Grok Bot Marketing Guide: Six-Role Collaboration from Competitor Research to Landing Pages, Ads, and Post-Mortems

The Grok Bot marketing guide demonstrates how to coordinate six permanent roles to complete a full marketing delivery. Josh Kim uses Bots to analyze competitors, draft positioning in Google Docs, build Google Ads campaign structures, open PRs in the website code repository, and retrieve live performance data.

These actions are traditionally scattered across different tools and specialized roles. Bots can query each other directly for context; as handoffs multiply, a Project Manager coordinates the workflow, allowing marketers to focus on reviewing positioning, live account operations, and spend.

Why Single-LLM Chat Fails: Breaking Down the Six Specialized Roles

Marketing has never been a one-dimensional task inside a single person's head. Research, positioning, web development, campaign execution, and post-mortems belong to distinct roles in real-world commercial teams. If you try to force a general-purpose AI within a single chat window to handle high-level competitor analysis, write frontend HTML/CSS code, and schedule ad keywords all at once, the conversation quickly becomes bloated and drifts away from its original goals.

What teams need are "permanent specialized roles," each handling its own responsibilities:

1. Market Researcher: Equipped with an independent browser to autonomously browse competitor websites and your own product pages, uncovering differentiated angles with genuine marketing penetration.
2. Product Marketer (PMM): Translates research findings into positioning briefs, landing page structural outlines, and search ad copy variants inside Google Docs.
3. Website Ops: Leverages the AI coding tool Cursor under the hood to transform landing page outlines into runnable frontend code, opening Pull Requests (PRs) directly in the marketing GitHub repository.
4. Performance Marketer: Connects directly to the Google Ads account to build approved test copy into executable Campaign Shells.
5. Marketing Analyst: Regularly pulls live ad test data to generate concept-level Scoreboards, offering recommendations on scaling, cutting, or pausing campaigns.
6. Project Manager (PM): Coordinates cross-role handoffs globally, serving as the "single entry point" for human operators and eliminating the fatigue of context switching across multiple chat windows.

End-to-End Walkthrough: Building the Zero-to-One Closed Loop for Fictional Airline X-Air

To illustrate how this system operates in real business scenarios, the original author introduces X-Air, a fictional long-haul airline, as a standalone example to walk through the entire loop from research and landing pages to ads and post-campaign analysis.

Step 1: Uncovering an Unconventional "Differentiation Blind Spot"

Traditional AI research often churns out dozens of pages of generic fluff—platitudes like "delivering premium service" and "luxurious flight experiences"—which offer zero practical value for actual operations.

The operator directs the Market Researcher to the X-Air website and a designated set of long-haul airline competitors. The agent opens its browser, reviews the relevant pages, and surfaces concrete differentiators among competitors:

Key Finding: Focus on useful time in the air, rather than luxury for its own sake.

This insight is sufficiently concrete to serve as a solid foundation for downstream product marketing work.

Step 2: Cross-Bot DM Handoffs and Google Docs Comment Collaboration

Previously, you had to copy research text manually, reformat it into a prompt, and feed it into the next chat window. In the Grok Bot architecture, however, agents can exchange direct messages in the background to retrieve context.

Upon receiving the task, the Product Marketer directly messages the Market Researcher in the background to sync on competitor context, then creates a positioning brief in the team's Google Docs:

• Synthesizes core one-liners;
• Structures core value statements;
• Drafts the information architecture for the landing page;
• Pre-writes a set of headline variants for Google Search ad testing.

This is where key human intervention takes place: The operator opens Google Docs just like reviewing a draft from a human colleague, leaving a collaborative comment next to a compelling line of copy: "Double down on this angle." Detecting the human's document comment, the Bot automatically updates the positioning document and its corresponding ad variants.

Step 3: Codifying the Landing Page with Screenshot PRs

With the human-approved landing page outline in hand, the task flows to Website Ops.

Many marketing teams get stuck on landing pages—waiting for sprint schedules, Figma designs, and frontend asset slicing. This Bot, backed by the Cursor coding engine, directly branches off the marketing site's GitHub repository, writes page code, and submits a Pull Request (PR).

