1.
Automation matrix
Chart the matrix at regular intervals.
High costhard to automate
High costeasy to automate
Low costhard to automate
Low costautomatable, nice to have
Start top-right. High cost, easy to automate. Y axis: your cost. X axis: automatability.
2.
Map external constraints
Detect stakeholder constraints and where approval is needed.
Understand which changes you can implement under your own authority, which need specific stakeholder approvals, and which are organisationally off-limits.
3.
Invest upfront context
Transfers in most orgs. Banks/health may lean red.
Matt Pocock's "grill-me" skill →
A skill that interviews you before drafting, so the AI starts with real context.
4.
Build your own AI council
Adapt with your own stakeholders and customer personas.
Ole Lehmann on the LLM Council pattern →
Multiple AI advisors debate the same question. Inspired by Andrej Karpathy.
5.
Rethink team structure
Organisational authority required. PM can advocate, not implement.
How can you gain speed and efficiency by giving more agency to your team members? And how do you, as a PM, ensure the things get built, the team is aligned, and the product remains coherent?
6.
Adapt stakeholder communication
"Finished-looking vs finished" is entirely within PM control.
Prioritise clear flagging of prototype stages, of test phases for products, and of the tech that accumulates in the background. Flag these up front.
7.
Protect learning capacity
Team-wide needs leadership alignment. Ideally an OKR.
Advocate for learning in your organisation: highlight the benefits to stakeholders. Give team members the opportunity to present new learnings, and consider including dedicated upskilling in the OKRs.
8.
Build a peer circle
Entirely personal. Next-day actionable.
Find fellow PMs and builders who are in active learning mode, and set up discussions at regular intervals. A hivemind approach to learning.
Start on the greens this week. Translate the ambers. Advocate for the reds.