The context layer every AI-native marketing team needs

Forwarded this email? Subscribe here for more.

👋 Hey there, I'm Nick. Each week, I share field notes from the people building the future of marketing. For more: LinkedIn | The Marketing Engineer | Profound University.

That is the real lesson from this episode with Reema Batta, VP of Growth Marketing at Figma.

Figma is operating in the middle of a very specific marketing problem. Product velocity is up. Customer expectations are up. AI can help teams create, test, and iterate faster than before.

The bottleneck moved upstream.

If production gets cheaper, the scarce work becomes deciding what should exist, what answer earns trust, what context the team should use, and which ideas deserve distribution. That is judgment work. It is upstream of writing. It is upstream of prompting. It is upstream of publishing.

The part of Reema's answer I liked most was how clearly she separated production speed from quality.

AI lets the team generate more versions, test more answers, and launch more programmatic pages. That matters. The human role shifts toward strategy, brand voice, quality control, and deciding what the right answer should be for a customer in a specific moment.

Before AI, the bottleneck was often production. After AI, the harder questions move earlier: what should we say, why should a customer believe us, which ideas deserve shipping, and what should stay unpublished even if it is easy to make?

The best marketing engineers are going to be good at building systems, but they still need marketing judgment. They need to know what not to do. They need enough channel, customer, and business context to recognize when the machine produced something plausible but wrong.

That is also why I think token maxxing is a temporary phase. It makes sense as an adoption driver: get people using the tools, experimenting, and building the habit. Then the metric has to move from usage to business value. Did it improve customer experience? Did it reduce cost? Did it create pipeline? Did it help the team learn faster?

Usage is the adoption metric. Better judgment in the system is the value.

The most tactical thing was how Figma handles product marketing context.

Figma ships quickly. That creates a marketing problem most fast-moving companies recognize: messaging, positioning, value props, launch notes, and audience context can go stale almost immediately.

The old answer is more coordination. More update meetings. More Slack follow-ups. More recordings. More product marketing handoffs.

Reema called that the coordination tax.

Figma's better answer is a shared context layer:

This is the workflow I would steal first.

The idea is simple: the product marketing team maintains a living source of truth, then channel teams can reference it through a skill when they need to write an email, update a page, launch a campaign, or refresh messaging.

It also connects directly to the Matt Swulinski episode. Matt's lesson was about turning personal workflows into team infrastructure. Reema shows the org-level version of the same idea.

If a team wants to move faster with AI, the context has to be shared, current, and legible to the system. Otherwise every workflow becomes a private universe with stale docs, hidden assumptions, and prompts carrying old facts.

Building more agents is only part of the work.

The deeper work is building the memory they can trust.

I keep coming back to the idea that teams may need an AI librarian: someone who keeps shared context clean, current, and usable. Maybe that is a new role. Maybe it is part of product marketing or marketing ops. The work is real either way.

One of the better tensions in the episode was the generalist versus specialist question.

The easy take is that AI makes everyone a generalist. I think Reema gave the more useful answer: it depends on scale.

At startup scale, the "barrel" generalist who can move across problems is incredibly valuable. In a small team, you need people who can scan an ocean of possible work, pick the thing that matters, and go do it.

At enterprise scale, the answer changes. Distribution gets more complex. Products multiply. Channels become deeper. The operating model turns into a matrix.

That is where deep channel expertise still matters.

Reema shared a line from a Figma board member, formerly the CMO of Salesforce, that captures it:

I love that framing.

The best marketing engineers will be marketers first. They will know the channel. They will know the customer. They will know the business. Then they will add systems thinking, AI fluency, and workflow design on top.

That combination is the hard part.

Pure generalists can move fast, but they can miss the nuance. Pure specialists can have taste and depth, but they can get trapped in old operating models. The rare person is the specialist who can also think in systems.

First, I want to judge AI marketing work by the quality of the decisions it improves.

Second, I want every repeatable workflow to have a context layer behind it. If the system depends on someone remembering where the latest positioning lives, the system is already fragile.

Third, I think the marketing engineer role is going to split by company stage. At early-stage companies, it may look like an AI-native generalist who can build across the whole funnel. At larger companies, it may look like a specialist who deeply understands a channel and can redesign that channel's operating system.

That is the macro trend I see here:

AI makes production abundant. The winners will be the teams that can turn context and judgment into infrastructure.

Listen here

If that's you, or you want it to be, stick around. Hit reply and tell me what you're building right now. I read everything.

See you next time,

Nick