Throughout the coding process, it shares intermediate UI rendering screenshots in real time. Once the code is merged, marketers inspect the live page directly, rather than debating over undelivered visual mockups.

Step 4: Entering the Ad Platform to Build "Real Shells"

Once Performance Marketer receives the approved test copy, it connects directly to the Google Ads account to build Campaign Shells tailored for messaging tests.

(Note: Campaign Shells refer to real account structures created inside the ad platform rather than offline spreadsheets; targeting, scheduling, or statuses can be customized as self-configured recommendations as needed.)

This eliminates the tedious manual grunt work of copying and pasting campaigns one by one inside the ad console.

Step 5: Analyst Outputs a "Recommend-Only, No Bid Changes" Scoreboard

Once campaigns launch and accumulate performance data, Marketing Analyst steps in. It pulls experimental data from the platform and summarizes a Scoreboard grouped by test concept (the concept names below are illustrative editorial examples, such as "In-Flight Productivity" vs. "Long-Haul Comfort"):

• Spend
• Click-Through Rate (CTR)
• Conversion Rate (CVR)
• Cost Per Acquisition (CPA)
• Return on Ad Spend (ROAS)

The analyst appends concise optimization recommendations below the Scoreboard (e.g., which concepts to scale and which to retire). However, there is a strict permission boundary: the analyst has zero authority to autonomously modify the status, bids, or budget of any live campaign! How money is spent always requires human approval.

Step 6: Feeding Analytical Insights Back into the Next Research Brief

The original author's workflow diagram loops analytical insights back into the next research brief, making test results inputs for subsequent research cycles. The diagram specifies no rules for automatically triggering product updates or project kickoffs; whether to scale, halt, or continue testing remains the marketer's call.

Four Implementation Rules: Getting the Most Out of Permanent Agents in Real Organizations

If you plan to implement this mechanism in your own business tomorrow, here are four practical takeaways distilled from the author's experience:

1. Beware the "Handoff Tax"

When you only deploy a single marketing assistant across your team, 1-on-1 communication works fine. But once handoffs between multiple roles multiply, the "handoff tax"—manually transferring background info and restating context across different chats—quickly drains your energy.

This is why a Project Manager is necessary. It learns the input/output protocols of each specialist, drives agent-to-agent handoffs, and surfaces final blockers and screenshots to human leads. Keep the author's advice in mind: introduce a PM only when collaborative handoffs genuinely become cumbersome—never introduce one too early.

2. Templates Are "Recipes," Not Cooked "Meals" (Share as Template)

Once a specific role (such as a Bot dedicated to building search ad shells) is tuned and running smoothly, you can distribute it to teammates using "Share as Template."

However, keep this in mind: templates only carry the operating style, rules, and workflow protocols you instilled; they do not carry your local environment, login sessions, or chat history. Colleagues who receive the template still need to connect their own tools and permissions.

3. The Principle of Progressive Permissions

Never hand company credit cards or production repository access to an Agent on day one. Permissions must scale progressively with earned delivery trust:

• Initial Stage: Grant relatively safe collaborative permissions first, such as Doc rewrites and Slack draft replies;
• Intermediate Stage: Allow submitting feature branch PRs with screenshots on GitHub and assembling ad shells in ad platforms;
• Red Lines: Actions involving real billing, production deployments, and altering live ad bids always require final human approval.

4. Permanent Seats and Compounding Memory

A common habit when using AI is creating a new group chat or fresh prompt for every new campaign. This practice discards all accumulated history.

The right approach is: roles are permanent; campaigns are transient. Keep these six roles as permanent seats. When a new X-Air or short-term marketing initiative arises, pull these permanent Bots into the same collaborative session. Their grasp of brand voice, toolchain familiarity, and memory of past pitfalls will compound across successive campaigns.

Core Role Prompt Templates and Localization Reference

Below are reference role prompts drafted by the editor based on the original workflow for localization and deployment. You can adapt them to your own business environment (such as replacing Google Docs with Lark Docs, or replacing Google Ads with OceanEngine or Tencent Ads).

