My AI Flow State: How I use AI to transform my marketing workflow

Most marketers are still figuring out how to use AI. We are in as much an experimentation phase as the engineers building the tools we’re using. It’s one of the most exciting times in the history of our industry.

As part of my own experiments, I've developed what I call my "AI flow state." It’s a specific workflow that uses multiple AI tools in sequence to create marketing strategies and tactics that actually work.

The result? I get 30% to 40% further on any project before I need to step in and finish the work myself. That might not sound like much, but it's revolutionary. It means I can handle three times the workload, or spend three times as much energy on the details that really matter.

Here's exactly how I do it.

The foundation: My marketing philosophy as AI constraint

Before I dive into the tools, I need to explain something important. I wrote a book called Picks and Shovels: Marketing to Developers During the AI Gold Rush that captures both my writing style and my complete marketing philosophy. This book isn't just sitting on a shelf it's the secret weapon that makes my AI workflow actually work.

Every AI tool I use gets constrained by this book. When I prompt ChatGPT, I reference specific chapters. When I ask Perplexity to research, I give it my framework. When Claude writes for me, it follows my voice and approach. This constraint is everything. Without it, AI gives you generic output that sounds like everyone else. With it, AI becomes an extension of your own thinking.

Most people make the mistake of asking AI to be creative or original. That's backwards. AI should be focused and aligned with your existing expertise. It should amplify what you already know, not replace what you think.

I like to think of using AI as akin to a wind-assisted marathon runner. You still need to train and condition yourselves for the race. But a little extra at your back can’t hurt.

The core workflow: ChatGPT to Perplexity to Claude

My main AI workflow follows a simple three-step pattern. Each tool has a specific job, and I never try to make one tool do everything. I also don’t adhere to these steps in sequence. Some projects call for different sequences, while others may not require the use of secondary and tertiary tools at all.

Step 1: ChatGPT for organization and prompting

I start every project with ChatGPT. Not for writing. ChatGPT is terrible at writing in my voice. But it's excellent at organization and structure.

I use ChatGPT to create project documents, scope outlines, and detailed prompts for blog posts. I'll tell it about the project, reference relevant sections from my book, and ask it to break everything down into clear steps. ChatGPT excels at taking messy ideas and turning them into organized plans.

For example, if I'm writing a blog post about email marketing, I don't just jump into research. I first ask ChatGPT to create a scope document. I give it my email marketing framework from Chapter 7 of my book, tell it the specific angle I want to take, and ask for a detailed outline with research questions.

The output becomes my roadmap. It lists exactly what I need to research, what arguments I need to support, and how the final piece should flow. I have a conversation with the AI to refine the outline, ask for second-level and third-level outlines, further refine those, and ultimately come out with a work product I can export into Notion or Gdocs to continue with.

Again, my expectations are clear: AI is not finishing the work for me. It’s giving me a solid starting point to think through the problem and refine my approach.

This might seem like extra work, but it saves hours later.

Step 2: Perplexity for deep research

Once I have my structured plan from ChatGPT, I move to Perplexity for research. Perplexity is the best AI research tool available right now. It doesn't just give you information it gives you sources, multiple perspectives, and recent data.

I take the research questions from my ChatGPT outline and feed them one by one into Perplexity. But I don't just ask generic questions. I frame every research query within my marketing philosophy. I ask Perplexity to find data that supports or challenges specific frameworks from my book.

This approach gives me research that's immediately useful. Instead of generic statistics about email open rates, I get data about how personalization affects engagement in developer products companies. It’s exactly what I need for the argument I'm building.

Perplexity also helps me find the holes in my thinking. Sometimes the research reveals that my initial angle won't work, or that there's a better story to tell. It's much better to learn this during research than after you've written 1,500 words.

Step 3: Claude for first draft writing

After I have my organized plan and solid research, I turn to Claude for the actual writing. Claude is the best AI writer I've used, especially when you give it the right constraints.

I provide Claude with three things: my outline from ChatGPT, my research from Perplexity, and specific instructions about my writing style from my book. I tell it exactly what voice to use, what frameworks to reference, and how to structure the arguments.

Claude consistently gives me good first drafts. The structure is solid, the main arguments are there, and the flow makes sense. But it's missing the nuance, the personal examples, and the specific insights that make content valuable.

This is perfect. I don't want AI to write finished content I want it to give me a strong foundation that I can build on. Claude saves me from staring at a blank page and gets me quickly to the point where I'm editing and improving rather than creating from scratch.

Can an AI have soul?

I should add that I am a novelist as well. Writing has always been my most expressive, most creative outlet. I won my first writing contest at twelve years old and I’ve always felt it was my one distinguishing talent.

I would never dream of asking an AI to write creative fiction for me.

Creative work comes from the heart and soul, not regurgitated transformations on prior work. My next novel is a deeply personal tribute to my mother. There’s no AI on the planet that can capture the vibrancy of those memories or help me plunge into the metaphors necessary to convey the sights, sounds, and emotions of my past and my view of her.

So, when I say I use AI to help me write project plans, content, and other materials in a work context, I am not saying that it does the work for me. I am saying that it helps me build a thought framework within which to use my experience, instincts, and creativity to build marketing plans that work.

