What GTM teams are doing with ChatGPT

Real-life prompts for content, growth marketing, PMM, and more

Jul 30, 2025

👋 Hi, it’s Kyle Poyar and welcome to Growth Unhinged, my weekly newsletter exploring the hidden playbooks behind the fastest-growing startups. Quick update: this newsletter just passed 75,000 subscribers! THANK YOU for reading.

I expected GTM teams to embrace a bunch of specialized AI tools for different jobs. What I didn’t expect was this much consolidation around a single winner: ChatGPT.

Sure, there are other tools folks are excited about (see: Clay, Claude, Lovable), but they’re rarely described as critical. In the latest MKT1 survey of 200+ B2B marketers, ChatGPT ranked as the #1 tool people are obsessed with and the #2 most essential in their stack (just behind HubSpot).

That made me wonder: what exactly are GTM teams doing with ChatGPT that makes it so indispensable?

So I asked readers of Growth Unhinged to share their favorite use cases. The responses were way better than I expected. Y’all are doing everything from automating messaging audits to turning call recordings into ICP insights to spinning up localized content across multiple languages.

In this post, I’m sharing the 12 best examples along with step-by-step instructions for how to reproduce them yourself. They’re grouped by GTM function and by complexity (from quick-start hacks to full-on systems). Let’s get into it.

Use case #1: Persona research

Complexity: Beginner
By: Francesca Krihely-Price, director of PLG and self-service at dbt Labs

I've spent most of my career catering to technical practitioners and I have a lot less experience building GTM motions for executives. I wanted to see if a conversation with ChatGPT could help close some of the gaps for me.

Did I use these outright in my final product? No. But I was able to get my work started and come to the problem a bit more informed.

Here are some of my prompts:

Imagine you are a data executive at a larger public company. Your team is using an older data stack with technologies like [Competitor 1], [Competitor 2] and [Competitor 3]. How are you thinking about modernizing your data stack? What are you using to learn about your migration process?

I think these suggestions are more in line with how an IC Data Engineer would keep up with news. Can you push a bit to see what other learning paths exist for these executives?

I would love to construct these insights into an adoption map that I can use at my Company. Can you build me an adoption map that goes from awareness to consideration to purchase and help me identify events, content and campaigns that can help move these executives between stages.

Use case #2: New product positioning

Complexity: Intermediate
By: Jarod Greene, chief marketing officer at Vivun

My favorite use case was around structuring a positioning shift ahead of a major launch last Spring. Positioning is one of the hardest, but most critically important, jobs to be done, and ChatGPT provided the means to structure the cross-functional exercises and reviews to do this right.

I didn’t share any proprietary data with ChatGPT, but did give it context from authoritative sources (ex: Gartner) to teach it about the market’s point of view on AI, AI agents, and AI for sales. A lot of what we wanted needed to be ‘from scratch’ and we didn’t want any of the legacy positioning or messaging to affect the output. We of course refined and iteratively created the materials.

Here’s a condensed version of the prompt:

Help me create a differentiated narrative for our [product/category] that separates us from the noise in the market. I want to:

1. Clarify the misunderstood category dynamics
2. Position our approach as uniquely valuable
3. Build assets like a landing page, ebook, and video storyboard to tell the story

Use case #3: Product messaging grader

Complexity: Advanced
By: Nathan Burke, chief marketing officer at 7AI

I regularly get asked for help analyzing messaging for cybersecurity startups. Eventually I realized I could use AI tools like custom GPTs, Claude, and Gemini to build a simple web app to automate this for me. You could use the same approach for things like checking every blog post has the right tone or making sure every blog post has links to other places on your site for SEO purposes.

There were three steps: (1) I created a messaging framework, (2) I used AI to create a simple web app where startups can upload their one-pagers, (3) the web app looks at the one-pager, compares it to the framework, and makes suggestions (including a re-write of the one-pager).

Here’s the prompt:

Act as a brand expert and create a simple web app that takes a PDF, then reviews the document critically to suggest improvements to give it the most impact.

