AI

AI Strategy

How multi-step LangGraph strategies plan and execute across your workspace.

What it is

AI Strategy builds a three-part marketing strategy — a Core Strategy, a Messaging Framework, and a Campaign Plan — through one guided conversation per strategy. Each conversation is a StrategyConversation row with its own strategy_uuid; the finished pipeline itself is stored as one JSON document alongside it, referenced by that uuid.

3

Strategy parts, in order

3

Mandatory inputs

1

AI agent per conversation

How it works

One AI agent runs the whole conversation and drives it toward a fixed goal: collect your product/service description, your target audience, and your founder’s intent (your main objective), and nothing else — it is explicitly instructed to infer market positioning, secondary audiences and messaging tone itself rather than asking you for them. From those three inputs it generates each part of the strategy in turn, shows you a short bullet summary rather than the underlying structured document, and waits for your approval before moving on.

  1. 1

    1. Core Strategy

    Business goals, target audience and unique value, generated from your three inputs.
  2. 2

    2. Messaging Framework

    Built to align with the approved Core Strategy — its core message reflects your unique value, and its audience messaging speaks to the audiences already defined.
  3. 3

    3. Campaign Plan

    Integrates both prior parts into goals, an email workflow plan and a content workflow plan — the last part of the pipeline.

Text alone never saves anything

The agent only persists a section by calling a save action for it, and only once you’ve approved it — describing the save in its reply is not the same as it happening. Each save reports success or failure; the agent is expected to never claim something was saved when it wasn’t. Approving one part also auto-advances to generating the next, in the same reply, without you having to ask.

Status and progress

A conversation’s status moves through draft → in_progress → review → active (with archived for an old one), and three independent booleans — one per part — track exactly which sections are approved so far, giving a simple completion percentage across the whole pipeline. You can revise any approved part later by asking for a change in the same conversation.

Screenshot pending — A completed strategy with all three parts approved
All three parts approved — Core Strategy, Messaging Framework and Campaign Plan.

Non-obvious rules

The three mandatory inputs are a hard floor, not a suggestion — the agent is instructed to ask for whichever of the three you haven’t given it and to stop asking about anything else, rather than interviewing you about market details or business goals it can infer on its own. Later parts genuinely depend on earlier ones: the Campaign Plan is built from both the approved Core Strategy and the approved Messaging Framework, so approving out of order isn’t possible.

Related

New to building a strategy? Start with the AI Strategy guide.