Most executives are still reading about agentic AI. Gérard is building with it every day.
Orion Playbook is Gérard Métrailler’s public working notebook on product, technology, GTM, operations, and the changes that occur when agentic AI becomes part of how the work actually gets done.
Every week, Gérard publishes what he is learning from direct experimentation: where AI creates leverage, where it introduces new failure modes, and what software leaders need to rethink when the technology moves faster than the operating model around it.
About a 5-minute read. Published weekly.
A working notebook on how software companies actually operate
Software companies rarely underperform because one function is completely broken. More often, Product, Engineering, GTM, Support, and Operations are each doing reasonable things within their own boundaries, while the system as a whole creates friction, delays, conflicting incentives, or poor decisions.
That is the territory Orion Playbook explores.
It is a working notebook on product, tech, GTM, and operations, and how to make them run as one system, plus the journey down the AI rabbit hole, weekly, from an operator continuously learning through hands-on experimentation.
Two threads run through it.
The software operating system
Gérard spent his career moving across the boundaries that software companies tend to organize into separate boxes: product, engineering, marketing, sales, services, operations, and the P&L. Orion Playbook looks at what happens between those boxes.
How does a product decision change the economics of GTM? When does a technically rational architecture choice become a commercial constraint? Why does a pricing decision expose an organizational problem rather than solve it? What changes when AI affects several of those functions at once?
The goal is not to produce another management framework. It is to examine the system closely enough to make better operating decisions.
Hands-on agentic AI
The second thread is more experimental. Gérard is not studying agentic AI only through reports, vendor presentations, or industry commentary. He uses it every day to build, write, research, challenge assumptions, automate workflows, and redesign how work gets done.
Orion Playbook itself is part of the experiment. The newsletter, the website, and much of the operating model behind them are being built in public, with Claude as a co-creator. The work on agentic AI for software companies is therefore not separated from the tools being examined. Orion Playbook is one of the environments in which Gérard tests them.
Some experiments work. Others expose weaknesses quickly. Both are useful.
Built with the technology it examines
AI commentary is easy to produce. Running the workflows is more revealing.
A model can generate an impressive answer while misunderstanding the premise. An agent can automate a process while quietly multiplying a bad assumption. A workflow can look dramatically faster until quality control, context, governance, or exception handling is included.
Those are not theoretical edge cases. They become visible when the technology is put into daily use. That is why Orion Playbook asks questions such as:
- Your AI Agent Will Never Question the Premise: What happens when an agent executes perfectly against the wrong objective?
- Every AI Answer Is a Bet Dressed as a Fact: How should leaders operate when fluent output can obscure uncertainty?
- The Most Durable Thing in Your Software Isn’t in the Code: Where does defensibility move when generating software becomes cheaper?
- Directionally Correct, Specifically Utterly Wrong: When does plausible AI output become more dangerous than an obvious failure?
The register is deliberately analytical and sometimes contrarian. Not because provocation is the objective, but because useful operating judgment often starts by questioning what everyone has begun to treat as obvious.
For executives evaluating agentic AI for software companies, that distinction matters. The important question is rarely whether an AI capability can produce an impressive demonstration. The question is whether it can survive contact with customers, economics, organizational boundaries, operating controls, and real accountability.
From experimentation to operating judgment
The connection between Orion Playbook and Gérard’s advisory work is straightforward. The newsletter is where much of the thinking gets tested in public. The advisory work is where operating context becomes specific.
Gérard has spent more than 25 years inside software businesses, moving from product leadership through CEO, President, GM, and full P&L responsibility. His operating experience spans software-to-SaaS transformations, portfolio strategy, global GTM, M&A, operational restructuring, and the constant tension between growth, profitability, and product velocity.
Agentic AI introduces a new variable into that system. It does not remove the old ones. That is why Orion Playbook focuses on the intersection between the technology and the operating model around it.
What happens to product management when software creation accelerates? What happens to support economics when agents can resolve more work but also create new escalation paths? What happens to GTM when content and enablement become abundant? What happens to management when autonomous workflows begin making decisions that previously sat with people?
And perhaps most importantly: Which assumptions inside an existing software company stop being true?
Those questions are central to any credible AI operating model for SaaS. They are difficult to answer from a slide deck. They become clearer when you build.
The point is not to predict AI. It is to build better judgment around it.
No one knows exactly where agentic AI will take the software industry. What leaders can do is shorten the distance between observation, experimentation, learning, and operating decisions. That is what Orion Playbook is for.
Each week, Gérard takes one problem apart, documents what he is seeing, and puts the reasoning into public view. Over time, that creates something more useful than a collection of AI predictions: a dated, public record of how thinking evolves as technology and evidence change.
For leaders responsible for a software business who are trying to separate durable operating change from temporary AI enthusiasm, Orion Playbook offers a way to track the work as it happens.
About 5 minutes. Once a week.
Product. Technology. GTM. Operations. Agentic AI.
No AI cheerleading. No certainty theater. Just the questions, experiments, and operating implications worth examining.