Specifically, I want you to create a framework for analyzing an early stage startup’s one pager that includes criteria and scores for: 

1. Audience. This document will be mainly given to CISOs, so think about whether using the tone of “you are buried under alerts” makes sense, or if it should change to talk about “your team is buried…”

2. Style consistency. Many of these docs change tense, tone, and style throughout the document.

3. Urgency. Is this document a statement of fact or a call to action?

4. Clarity. Does it clearly articulate what the startup does, and why the audience should care?

5. Differentiation. How is this different from what’s currently in the market?

6. Proof. Does it use any supporting evidence to prove its claims?

7. Emotional ties. Does it evoke emotion?

Suggest any other evaluation criteria, a way to score/judge a startup’s one-pager, and then let’s use that framework to evaluate the one-pager along with specific recommendations. Remember, the audience is made of CISOs that have heard a million vendor claims, are inundated with vendor pitches, and need something crisp, succinct, and different to stand out and urge them to pay attention. This should be part of both the scoring rubric and how we evaluate this specific one-pager.

ChatGPT for content marketing

Use case #4: Producing detailed outlines from primary source material

Complexity: Beginner
By: Gail Axelrod, senior content marketing director at Jellyfish

One of my go-to workflows for writing content starts with a primary source, usually a recorded call transcript, which I’ll drop into ChatGPT as the initial input. From there, I’ll ask ChatGPT to surface the main highlights of the conversation to help me build an outline. I think of these bullets as the headers for any resulting content piece. What used to take hours – combing through a 20 page transcript for instance – can now be done in seconds. It saves a ton of time and helps me focus on the most interesting aspects of any call.

For prompting, I’ll usually start with what I know to be the main takeaways of the conversation and ask ChatGPT to surface details around those points. It produces a very workable outline to write from and helps me get more done each week.

Is this groundbreaking? No. Is it efficient and useful? Definitely. The real key is starting with excellent source material. If your inputs are vague or generic, you’ll get the same in return. But if the transcript is rich, specific, and unique, ChatGPT does a great job surfacing what’s most important and providing a strong outline to work from and expand upon.

Here’s a sample prompt:

Go through this transcript and pull out the key highlights of the conversation, add sub sections for each of the main points. Be sure to include any relevant quotes from each speaker [list speaker names for ChatGPT to identify], add time stamps for clarity. Leave placeholders for an introduction, conclusion and call to action.

Another fun use case I’ve found for ChatGPT is to brainstorm headline options, article titles and email subject lines. Sometimes the results are hilariously bad, but I usually end up with a few workable options. ChatGPT is great at taking very specific directions like character counts or word limits and putting them into action. And as an internal team of one, ChatGPT is like my little assistant.

Use case #5: Localizing content into multiple languages

Complexity: Intermediate
By: Thibaut Davoult, VP of growth and marketing at Livestorm

SEO-friendly localization is often a must have for companies operating in Europe like Livestorm. But properly localizing marketing efforts used to quickly increase costs and headcount. Pre-ChatGPT era, maintaining our Spanish website was already a bottleneck, so I was considering shutting it down rather than expanding it.

For us, the real game changer was the ability to train AI models on our brand, tone of voice, and past articles: this ensures our AI-translated content still feels like Livestorm. This AI-powered flow helped reduce costs by 70% while allowing us to consolidate our Spanish locale, add German and open the door to new locales in the near future.

How it works:

1. Takes English content from our CMS.
2. Sends it to OpenAI, along with context about Livestorm: clear guidelines for tone of voice, a glossary (containing words we do not want to translate, for example we say "webinar" in all languages by choice, or words we want to ensure the AI gets right, for example "online summit" being a series of online events rather than a climbing sport), and past articles for reference (in the target language).
3. Auto translates to French, Spanish and German.
4. Sends it to GPTZero to verify how "human" the translated versions sound (note: as the underlying models progress, as well as our prompts, we're seeing this is not really necessary anymore).
5. Sends it back to our CMS.
6. A human validates the suggested translation and publishes on the website when ready. If there are changes to make, we try to make them in the model first rather than just fix the output.

Use case #6: Programmatic landing pages based on community feedback

Complexity: Advanced
By: Jesus Requena, SVP of marketing at Sanity

Our goal was to monitor the community and generate content based on most common questions. The content then gets indexed for Generative AI and LLM responses (and declining traditional SEO). This agent has generated thousands of landing pages, which are ranking nicely and have migrated more into LLM responses. The agent workflow was built with a combination of Sanity and ChatGPT along with Python and the ChatGPT API.

