AI as your Marketing Strategist - Complete Workflow

In this post, we'll walk through building a scrappy 90-day marketing plan built for a team of fewer than 10 marketers

Is ChatGPT capable of performing the marketing strategy work of a world-class CMO?

I often hear from marketers that ChatGPT and AI isn't good at strategy. Let's test that to see if it's true.

In this post, we'll walk through building a scrappy 90-day marketing plan built for a team of fewer than 10 marketers. We'll use ChatGPT + Perplexity Labs (if you don't have a Perplexity Pro subscription, you can use all ChatGPT or the LLM of your choice) to do the following:

1. Extract customer and competitor insights

2. Condense those findings into a one-page strategy canvas

3. Spin up three channel playbooks tailored to your ICP

4. Sequence everything into a 30-60-90 day execution roadmap

5. Provide an assessment of this workflow and how it can be improved

By the end, you'll have a prompt playbook that you can copy and paste into ChatGPT to turn it into your CMO-level strategist.

1. Extract Customer and Competitor Insight

I've shown previously how to create an ICP profile using external and internal data - 'The New Marketing Playbook for a Clickless World' and 'How to do Micro-Audience Marketing with AI'.

But this is for a scrappy marketing team so we’ll build an ICP snapshot because it’s likely a smaller company won’t have a lot of internal data on their customers.

Extract Customer Insight

Unstructured data is excellent for AI and to help inform your marketing strategy. In this example, we'll feed ChatGPT ten short customer call transcripts (7 wins, 3 losses) and ask it to create an ICP sheet.

Here is the real AHA when considering unstructured data. Your sales team, chat team, customer support team, etc, are now a great data source. You'll want to have them purposely ask questions so you can capture key data points for your AI initiatives.

Note: I can't use real HubSpot customer data, so I've used AI to create synthetic data to demonstrate how this works.

---

PROMPT:

You are a senior market researcher.

Input = raw transcripts from 7 buyer interviews + 3 lost-deal interviews (pasted below between ***).

TASKS

1. Extract: biggest pains, triggering events, desired outcomes, common objections.

2. Cluster each into 3-5 themes.

3. Identify preferred information channels and buying-process cues.

4. Output a concise **ICP Sheet** with these sections:

• Snapshot sentence // One-liner: who the ICP is + the #1 result they want

• Top pains (ranked) // 3–5 pains ordered by frequency or urgency

• Trigger events // Situations that push them to seek a solution (summarised themes)

• Desired outcomes // What "success" looks like in their own words

• Objections to overcome // Common reasons they hesitate or churn during evaluation

• Preferred channels // Where they research, learn and chat (Slack, podcasts, etc.)

Format the ICP Sheet in markdown.

***<paste transcripts>***

---

This is how it looks.

For a scrappy team, this is a great starting point to build your 90 day sprint around.

Create Wedge Matrix

Next, we'll use Perplexity Pro to scrape public competitor material, surface white-space gaps for your strategy, and attempt to distill them into a statement on what is different about your product and why that will win with your ICP, e.g., obsess about your differentiation.

---

PROMPT:

You are a positioning strategist with live-web access.

GOAL

• Identify my three to five closest competitors.

• Analyse their public positioning (home pages, pricing pages, ads, review sites).

• Surface white-space gaps they leave open.

• Craft a Wedge Matrix and an Only-ness Statement I can use for messaging.

INPUT

Our product: <ONE-LINE DESCRIPTION GOES HERE>

Example: "DataPilot is a SaaS dashboard that gives DTC e-commerce brands a real-time view of CAC-to-LTV across all channels."

TASKS

1. Use web search to list 3-5 direct competitors (name + URL).

2. For each competitor, pull key claims, killer features, obvious weaknesses—cite sources.

3. Cluster weaknesses into 3-5 opportunity gaps (white space).

4. Output the **Wedge Matrix** in this text format:

– Gap #1: <name>

• Evidence: <snippet + source>

• Leverage: <how we can exploit it>

(repeat for each gap)

5. Write a one-sentence **Only-ness Statement**:

"We are the ONLY {category} that {unique edge}, so {ICP} can {benefit}."

6. Finish with three tactical suggestions on how to prove the wedge in marketing copy or content.

STYLE

• Markdown, no tables (bullets and sub-bullets are fine).

• Keep it concise—max 400 words.

---

I used Perplexity Labs for this step. It's awesome. To see the results, check out the video walkthrough (sample doc with some of the insights it captured). Below is just one graphic it created to show 'Key Opportunity Gaps'.

Here is the ‘Only-ness Edge’ it created for me that speaks to our differentiation.

“We’re the only growth platform that unifies sales, marketing, and service on a generous free tier—so scaling companies get enterprise-grade insights without lock-in or surprise costs.”

2. One-page strategy canvas

We now have customer pain points and your product wedge figured out, the next move is to condense into a single sheet (e.g. no over-complicated marketing presentation).

The purpose of the one-page canvas is to ruthlessly focus on who you serve, the problem you solve, why you're suited to win, and what marketing channels focus on.

Paste the two insight docs into the prompt below.

Note: My output isn't indicative of how good this would be, as I used synthetic data for the customer insights. That means the competitor insights and wedge are disconnected, as our customer insights are fake and not specific to HubSpot's actual wedge. Using your actual data will make this much better.

---

PROMPT

You are a senior go-to-market strategist.

