-pm-live…
AI product managers, please learn about AI agents.
Here is the AI Agentic Cheat Sheet every PM needs 🧵
The only rule that matters
Start with a single ReAct agent + good tools.
Add hierarchical or parallel only when you can name the specific failure of the simpler version.
Most production systems don’t need more than one strong agent with skills.
Complexity is a liability until it’s an asset.
Over the next 2 Saturdays, I will show you live I
1) how agents work
2) different types of agents
3) how I build a single agent
4) how I use multi agent systems to run my life
5) how you can build your first agent
Please join these sessions, they are completely free. 100% practical.
Next event: Saturday, 31 Oct. (link in the first comment). We start with understanding agentic architecture and building single agent systems.
ign up here. Seats are filling fast. Free to attend. But seats limited:
justanotherpm.com/ai-pm-live…
To be a great AI product manager, you NEED to understand how agents work.
Here are 8 videos that will teach you everything you need to know:
(Save this post. you will NEED it.)
1. Generative AI vs agentic AI youtu.be/EDb37y_MhRw
2. How the agent loop works youtu.be/EuzYhzB0vbI
3. MCP vs API youtu.be/7j1t3UZA1TY
4. 4 kinds of memory for agents youtu.be/BacJ6sEhqMo
5. How Anthropic builds agents youtu.be/D7_ipDqhtwk
6. Agents and skills youtu.be/CEvIs9y1uog
7. Build an agent in 8 minutes youtu.be/Zqno_vux6d8
8. Agents working together youtu.be/sWH0T4Zez6I
Over the next 2 Saturdays, I will show you live I
1) how agents work
2) different types of agents
3) how I build a single agent
4) how I use multi agent systems to run my life
5) how you can build your first agent
Please join these sessions, they are completely free. 100% practical.
Next event: Saturday, 31 Oct. (link in the first comment). We start with understanding agentic architecture and building single agent systems.
Sign up here. Seats are filling fast. Free to attend. But seats limited:
justanotherpm.com/ai-pm-live…
To be a great AI product manager, you NEED to understand the basics of AI.
Start here. 10 videos. 85 minutes.
(Save it for easy reference)
1. AI vs ML vs DL vs GenAI youtu.be/qYNweeDHiyU
2. How LLMs work youtu.be/LPZh9BOjkQs
3. How to write good prompts youtu.be/jC4v5AS4RIM
4. Why AI makes things up youtu.be/005JLRt3gXI
5. How RAG gives AI the facts youtu.be/T-D1OfcDW1M
6. What a vector DB does youtu.be/gl1r1XV0SLw
7. RAG, fine-tuning, prompting youtu.be/UabBYexBD4k
8. What is MCP youtu.be/_d0duu3dED4
9. AI agents, clearly explained youtu.be/FwOTs4UxQS4
10. 7 terms for every AI PM youtu.be/VSFuqMh4hus
Over the next 5 Saturdays, I will be showing live how I apply these learnings to build real AI products. I'd love it if you joined our online sessions. We start this Saturday, 24 Oct.
First up: your own AI second brain.
Sign up free
Sign up here. Seats are filling fast. Free to attend. But seats limited:
justanotherpm.com/ai-pm-live…
The nature of the work we do has already changed. Most of the change is already irreversible.
AI does not simply assist the product manager. It redefines the constraints (and opportunities) under which we work. The old world--where we translated ambiguity into documents, coordinating across teams, built roadmaps with narrativ--was built for a world with bottlenecks and expensive coordination.
Those bottlenecks are disappearing.
What remains is judgment:
- the ability to define what “good” means in measurable terms
- to build the systems that stay true to it
- to detect, early, when the system has drifted from the goals
Those who continue to operate as if the old constraints still hold will discover that their work is increasingly substitutable.
Data analysis, user research, writing requirements, brainstorming solutions. These are no longer super powers
What is/will be a real super power is the clarity to say which outcomes matter, which failures are unacceptable, and which trade-offs the product will accept.
The same prowess defined how we did the work in the old world too. But the unpredictabilty in our products has madethe same work more challenging.
The world will keep changing. Models will become better. Your learnings and expertise will be obsolete. The only way to to stay relevant (and try to win) is to be part of the movement.
As builders, leverage AI to become more efficient not to sound cool.
As PMs, build products that are forward looking. Build solutions solve the problem of tomorrow. Embrace the unknown, bet big, learn fast.
But learn.
Invest time to continuously learn --> do --> learn
That is the only way to survive in the new world.
The role will continue. The individuals who occupy it will be those who have made the new constraints their own.
AI PMs who can’t build and own evals are already behind.
Here’s why this is the highest-leverage skill right now, and the exact path to get good at it.
1/ Evals are the new unit tests + PRDs for AI products.
Traditional software has deterministic behavior. LLMs don’t. Same prompt → different outputs. Without systematic evals you’re shipping on vibes, and vibes don’t scale.
8/ Tools to actually practice with (start simple):
Promptfoo (great CLI + YAML configs)
DeepEval (pytest-style)
Ragas (if you’re doing RAG)
LangSmith / Braintrust / Phoenix for traces + online evals
You don’t need to master all of them. Pick one and build something.
9/ The meta skill is error analysis + calibration.
Look at failures, categorize them, write better rubrics, validate your LLM-as-judge against human scores, then automate. Iterate the eval as hard as you iterate the product.
Evals aren’t an engineering side quest. They’re how you turn non-deterministic AI into a product you can actually ship, improve, and defend.
Start with one feature and 50 examples this week. The rest compounds fast.