Phase 3. Marketing: the agent plans the go-to-market
This article is about planning marketing, not executing it. How the product goes to market: who it's for, what sets it apart, what it costs, which channels sell it, and what metrics will tell us the launch worked. Launch execution — turning on channels, the landing page, first touches — comes later, woven into Release (11) and Maintenance (13). Here we build the plan.
Marketing planning is traditionally wrapped in an aura of creativity and market instinct — the same aura as the product idea in phase 1. And it breaks against the same argument: under the instinct sits data synthesis and optimization against criteria; the criteria are just usually unspoken.
The human role today
A marketer, a product marketing manager, or — early on — the founder themselves. They build the segmentation, formulate the positioning, set the price, choose the channels, draw up the launch plan and metrics. Classically this is strategy sessions, competitive analysis, back-of-napkin unit economics, and channel decisions made "from experience."
A significant share of this work is a conclusion drawn from data already collected about the market and the pain. Who the product is for follows from the discovery segments. What sets it apart follows from the competitive field and the unmet pain. What it costs follows from willingness to pay and unit economics. Which channels follows from where the segment lives. It's a chain of inferences, not a series of flashes of insight.
What we hand to the agent
The agent holds the role of marketing strategist. From discovery-report and opportunity-brief it builds the segmentation, formulates the positioning and UVP, proposes a pricing hypothesis, runs rough unit economics, chooses channels per segment, and assembles a GTM plan with success metrics. It's not deciding "write a post" — it's deciding "how does this product go to market at all."
And here the agent does something a human rarely does: it explicitly ties every plan decision back to a source in discovery. The positioning isn't "we came up with something nice" — it's derived from a specific pain and a specific contradiction flagged in phase 2.
Agent architecture
Phase state-machine
The agent holds the role
Tools: segmentation, competitive positioning, pricing models, rough CAC/LTV and unit economics, channel selection, launch-plan and metrics generation.
Artifact
gtm-plan — segments, positioning (UVP), pricing model, channels, funnel, launch success metrics, traced back to discovery.
Handoff: gtm-plan → Requirements (what from the positioning must be in the product at launch) and Release/Maintenance (launch execution). The gtm-plan → Requirements link matters, and it's the one that breaks most often with humans: positioning promises the market one thing, and the launch product gets built to a different priority. The agent makes the trace explicit — a promise from the plan becomes a requirement in the backlog rather than staying in a slide deck.
- Inputs:
opportunity-brief(phase 1),discovery-report(phase 2), market and competitive data. - Tools: segmentation, competitive positioning, pricing models, rough CAC/LTV and unit economics, channel selection, launch-plan and metrics generation.
- Artifact:
gtm-plan— segments, positioning (UVP), pricing model, channels, funnel, launch success metrics, traced back to discovery. - Trigger: completed
opportunity-briefanddiscovery-report. - Handoff:
gtm-plan→ Requirements (what from the positioning must be in the product at launch) and Release/Maintenance (launch execution).
The gtm-plan → Requirements link matters, and it's the one that breaks most often with humans: positioning promises the market one thing, and the launch product gets built to a different priority. The agent makes the trace explicit — a promise from the plan becomes a requirement in the backlog rather than staying in a slide deck.
Where it breaks
Brand taste. The agent produces correct positioning — but average positioning. It lands on "reasonable and right" more often than "sharp and distinctive." A brand voice that hooks precisely because it's a little uncomfortable and not optimal is harder for the agent: it optimizes toward the center, and a brand often lives on the edge.
Channels built on human trust. Partnerships, community, long sales cycles, reputation within a narrow circle — these are channels that can't be spun up by planning; they grow on live relationships. The agent will account for them in the plan but won't replace the person holding those relationships.
Signing off on the budget. Choosing a channel is a bet with money. The agent will calculate CAC/LTV better than a human, but the decision "we're putting the marketing budget here" is owned by an accountable person.
What stays human
Approving the positioning and the budget, the brand's tone. A candidate for compression on the calculation and channel-selection side; the durable remainder is brand taste and the signature on the budget, and the latter, again, traces back to the principal (phase 14).
human remainder ≈ 80%
Provocation / thesis
A GTM plan is a structured inference from market and pain data, not a creative act. Once discovery is digitized, segments, price, and channels are computed, not "felt." There's exactly as much creativity in marketing planning as there is in unspoken criteria — and the agent is forced to speak them. A human is needed where brand taste and the signature on money are required; everything else in planning is optimization the machine runs more transparently, with an explicit trace back to the data.
Vitrina in this phase
The agent builds Vitrina's plan from discovery — and resolves the contradiction flagged in phase 2 (the founder's "beautiful design builder" versus the market's "payment pipe"). The data sides with the market: the value is in capturing the order and the payment, not in the design. The agent puts this at the core of the positioning — "a storefront in an hour, not a website in a month": sell speed and payment, not a design builder.
The segment is local SMB with no tech team. The pricing model is freemium on the storefront plus a transaction cut: whoever's making sales pays, and the barrier to entry is zero. The channels are local entrepreneur communities and service marketplaces where the segment already lives, not abstract performance traffic. Launch success isn't measured in registrations — it's the share of stores that processed at least one paid transaction.
Artifact → Vitrina's gtm-plan. And the trace outward: the promise "payment in an hour" goes into Requirements as a hard priority — accepting payment must be in the first release, or the positioning is empty. That's how a marketing decision becomes a checkable product requirement and lands in the backlog instead of settling into a slide deck.
Read next
Building AI-driven delivery in-house?
Designing the ADLC loop: where the agent holds the role and where the human principal remains — for your team and product.
Email meThe transition engine
Next Move Engine — the system that takes a team to an autonomous delivery loop.
Next Move Engine →