The 90 Minute Setup That Saves PMs 5+ Hours Every Week

The evolution of the pm

You’ve used AI. You’ve pasted a user story into a chat window, asked it to tidy up your PRD, maybe even drafted a stakeholder update. Then you closed the tab and started from scratch next time.

That’s just the basics. And most PMs never leave it.

The problem isn’t the tool. It’s the operating model. You’re re-briefing from zero on every interaction. Your domain, your users, your product’s quirks, the political dynamics of your stakeholder map. Every. Single. Time.

That’s not a productivity gain. That’s a context-assembly tax you’re paying on repeat.

The shift: tool to team member

Think about what happens when a new PM joins your team. You don’t hand them a task on day one and expect a polished output. You onboard them. They learn the product, the customers, the codebase constraints, how decisions get made, and who needs to be in the room. Only then do they become useful.

AI works the same way. The unlock isn’t better prompts. It’s better onboarding.

This comes down to three things.

“You don’t need better prompts. You need better infrastructure.”

Mustafa Shairani, Product Director @ Polaris Software

1. Give it your context (Once)

Most AI platforms now let you store persistent preferences or system-level instructions. Use them. But don’t just write “I’m a PM at a B2B SaaS company.” That’s a LinkedIn headline, not a briefing.

Go deeper. Define your product domain. Name your user segments. Describe how decisions actually get made (not the org chart version, the real version). State your defaults: do you prefer concise over thorough? Bullet points or prose? Do you write for engineers or for commercial stakeholders?

When I set up my own workspace, I treated it like writing a company handbook for a new hire. I documented my role, my product area, the tools I use daily, even my writing style preferences. That context now loads automatically. I don’t re-explain who I am or what I’m working on. The AI already knows.

2. Wire it into where your work actually lives

A PM’s job is fundamentally about synthesis. You pull signals from customer calls, analytics dashboards, engineering tickets, Slack threads, and strategy docs. Then you compress all of that into a decision or a document.

If your AI can’t access any of those inputs, you’re still the middleware. You’re copy-pasting between systems and manually reconstructing context that already exists somewhere in your stack.

The fix is integration. Most modern AI tools support connectors (sometimes called MCP integrations) that link directly to your email, calendar, document storage, and project management tools. I’ve connected mine to Gmail, Google Calendar, and Notion. The result is that when I ask for help preparing for a meeting, it can pull the actual calendar invite, the relevant docs, and recent email threads without me hunting for links.

You don’t need to connect everything on day one. Start with two: your primary document store and your communication tool. That alone eliminates a surprising amount of tab-switching.

3. Codify your repeatable work into playbooks

This is where most PMs leave serious time on the table.

Think about the tasks you repeat weekly. Sprint summaries. Stakeholder updates. Competitive scans. PRD scaffolding. Release notes. These aren’t creative exercises. They’re structured outputs that follow a pattern you already know.

Document that pattern. Save it as a reusable instruction file (most platforms call these skills, templates, or custom instructions). The next time you need that output, you trigger the playbook instead of explaining the format from scratch.

I’ve built skill files for specific deliverables I produce regularly. Each one defines the structure, the tone, the audience, and the level of detail expected. The AI follows the playbook. I review and refine the output. The cycle takes minutes instead of the hour it used to take when I was drafting from blank.

The compounding effect matters here. Every playbook you build makes the next one faster to create, and every future model upgrade inherits the library you’ve already built.

What this actually looks like in practice

Here’s a realistic first week if you’re starting from scratch:

Day 1:

Populate your persistent context. Spend 20 minutes describing your role, your product, your users, and your working style. Talk to the AI like you’re onboarding a colleague.

Day 2-3:

Connect two integrations. Your document store and your email or calendar. Don’t overthink permissions. Start with read access and expand later.

Day 4-5:

Pick one repeatable deliverable you do every week. Write the steps as a skill file. Test it. Refine it.

That’s it. Five days, maybe 60 to 90 minutes of total setup. The return shows up every week after that.

TLDR: The real point

This isn’t about becoming an “AI power user.” It’s about recognising that the bottleneck in product work has never been typing speed. It’s context assembly and synthesis. If you can offload even 30% of that to a system that already understands your domain, your stakeholders, and your deliverable formats, you free up capacity for the work that actually moves the product forward: talking to customers, making trade-off decisions, and aligning teams around what matters.

Stop briefing your AI from scratch. Start onboarding it properly.