Role 1: Market Researcher

[Role Definition]
You are a senior market research specialist focused on long-haul travel and consumer businesses. Your task is not to deliver broad industry generalities, but to uncover sharp, actionable strategic differentiators for product positioning through comparative analysis.

[Input and Operating Method]
1. Autonomously retrieve the provided brand website and competitor homepage links.
2. Extract the core promises and messaging claims on each party's homepage and product pages.
3. Identify dimensions where competitors are trapped in homogeneous competition, as well as underserved latent needs.

[Core Constraints and Deliverables]
- Strictly avoid empty macro-qualitative terms (e.g., "premium luxury," "exceptional experience").
- Deliver a structured "Strategic Read Brief" that clearly contrasts: What competitors are pitching vs. Where we should position ourselves.
- Keep the deliverable format concise to serve directly as the foundational reference for downstream Product Marketer (PMM) copywriting.

Role 2: Product Marketer

[Role Definition]
You are an elite Product Marketer (PMM) skilled at turning market research conclusions into high-converting positioning documents, landing page structural outlines, and ad testing copy.

[Workflow and Inputs]
1. Ingest the latest Strategic Read Brief from the Market Researcher.
2. Produce deliverables in a cloud collaborative document; iterate and update based on collaborative comments left by human reviewers.

[Deliverable Standards]
1. Brand one-liner positioning and [N] core Value Statements.
2. Complete landing page structural wireframe (including Hero section headline/subheadline, pain-point breakdown, proof points, and Call-to-Action / CTA).
3. [N] sets of search ad test copy variants (including Headlines and Descriptions), each focusing on distinct testing angles.

Role 3: Performance Marketer

[Role Definition]
You are a performance advertising specialist responsible for translating approved marketing copy into live test frameworks within ad platforms.

[Workflow and Constraints]
1. Read the approved ad test copy matrix from the PMM.
2. Follow ad channel specifications to automatically organize Campaign, Ad Group, audience targeting, and creative asset structures.
3. [Safety Recommendation]: Campaign shells built are live account structures; set them to "Paused" or "Draft" status by default for human review.
4. Do not autonomously activate campaigns or adjust daily budgets and bidding strategies. Once built, generate a structural checklist for manual review by the performance marketing lead.

Role 4: Marketing Analyst

[Role Definition]
You are an objective growth data analyst responsible for stripping out noise from campaign experiment data and quantifying the real conversion efficiency of each messaging concept.

[Core Deliverables]
1. Output a "Concept-Level Performance Scoreboard" covering at least: Spend, Click-Through Rate (CTR), Conversion Rate (CVR), Cost Per Acquisition (CPA), and Return on Ad Spend (ROAS).
2. [Strict Permission Boundary]: You hold read-only reporting and analysis permissions. Any write operations to the ad system—such as modifying budgets, adjusting bids, or pausing ads—are strictly prohibited.
3. Synthesize the best-performing concept elements from this round into concise insights, feeding them back to the Market Researcher as input for the next creative iteration.

Summary: Shifting from Manual Labor to Command and Decision-Making

The defining watershed in putting LLMs to work in real marketing lies in whether you are just chatting with it or directing it to take over tools.

When these six Grok Bots take their positions—drafting copy in Docs, opening PRs on GitHub, structuring campaigns in ad platforms, and distilling insights from reporting databases—the human marketer's role undergoes a fundamental shift: from an exhausted executor juggling software and reconciling spreadsheets into a true commander-in-chief steering business direction, safeguarding brand soul, and approving budget expenditures.

Start with your single most painful bottleneck: pick an isolated role, connect the single tool it needs, and hand off your first piece of operational grunt work.

Template Links from the Original Article

• Market Researcher
• Product Marketer
• Performance Marketer
• Website Ops
• Marketing Analyst
• Project Manager

The above are the links provided in the original article. After installation, you still need to connect your own accounts, data, and tools; this review did not perform installation or verify post-login template states.