Deep Research-to-Campaign Workflow
Turn product research into a strategy and ready-to-edit campaign assets
- Difficulty
- Moderate
- Time to result
- ~days to results
- Steps
- 7
- Confidence
- 98%
The workflow begins with deep research into the product, including its features, use cases, integrations, audiences, and desired customer outcomes. That research is then used to generate a detailed prompt rather than asking a model for a campaign from a thin brief. The marketer supplies both the prompt and the underlying research to a capable reasoning model, which produces positioning, personas, channel plans, content ideas, spending guidance, timelines, and measurement criteria. A human reviews the result for market-specific weaknesses, such as bland thought leadership or questionable channel economics. Once the strategy is sound, each section becomes an asset-generation request, producing landing-page copy, articles, messages, and other campaign components that can be edited and deployed.
Origin
Extracted from Marketing Against the Grain through a demonstration using HubSpot Breeze customer agent and Gemini 2.5 Pro.
Core principles
- 01Research quality determines strategy quality
- 02Give the model more relevant context than a summary alone
- 03Match each task to the model best suited to perform it
- 04Review AI strategy with human market knowledge
- 05Expand the approved strategy into channel-specific assets
How to run it
- 1
Run deep product research
Compile reliable information about the product, its capabilities, integrations, use cases, customer outcomes, target buyers, and competitive context. Use research tools or an LLM research mode to organize the evidence.
Pro tip The same process can research your own product, a competitor, or an unfamiliar category.
Watch out Do not let unverified or outdated research become the factual foundation of the campaign.
- 2
Generate the campaign prompt
Ask an LLM to transform the research into a detailed prompt for building the marketing campaign. Specify the required channels, deliverables, audiences, and strategic components.
Pro tip Use the research itself to make the prompt more complete than one written from memory.
Watch out Check that the prompt actually asks for a campaign rather than only a product summary.
- 3
Supply the full context
Give the selected reasoning model both the campaign prompt and the raw research. The underlying evidence helps preserve detail that may have been lost in the prompt summary.
Pro tip Include relevant positioning, customer proof, brand beliefs, and provocative market perspectives.
Watch out More context is useful only when it is relevant and internally consistent.
- 4
Generate the strategy
Have the model produce the campaign's audience segments, personas, value proposition, messaging pillars, channel tactics, budget allocation, timeline, KPIs, and optimization plan.
Pro tip Use a model with strong reasoning rather than choosing solely by familiarity.
Watch out A polished response can still contain weak channel assumptions or generic positioning.
- 5
Review with expert judgment
Evaluate whether the model understood the assignment and whether its personas, messages, channels, and economics fit the real market. Identify sections that need stronger differentiation or adjustment.
Pro tip Compare the output against existing positioning and firsthand campaign performance.
Watch out Do not treat AI recommendations as automatically correct because they are detailed.
- 6
Expand sections into assets
Prompt the model to turn each approved strategy section into a complete first-draft asset, such as a landing page, article, advertisement, or persona-specific message.
Pro tip Keep asset generation in the same thread so the model retains campaign context.
Watch out Visuals and final copy may require separate tools and human editing.
- 7
Edit and deploy
Review every generated asset for factual accuracy, brand voice, differentiation, and channel fit. Move approved versions into the marketing platform and launch the campaign.
Pro tip Treat generated assets as strong V1 drafts that accelerate execution rather than eliminate review.
Watch out Never publish customer claims, competitive statements, or performance promises without verification.
In the wild
The host researched HubSpot's customer agent, used that research to create a detailed campaign prompt, and supplied both the prompt and raw findings to Gemini 2.5 Pro. The model generated target segments, buyer personas, value propositions, channel tactics, content pillars, budget guidance, a timeline, and KPIs. The host then requested a complete landing page from one section of the plan.
→ A broad B2B marketing brief became a detailed campaign strategy and a substantial first-draft product page.
A two-person software company researches its product, customer interviews, support records, and competitors. It uses that evidence to build a structured campaign prompt, reviews the resulting positioning and channels, and then generates landing-page copy, nurture emails, and social posts from the approved strategy.
→ The team produces a coordinated first campaign in days while retaining human control over claims and positioning.
Common mistakes
Prompting without research
A clever prompt cannot compensate for missing product facts, customer context, or market evidence. Thin inputs tend to produce generic campaign recommendations.
Omitting the brand's point of view
Without explicit beliefs, differentiation, and provocative market takes, the model may suggest reasonable but undistinguished content.
Deploying without expert review
AI can identify plausible channels and budgets while still missing real-world cost or performance constraints. A marketer should validate the recommendations before spending money.
Is it for you?
Best for
It is best for marketers who understand their product but need to develop and execute a multi-channel campaign quickly.
Not ideal for
It is not ideal when product facts, customer evidence, or strategic positioning are unavailable or unreliable.
From the transcript
“a killer workflow is deep research use that deep research to write a prompt for you and then use that prompt in the best model…”
“I find this is another tip that the more context you provide the better”
“I can literally go through each of the sess sections on this plan and just say build it for me and I'll have one massive…”
From the episode
This AI Marketing Strategy Would Cost You $25K - But It’s FREE