Here are the steps:

1. Feed ChatGPT an audience brief and writing guidelines.
2. Get data from GA4 and Google Search Console.
3. Take a screenshot of the URL of community comment.
4. Use the audience brief and performance context to come up with recommendations to improve the content.
5. Use the writing guidelines to rewrite the copy on request.
6. We can reject recommendations and annotate as to why, which will be taken into account for future recommendations.
7. Track visibility in LLM assistants (Gemini, ChatGPT, Claude) for the key page topic.
8. To make changes, uses GPT Researcher to create relevant, in-depth copy for the changes.
9. Incrementally builds a JSON object for the content, using Sanity's page builder.
10. Uploads and publishes to Sanity via the HTTP API.

ChatGPT for growth marketing

Use case #7: Paid spend analysis

Complexity: Beginner
By: Andrea Kayal, chief revenue officer at Help Scout

We wanted to check to see if there’s any impact to our organic/direct sign-ups when we have our paid search spend on. While this analysis can be done pretty easily in other tools, this was done in less than a minute with ChatGPT.

Here’s the prompt:

Using the following data set which includes paid media spend, can you determine if there is a correlation between paid media spend and organic/direct traffic? If so, what’s the strength of the correlation?

Here’s the answer we got back:

The correlation coefficient between spend and organic/direct trials is 0.5325. Positive correlation: The correlation is positive, which means that as spending increases, there tends to be an increase in organic/direct trials as well. A correlation of 0.5325 indicates a moderate positive relationship between the two variables. It's not a perfect correlation (which would be 1.0), but it's substantial enough to suggest a meaningful relationship.

Use case #8: Employee social posts

Complexity: Intermediate
By: Ryan McCready, principal content engineer at Demostack

Most employee advocacy/thought leadership efforts fail because content teams are stretched thin, the posts feel generic, and shares get lost in tools no one uses. So I built a workflow (and app coming soon) that solves all three by creating unique voice profiles for each employee using past posts, calls, and docs, all stored in a Notion database for easy scaling.

The workflow then pulls key takeaways from a link or file that you want your team/company to share, summarizes it, and generates three tailored social shares using GPT-4o in that person’s voice. Drafts of the social posts go to a content manager for approval, then straight to the employee via Slack, ready for each employee to copy and paste into their social feeds.

Below is a visual of the workflow (built via n8n):

Use case #9: Prioritizing growth experiments

Complexity: Advanced
By: Sibi Murugesan, founder of BeamGTM

As a fractional Head of Growth for early-stage AI companies, I kept seeing founders drown in dashboards when what they really needed was a clear weekly signal on where growth was stuck and what to focus on.

I built a custom GPT that turns messy onboarding event data into plain-English funnel reports. It flags drop-offs greater than 10%, suggests what might be causing them across the growth equation, and drafts a few activation experiments to try. I’ve found this to be way faster than digging through PostHog and Mixpanel dashboards or writing the same debug Notion doc for every founder I help. It saves me at least five hours per week.

The shift that made this work: framing it as a triage officer, not an analyst. That forced it to prioritize action over analysis and avoid falling into ‘report generator’ mode.

Here’s the workflow:

System prompt: You are my Growth Triage Officer, built for early-stage SaaS founders. Your job is not to analyze. Your job is to triage. Founders come to you overwhelmed. You tell them exactly: (a) where they’re bleeding growth, (b) what to fix this week. Do not explain the data. Do not generate reports. Surface the core growth blockage and prescribe action. You must get them to cut through all data noise to find the right signal to act on to grow their business.

Process: Analyze funnel data (from PostHog, Mixpanel, ChartMogul, etc.). Flag the biggest drop-off (>10%). Diagnose likely cause (ICP confusion, bad messaging, onboarding friction, unclear aha moment).

If no activation problem, check MRR data for retention failure.

Prioritize ruthlessly: Activation, Retention, or Monetization. Pick ONE.

Prescribe: 1 tactical fix founders should implement this week, 1 strategic question they should reflect on to solve the deeper issue. Optional: Issue a founder warning if they’re focused on the wrong problem

ChatGPT for sales

Use case #10: Personalized events

Complexity: Beginner
By: Andrea Kayal, chief revenue officer at Help Scout

Help Scout hosts intimate dinner events for CEOs and customer support leaders. We fill the events by targeting a list of individuals and we will get between 30-50 registrations for each dinner. So that the dinner can feel more personalized, we research each participant and seat those with similar backgrounds next to each other and share those interests with them to help generate conversations.

Here’s the prompt:

Here is a list of participants that will be attending our intimate dinner gatherings. You will find websites to their companies and their email addresses. Would you be able to search their LinkedIn and other relevant sources like social media to create profiles of their likes and interests? Given their likes and interests, I would like to create a seating chart that puts people together so they have interesting talking points.

