Anthropic just shipped their new "Claude Cowork" tool in less than two weeks, once the opportunity was identified.
They didn’t do it because it was on a roadmap defined six months ago. They didn’t do it because a stakeholder demanded it. They did it because they looked at their data, saw users (in this case, their own employees) hacking their terminal agent to organize personal receipts, and realized they had accidentally built a general-purpose product.
They observed, they built, and they shipped. In less than two weeks. Rather than months of traditional engineering work, Anthropic’s Cowork was developed in 10-14 days once the need surfaced.
This is in a sense, a masterclass in Agile Development and Empiricism, in the age of generative AI.
The Engineer-to-Product Manager ratio
For years, the tech industry’s standard for the Engineer-to-Product Manager ratio hovered anywhere from 4:1 to 10:1. Those numbers are highly contextual and should be taken with a grain of salt, but essentially: coding is slow and expensive, so you need a lot of hands on keyboards for every one person deciding what those hands should do.
As AI tools dramatically reduce the cost and time of software development, the bottleneck has shifted upstream. If a team can build a functional product in 10 days, the constraint isn't how to build it. The constraint is knowing what to build. According to Andrew Ng, the US tech industry is seeing more and more teams moving toward a 1:1 ratio.
This isn't just a staffing metric. It’s a change of paradigm.
What that means for us?
As a developer
If you are a developer, the days of being a "ticket taker" are numbered. When code becomes cheaper, the value shifts to the context. Developers need to understand the user, the business case, and the market as well as they understand their IDE. It is becoming more and more clear that developers’ ability to contribute and generate value will dramatically drop if they fail to focus on this context.
As a leader
If you are a leader, look at your org chart, your governance processes and your employee’s capabilities:
- Are you organized around functional silos or Products?
- How long do you need to get a new budget approved?
- How long do you need to push a new solution into the hands of the user?
- How long do you need to run an A/B test and collect data?
- Are your employees committed to narrow career paths or are they encouraged to take ownership end-to-end and diversify?
- Are teams able to learn and adopt new technologies and tools proactively?
- etc.
Anthropic’s Claude Cowork’s release wasn’t random luck. They have talked a lot in the past about how they approach Product Management, and how they have organized themselves so that this keeps on happening again and again.
They are just one example of how to apply transparency, inspection, and adaptation - the fundamentals of Scrum - at high speed. And if most organizations can’t move that fast, it’s not because they lack the AI tools. It’s because they lack the culture, the structures, the leadership and the capabilities that would allow them to deal with complexity and change effectively.
If you're looking to start discussions within your team or across your entire organization about how AI will impact future product development approaches and organizational structures, the “AI Sense-Making Workshop” should also be a useful resource.