The supporting cast

I often write sales materials and as every marketer knows, when you are asked by sales to write a datasheet or a battlecard, there’s a 94.7% chance no one will actually read it.

So it’s important to build materials in formats people do actually use.

• ElevenLabs turns your words into audio files that people can actually consume.
• NotebookLM can take your materials and turn them into an engaging podcast.

I’m still searching for tools that generate explainer videos and quality diagrams. The 8 second video clips generated by Veo aren’t useful in this context.

Six real-world examples to spark your imagination

It’s one thing to describe a workflow in theory. It’s another to show how it applies in practice. Here are four ways I’ve used the practical guidance in my book coupled with my AI flow state to make me significantly more efficient.

• Sales enablement materials. In Chapter 17, I wrote that “sales enablement equips sales teams with the training, resources, tools, and strategies to effectively engage with potential customers and close deals” . AI helps me skip the blank page. I’ve used it to generate outlines for datasheets, first-meeting decks, and battlecards, then constrained them with my positioning framework so they didn’t sound like generic collateral but also maintained consistency with my messaging. I then created strong first drafts that could be quickly turned into audio or other consumable formats for sales teams.
• Case studies. Chapter 7 explains that “a great case study tells a compelling story about how your product solves a problem of wide applicability for a given customer”. AI speeds the painful part. I run transcripts through ChatGPT to extract themes, then use research tools to bring in supporting data. Claude drafts the narrative, leaving me free to shape the arc and align it with my voice.
• Launch campaigns. In Chapter 15, I noted that “a launch is an opportunity to align the entire organization around a single deliverable. It is the culmination of everyone’s great work” . I use AI to build scope documents, research competitor positioning, and generate first-pass copy for launch blogs, emails, and demo scripts. That allows me to spend my time on the bigger question: what story are we really trying to tell? And the harder problem: getting everyone aligned.
• Vibe coded utilities. In this new era, lightweight utilities often power the real work of marketing. I’ve built dozens of small Python scripts and Next.js apps with AI assistance to automate the chores no one else wants to do: ingesting data into lifecycle marketing tools, auto-generating shortlinks, programming ad campaigns, and cleaning up analytics pipelines. These aren’t polished products, they are utilities. AI helps me spin them up quickly and safely by generating scaffolding code, suggesting database schemas, or even wiring in APIs. What used to require me to bug an engineer for help, now takes an afternoon by myself, giving me leverage to focus on the campaigns and strategies that matter.

And, now the obverse. What about marketing projects where AI isn’t as useful as you could hope? Here are a couple of examples where AI does not meet the bar for me, and never may.

• Planning a user conference. Chapter 8 describes first-party conferences as “an all-company affair and a major milestone, no matter how big or small your company is” . These are chaotic undertakings with dozens of moving parts. I’ve used AI to organize the program tracks, benchmark against similar events, and produce early drafts of keynote copy and promotional material. But more importantly, this is also a great example of the limitations of AI. User conferences are not things that can be automated. They require a significant level of cross-organizational collaboration, and a large, large dose of human soul to get right. A conference with the right vibe cannot be vibe coded.
• Building a long-term strategic plan. Beyond campaigns and collateral, I have used AI to draft the skeleton of multi-year marketing strategies. Drawing on the frameworks in Chapter 16 about alignment with product , I built plans that tie go-to-market to product roadmaps, competitive pressures, and community investment. AI handled the scaffolding, but the real value of strategic planning is bringing your experience and judgment to making decisions now such that they accrue strategic benefit years later. Just as humans cannot predict the future, neither can AI. Ultimately, strategic planning is an exercise in human resiliency: being able to make micro- and macro-adjustments on the go.

The daily reality: AI as creative partner

In practice, my AI flow state has changed how I think about marketing work. Instead of dreading the blank page, I look forward to the creative collaboration with my AI tools. Each tool brings specific strengths to the process, and I've learned to leverage those strengths effectively.

The workflow also keeps me focused on strategy rather than execution. Instead of spending hours on research and first drafts, I spend that time thinking about positioning, messaging, and campaign strategy. The AI handles the tactical work, and I focus on the decisions that actually move the business forward.

This isn't about replacing human creativity it's about amplifying it. AI gives me a strong foundation so I can spend my energy on the insights, connections, and refinements that only human intelligence can provide.

What's next: Evolving the flow state

My AI workflow continues to evolve as new tools emerge and existing tools improve. I'm constantly testing new approaches, refining my prompts, and updating my constraints based on what I learn.

For example, I’d like to build a dedicated LLM for sales materials. Instead of writing datasheets and battlecards nobody uses, why not create a chatbot that just answers questions on the fly, when the answer is needed the most: when a rep is on a call, or researching a customer, or negotiating a deal?

The key is treating this as a system rather than a collection of tools. Each component needs to work well with the others, and the whole workflow needs to align with my marketing philosophy and business goals.

The future of marketing isn't about choosing between human creativity and AI efficiency. It's about developing systems that combine both effectively. My AI flow state is my answer to that challenge, and it's transformed how I work.

As marketers, we must reinvent ourselves just as the world is. We must embrace innovation over tradition. That is how we thrive, now and well into the future.

Picks and Shovels is available now! Get it wherever you buy your books.