Skip to content

Custom GPTs for Marketing Teams: How We Build Ours

How to build custom GPTs for marketing teams that teach, not just do the work: the 4 GPTs we use and a 5-step process to upskill your whole team.

Hey, Happy Friday.

Last week, I was rewatching Good Will Hunting while my daughter practiced her 3am wake-up call. There's this scene where Matt Damon points to Robin Williams' bookshelf: "You surround yourself with the wrong f'ing books." or something like that.

The line hits me every time I scroll LinkedIn or X - "wisdom" from people who've never shipped anything that actually grew a business. No supporting data, context, or lived experience to back up their claims. Sometimes I worry my team may take these "insights" as fact.

So we're building something to fix it: custom GPTs trained on the right sources. The timeless stuff. The principles that actually work.

And today, I figured I'd share how we're building them.

Why do marketing teams need custom GPTs now?

A writer asked me last week: "what skills should marketers focus on?" Made me think - we've spent a decade training specialists. SEO experts. Media buyers. Social managers. CRO specialists.

But things are changing. The winners over the next 5 years will be generalists. Marketers who understand strategic planning AND brand AND copywriting AND media AND acquisition.

The problem? How do you upskill an entire team without overwhelming them?

I think custom GPTs are a great start.

But not the kind that just do the work. The kind that teach your team how to think about the work.

What custom GPTs are we building?

So far, we've created 4 GPTs:

  • Copy Chief GPT: Trained on Bly's Copywriter's Handbook, plus our best swipe file examples.
  • Brief Builder GPT: Trained on JWT/BBDO brief frameworks + internal templates.
  • Creator Compass GPT: Trained on editorial values, audience, topic/offer alignment.
  • Performance Pilot GPT: Trained on performance principles and internal benchmarks for paid media.

Each one acts like having a senior marketer next to you - asking the right questions, pushing your thinking, then helping you create something that meets our expectations.

How do you build custom GPTs for a marketing team?

Here's how we pull them together.

Step 1: Map Jobs to be Done - Work with your managers. Ask "What tasks do our marketers do weekly? Which ones could a GPT enhance and standardize?"

We want GPTs that make people better, by upleveling their thinking and deliverables.

Step 2: Build Your Knowledge-base - For each GPT, collect:

  • Snippets from trusted books (Bly, King, Halbert)
  • Swipe file examples (top performing creative, strategy, frameworks)
  • Company guidelines (tone, style, benchmarks)
  • Industry research (that's actually proven from trusted sources)

Find sources that your best people reference constantly. Data backed. Trusted. Your flavor.

Step 3: GPT Instructions - Write instructions that coach. This is critical. Your GPT instructions should force collaboration with the user.

For example, our Brief Builder works with the user in a Q&A format and asks follow-up questions:

  • What's the business objective?
  • Who's the target audience? (then ask them to elaborate)
  • What's the key insight driving this?
  • What action do we want people to take?

It challenges weak inputs. Ask "why" and "tell me more."

Only after the discussion does it draft the brief.

Step 4: Test with Your Best - Send the GPTs to 1-2 strong team members. Have them use it for a week on real projects. Ask for feedback.

  • What worked?
  • What felt clunky?
  • What's missing?

Iterate based on feedback. Keep improving the instructions.

Step 5: Launch as a Toolkit - Don't drip out GPTs one by one. Package them as a complete toolkit for the team.

What happens when the whole team levels up

What happens next is pretty cool. Your paid media person starts using Copy Chief to write better ads. Your brand marketer uses Performance Pilot to understand spend and performance metrics. Everyone levels up.

We're creating versatile marketers who use AI to improve across the board. They build range. They become 0 to 1 capable across all channels, while staying advanced in their specialty.

The goal isn't to have the most GPTs. It's to give our team the right foundation to think from - no matter what program or channel they're working on.

Are you using custom GPTs with your team? What's working? What's not?

Hit me up. Would love to compare notes.

Until next time ✌️
Kyle

Questions (FAQ)

How can custom GPTs help a marketing team?

Custom GPTs are a great start to upskill an entire team without overwhelming them. Not the kind that just do the work. The kind that teach your team how to think about the work.

What should you train a marketing GPT on?

Find sources that your best people reference constantly: snippets from trusted books, top performing creative, strategy, and frameworks from your swipe file, and company guidelines on tone, style, and benchmarks. Data backed. Trusted. Your flavor.

How do you write instructions for a custom GPT?

Write instructions that coach. Your GPT instructions should force collaboration with the user. Our Brief Builder challenges weak inputs, asks "why" and "tell me more," and only drafts the brief after the discussion.

What custom GPTs are useful for marketers?

Copy Chief GPT, trained on Bly's Copywriter's Handbook. Brief Builder GPT, trained on JWT/BBDO brief frameworks + internal templates. Creator Compass GPT for editorial values and topic/offer alignment, and Performance Pilot GPT for performance principles and internal benchmarks for paid media.

How do you roll out custom GPTs to a team?

Test with your best: have 1-2 strong team members use them for a week on real projects, and ask what worked, what felt clunky, and what's missing. Don't drip out GPTs one by one. Package them as a complete toolkit for the team.

Sources

  1. Robert W. Bly, The Copywriter's Handbook
  2. OpenAI, Introducing GPTs
  3. IMDb, Good Will Hunting (1997)

How this was made Originally published in Kyle's newsletter, Brand to Demand (Issue 17, June 6, 2025). Written by Kyle Denhoff from first-hand experience growing the HubSpot Media Network. Edited for the web with AI help: headings, the summary and questions were added, and claims were checked against the sources above.

Denhoff Digital
  • Marketing executive on the HubSpot Media Network for nearly five years
  • Helped build HubSpot's media brands, including The Hustle and My First Million
  • Led content for HubSpot's YouTube channel and State of Marketing report

The newsletter

Get one idea a week.

On content, distribution and demand.

Free. Unsubscribe anytime.

Keep reading