Babe, wake up - it's 3/6/9 weeks now.

Michael Watkins' The First 90 Days needs to drop the zero and shorten the timeline - with AI, the 30/60/90 day plan is now the 3/6/9 week plan.

You might be the first marketing hire, leading a small team, or taking on part of a larger marketing org. Get your knowledge graph and agents working before day one. Three weeks in, I'd expect you to understand the business well enough to make decisions. By week six, you should have picked your bets and have something to show for them. By week nine, you should be running the work you were hired to lead.

On day one, connect your AI harness of choice, for me it's Codex, to the systems relevant to your work and start making progress with the context already available to you. Take what you've learned and speak with your new coworkers about it, share it, ask if you're missing anything, then start to establish your baseline understanding of the business.

T-minus 2: Onboard your agents first

Set aside time in the two weeks before starting to get your knowledge graph and a small set of agents ready. The first week in a new job has enough admin. Teaching a blank chat window your frameworks and what the company does doesn't need to be part of it.

For me, the day I mentally commit to a new role is the day I start. Before the official start date, ask your hiring manager what the team already uses for AI, where company knowledge lives, and what you can access. Use what's available to start building context and preparing your agents. As the first marketing hire, you may need to set up the basics. In a larger organization, you may be fitting your workflows into systems the team already uses. Find out what's there before deciding what to stand up.

I started both my knowledge graph and the Market Signal agent before day one. There's enough in product docs, customer stories, competitor sites, and Reddit/LinkedIn/X threads to start doing useful work. For Blaxel, I built a Market Signal agent to bring together competitor activity and developer discussions, then propose recommendations with sources attached. By the time I joined, I had specific questions to work through with the team and a system I could keep using.

You can use those same sources to build your initial understanding of the product, who it's for, how the company positions it, and what backs up those claims. Keep that context in your knowledge graph, including the assumptions you want to check with your new coworkers, so you and your agents can build on it as you learn. I've covered the setup in You Don't Need Better Prompts. You Need a Knowledge Graph.

Week 3: Start making calls

Sit in on live prospect and customer calls, review old ones and use the product yourself. Discuss with the team what they're working to achieve and what's already in motion. Put your agents to work connecting what customers are telling you with what Sales and Product are seeing. Bring that back to the team, compare notes, and work out what it means for marketing.

Here's what I've found immensely useful:

• MCPs: Slack, Granola, Gong, HubSpot/Salesforce, Linear, Notion, PostHog, X Ads, ChatGPT Ads, Profound.
• APIs: Google Ads, LinkedIn Ads, Reddit Ads.

Use what's available and relevant. If the call archive or useful reporting doesn't exist yet, include those gaps in your assessment and work out what's needed to fill them.

Agree on what the numbers mean, too. The paid-growth dashboard I built at Blaxel keeps ad-platform conversions separate from product and revenue outcomes. Check whether the ads are reaching the people you want to reach, what counts as a conversion, and whether those conversions are turning into product usage or revenue.

By week three, you should have worked through what you've learned with your leaders and the people you work with, enough to agree on how things are going and where to focus next. Share your thinking as it develops. Give people something they can contribute to while you're still figuring it out together.

Week 6: Pick your bets and ship

By week six, you should have an agreed plan for the work you own, at least one meaningful initiative underway, and a quick win you can point to. A strategy deck doesn't count as the quick win - we don't need any more decks!

Put the choices in one markdown file: what are we going to stop, start, and continue? Pick three core bets to focus the work, and be clear about what you'll do yourself and what you need from others.

In my ongoing consulting work with Guild, I've built a workflow to capture ideas coming out of meetings and turn them into prototypes. It takes a transcript in Slack, proposes directions, and turns the selected idea into a browser-viewable prototype that comes back to the same thread. I'd much rather try the idea than book another meeting to check whether we all imagined the same thing. Put the prototype in front of someone who'd actually use it and find out where it falls apart.

Whatever you count as the quick win, show what changed or which decision you were able to make. Be honest about what you've learned and what still needs testing before you commit more time or money.

Week 9: It's your job now

You're running the marketing work you were hired to own. The plan is in motion, work is shipping, and you're using what you learn to decide what happens next.

The knowledge graph should now reflect what you and the people you work with have learned, with everyone contributing context, correcting assumptions, and recording decisions as they do the work. You, your coworkers, and your agents should be able to use that shared context without someone having to explain it all again for every task. If you're looking for a set of agents to start with, I built the Marketing OS for Guild. Its specialists share an approved company brief.

Doing the work yourself may be how you prove a program before deciding what needs more investment or support. When I'm leading other people, I like being microinterested: get into the details where the risk warrants it, build context together, then give people room to own the work.

There'll still be things you don't know, whether you're waiting for a sales cycle to play out or gathering more evidence for a difficult people decision. Say which parts need time and why.

By week nine, ask the people you work with what got better because you joined.