Building the Brief Before the Brief

Context

HubSpot's Marketing Studio is building AI-powered campaign creation — giving marketers tools to generate assets, copy, and campaign structures using AI. The product works. The output often doesn't.

Not because the models are wrong. Because the inputs are thin.

Marketers come to the creation experience with a campaign idea in their head — a goal, an audience, a tone, a set of constraints built up over years of brand work. None of that context makes it into the prompt. The AI gets "write a campaign for our spring sale" and produces something technically correct and entirely generic.

The Problem

The gap between what marketers know and what they give the AI is the entire product opportunity. A marketer who's been working a brand for three years has rich mental context — audience segments, brand voice rules, campaign history, competitive positioning. None of it is structured. None of it is accessible to the AI at the moment it needs it most.

The result: AI-generated assets that feel off-brand, require heavy editing, and undermine trust in the tool. Marketers stop using AI for anything important and go back to writing from scratch.

The Insight

The opportunity isn't in the generation step. It's in the step before the generation step.

Campaign Planning Hub is a structured workspace that lives upstream of creation. The job: take unstructured planning artifacts — briefs, strategy decks, audience notes, past campaign learnings — and convert them into the context layer that makes AI generation accurate.

When the AI knows who you're talking to, what you're trying to say, what the brand sounds like, and what's worked before — the output changes. Not because the model changed. Because the input changed.

Discovery Process

Leading continuous discovery with marketing teams at companies across HubSpot's customer base. The process follows the Opportunity Solutions Tree framework: starting from the desired outcome, mapping the opportunity space through user interviews, and validating solution directions against real user behavior rather than assumed preferences.

Opportunity Solutions Tree

The OST is complete. The desired outcome: AI-generated campaign assets that are accurate and on-brand without requiring heavy editing. Three main opportunity areas have emerged from the discovery work:

• Context capture: Marketers don't have a system for their brand knowledge. It lives in their heads, in old decks, in Slack threads. Making it structured and accessible is the first unlock.
• Campaign memory: What worked before should inform what's generated next. AI that can reference past campaign performance is meaningfully better than AI that starts from zero every time.
• Audience grounding: Generic audience descriptions produce generic copy. Specific audience context — who they are, what they care about, what language resonates — produces copy that lands.

Interview Findings

Marketers consistently describe the same pain: they know exactly what they want, but expressing it to an AI feels like re-explaining themselves every time. The ask isn't a smarter AI. It's a system that remembers.

The strongest signal: marketers who use AI most are the ones who've already built their own context systems — saved prompt templates, brand voice documents, audience persona files they paste in manually. Campaign Planning Hub is the product version of the workaround they've already invented.

Solution Direction

Campaign Planning Hub is a structured workspace with three components:

• Brand context store — voice, tone, audience segments, visual guidelines, non-negotiables. Captured once, available to every generation request.
• Campaign brief builder — a structured form that converts a marketer's intent into the specific context fields that matter for generation: goal, audience, message hierarchy, tone, constraints.
• Campaign memory — post-campaign learnings stored in a way that informs future briefs. What performed, what didn't, what the audience responded to.

The north star: a marketer should be able to sit down, describe what they're trying to do in plain language, and get AI output that feels like it was written by someone who actually knows their brand.

Early Signals