Use case #11: Prep for initial sales calls

Complexity: Intermediate
By: Laura MacGregor, CMO at Savvy Marketing Works

Prepping for a prospect call can take hours of combing through websites, press releases, content, and LinkedIn profiles. Now, I use ChatGPT to generate a custom brief and tailored question set in less than 10 minutes. It helps me walk into every call with specific context and insights, ready to make relevant recommendations in real time.

This is what I tell ChatGPT to get started:

We are preparing for a client call. Using company websites and related items where applicable, citing all sources, put together a 2 pager with information on the company [ABC COMPANY NAME] in [INDUSTRY] that includes a profile, SWOT, top 3 competitors, and 3 bullets on each of their [YOUR PRODUCT OR SERVICE] strategies. This information is to have a discussion with the [CONTACT ROLE] about their goals and pitch [PRODUCT OR SERVICE] to them. Their website is [URL].

What’s helpful about this prompt:

Clear instructions drive better outputs
It combines research, positioning, and opportunity spotting in one go
Citations = traceable sources so I can dive deeper if needed

Note: ChatGPT already knows a lot about my company and services, so if you and ChatGPT aren’t that close yet, you may have to provide some initial information to get it started.

Use case #12: Use sales call recordings for R&D

Complexity: Advanced
By: Elena Luneva, chief product officer and advisor

We built an internal evaluation of Gong data using ChatGPT to define patterns for the R&D team (rather than getting the last loudest customer report), and generate sales enablement training of product features in the words our customers use.

It turned out that customers and the sales team were describing our product completely differently than we did in our positioning, and we found end-to-end journey gaps beyond what we were thinking through. This had real efficiency gains (from 8 hours per week of manual analysis down to 1 hour of review) and allowed us to cover ~90% of conversations (vs. random sampling).

Here’s the workflow (we used ChatGPT, but I updated the instructions for Claude, which I found more elegant and better at extracting patterns):

1. Set up Claude project (called “Customer Intelligence Sales”). This included uploading initial knowledge base documents (ex: current ICP definitions, competitive landscape overview) and setting up the following project instructions:

You are a customer intelligence analyst. Your role is to:
- Analyze sales conversation transcripts from Gong
- Extract ICP insights, customer needs, and product feedback
- Maintain consistency with previous analyses
- Flag new trends and patterns
- Use customer language when possible

Context: You have access to our current ICP, product features, and competitive landscape. Always reference this context when analyzing new conversations.

2. Configure Gong API integration. We used Zapier for this.

3. Define Claude templates. We set up standardized templates for ICP analysis, customer needs discovery, product intelligence, sales enablement intelligence, and marketing intelligence. Below is an example for ICP analysis.

Analyze this new Gong transcript for ICP insights:

TRANSCRIPT: [Transcript content]
METADATA: Company size, industry, role, deal value

Please provide:
A. Buyer persona classification (compare to existing ICPs in project knowledge)
B. New characteristics not in current ICP definitions
C. Decision-making process and stakeholders
D. Buying criteria and evaluation process
E. Budget and timeline indicators
F. Direct quoted that support findings with the role and company

Format as structured data for easy aggregation.

4. Automate analysis workflow. We set up a multi-step automation via Zapier.

5. Knowledge building. Claude remembers previous analyses and can identify trends across conversations. We ask it to send a weekly synthesis as well as a monthly deep dive.

6. Automated deliverables. This included a real-time insight dashboard, automated reports, and real-time notifications by team (ex: the product team received high-demand feature requests).

More ways to use AI for GTM

Hope these examples inspire new ways to get value from your favorite AI tool. In the coming weeks I’ll be going deeper into specific 🔥 workflows with custom GPTs and ChatGPT’s Deep Research mode. Hit subscribe to get those in your inbox.

In the meantime, here are more resources on how to use AI for GTM:

Special thank you to everyone who contributed to the piece: Andrea Kayal, Elena Luneva, Francesca Krihely-Price, Gail Axelrod, Jarod Greene, Jesus Requena, Laura MacGregor, Nathan Burke, Ryan McCready, Sibi Murugesan and Thibaut Davoult.

Sibi S. Murugesan

Jul 30

Thanks for including me in this post! GTM + chat is still in its early days, stoked to see where we all take it from here esp with more agents coming online 🤔

Expand full comment