INPUTS

---ICP SHEET---

<PASTE ICP SHEET>

---Wyou'reATRIX + ONLY-NESS---

<PASTE WEDGE MATRIX + ONLY-NESS>

GOAL

Create a concise **One-Page Strategy Canvas** that a 10-peteam'sarketing team can execute over the next 90 days.

CANVAS MUST INCLUDE (in this order)

1. **ICP Snapshot** – one sentence naming the buyer and their #1 desired outcome.

2. *you'llem Statement** – “<ICP> struggle to ___ because ___.”

3. **Only-ness Edge** – the one-sentence differentiation pulled from the inputs.

4. **Core Channels (max 3)** – each followed by a one-line rationale.

5. **Hero Offer** – the entry-point asset/product that proves value fastest.

6. **North-Star Metric** – single KPI that signals growth is working.

7. **90-Day Goal** – quantified outcome tied to revenue or pipeline.

CHANNEL LOGIC

• Start with the "Preferred Channels" listed in the ICP Sheet.

• Keep only channels where our Only-ness Edge is easy to demonstrate (e.g., demo-friendly formats for a "real-time dashboard" wedge).

• Assume a lean (<10) marketing team and limited budget; avoid channels requiring heavy creative production or large media buys.

• Rank the final three channels #1 – #3 and explain *why* each fits the criteria in one short sentence.

STYLE & FORMAT

• Write in punchy, business-rea "y language—no jarg" n or fluff.

• Use markdown headings and bullet lists; **do not** use tables.

• Keep the whole canvas to ~200 words.

OUTPUT ONLY "THE CANVAS—nothing else"

---

See example result below (can see in the video walkthrough)

I’ll cover this in the assessment at the end. The channel selection is something I believe can be drastically improved.

3. Create Playbooks for Three Marketing Channels

The canvas tells you what matters; now we'll build channel playbooks to tell your team how to execute. Each playbook is a one-pager actionable plan structured to be run in an agile way.

Each Playbook has the following:

1. Objective – the KPI or constraint (e.g., "< $250 CAC per demo-booked").

2. Signature Tactic – the repeatable move that proves your Only-ness (a live demo carousel, a Slack AMA, etc.).

3. Creative Guard-Rails – hook, POV, CTA, tone.

4. Weekly Cadence – posting frequency, budget envelope, DM quotas, follow-ups.

5. Rapid-Fire Experiments – a backlog of 5 ICE-scored tests; kill or scale each Friday.

---

PROMPT:

You are a senior growth operator.

INPUT

---STRATEGY CANVAS---

<PASTE YOUR ONE-PAGE CANVAS HERE>

GOAL

For every item listed under **Core Channels** in the canvas, create a 90-Day Channel Roadmap the 10-person marketing team can run.

INSTRUCTIONS

1. Parse the “Core Channels” section.

• Each channel name may be followed by a dash and a short rationale—capture only the channel name (text before the dash).

• Treat each unique channel as its own roadmap.

2. For each channel, output the roadmap in this structure (markdown, no tables):

### <Channel Name>

**Quarterly North-Star KPI:** <single metric this channel owns—tie to canvas KPI>

#### Month 1 – Traction Tests

- **Goal:** …

- **Plays (3–4):** …

- **Experiments (3):** Hypothesis – Success metric

- **Exit Criteria:** …

#### Month 2 – Scale the Winners

- **Goal:** …

- **Plays (3–4):** …

- **Experiments (2):** …

- **KPIs:** …

#### Month 3 – Systemise & Optimise

- **Goal:** …

- **Plays (3):** …

- **Experiments (1–2):** …

- **KPIs:** …

RULES

• Derive rationale from the canvas (ICP, Only-ness, resource limits).

• Assume a lean team & scrappy budget—avoid tactics needing heavy production or > $5 k/mo ad spend.

• Keep language punchy; no jargon.

• Output the complete set of channel roadmaps, one after another, and **nothing else**.

---

Here’s an example of the playbook it produced for one of the 3 marketing channels in the Canvas. It splits each channel out across the quarter into Traction Tests (Month 1), Scale the Winners (Month 2), Systemise & Optimise (Month 3).

It will do this for each of your channels. And there you have a 90 day scrappy marketing plan tailored for your ICP and what your key differentiation is against competitors. But, how did it do?

Summary of AI as a Marketing Strategist

The purpose of this post was to illustrate that AI can be a great strategist *if* you do a couple of things well

• Understand what kind of marketing strategy you want. The best thing to do with AI is to show it an outcome e.g. what does good look like and reverse engineer from that. This example was focused on a highly actionable 90-day plan.

• Context is everything. People ask AI to ‘create a marketing strategy for a SaaS healthcare app” and then wonder why it sucks. It has zero context. Even in this example, I haven’t given it anywhere near as much context as it needs.

But, what about the result of this workflow.

It replicated the strategy of an average marketing leader - a C+.

I can make this better and will do in future posts:

• It needs much more context. I’ll do an entire post on the best context to give an AI so it can craft better strategy for you.

• I’m going to work a lot more on how it chooses and ranks marketing channels. In this example, I didn’t give it access to marketing channel data as context. This would have been a game changer.

• It didn’t use the differentiation enough.

• The cool thing is, and I’ll also show this in a future post, you can now build self-hosted apps on Claude. I could easily convert this entire thing to an app for a marketing team to create and manage a strategy custom to their business. Episode incoming!

As mentioned in the video, I like to think some of these posts are me working in the ‘AI Lab’ on concepts and bringing you along with each version. I hope this is useful

Until Next Time,

Happy AI’fying

Kieran