The above narrative is everywhere. If you’re a marketer trying to build a career, it’s intense.

You’ll hear time and again - you need to become a systems builder. I’ve even said this myself.

But what does it even mean?

For people from a non-engineer or technical background, it can sound like AI jargon.

I want to demystify it and help people create their v1 marketing system.

My relatively straightforward explanation of a systems builder is: “A systems builder can look at repeatable work and turn it into a system where AI handles parts of the thinking and execution.”

That’s it.

For a marketer, that could be competitor monitoring, campaign creation, customer research, reporting, content creation, preparing for 1:1s, or figuring out what you should work on today.

You don’t need to become an engineer. You need to get good at spotting repeatable work and asking: can AI make this faster, better, or cheaper?

Let’s take competitor research as a simple example: a very common marketing task, and one you can definitely build a ‘system’ around.

The Non-System Approach

All marketing teams have to do competitor research periodically. Check what competitors are doing, open a bunch of tabs, read product pages, case studies, and recent launches on social media, look at their ad creative, dump some observations into a document, do an exec summary, and send it to a bunch of people.

The following month, you have to repeat the same process.

The Systems Builder Approach

A systems builder looks at the process differently. Could I build an intelligence layer that already knows who our competitors are, our ICP, and our positioning? Could I build a skill that continually updates that intelligence with new launches, messaging, case studies, pricing, and advertising? Could I then build skills on top of that intelligence to help with positioning, campaigns, sales enablement, and product strategy?

Now you are not repeatedly doing competitor research. You’ve built a competitor intelligence system that compounds over time.

As a systems builder, you’re constantly asking, how do I make this work reusable?

If you’re a marketer and want to build your very first ‘AI system’, I have a starter one for you.

The Starter AI Marketing System

I’ve built a V1 Starter AI Marketing System that you can copy and customise.

Versions of many of these things live inside my own system. Mine is a little more complex and, at times, wildly over-engineered :)

This is where I would start.

The system has 5 layers:

1. Intelligence: the context everything else reads from

2. More of You: skills that scale how you work

3. Better Thinking: skills that challenge your judgement

4. Product Story: skills that help you understand customers, competitors and differentiation

5. Run My Day: skills that help you focus on the work that matters

I’m also giving you the prompt to build the whole thing.

→ [COPY THE STARTER AI MARKETING SYSTEM]

Let’s go layer by layer.

Layer 1: Intelligence

This is the true game changer and what separates any system you build from a series of prompts. If you want AI to produce consistently good work for you, it needs to understand the world that work exists inside of.

Take the time to build your context/intelligence.

I’d start with 6 core intelligence files you can build

1. Goals

What are you trying to accomplish this quarter? What metrics matter? What are the priorities? Where are you behind? What explicitly does not matter right now?

This becomes incredibly useful because every skill can anchor its recommendations to what you’re actually trying to achieve.

2. ICP

Who is your ideal customer? What problems are they trying to solve? What language do they use? Why do they buy? Why don’t they buy?

Your ICP file should continually improve as you learn more from sales calls, support tickets, interviews and product usage. This is what sets it apart from traditional marketing teams with ICPs stuck in PowerPoints that no one reads.

3. Competitors

Who are the competitors you care about? How are they positioned? What have they launched? What are their latest case studies? How has their paid messaging changed?

Again, this should not be a static document you create once and forget.

The system should keep it current.

4. Positioning

Why should someone choose you?

What do you stand for?

Where are you genuinely different?

What claims can you credibly own?

This becomes the underlying context for a lot of the marketing work you do. In today’s noisy world, your product story is everything.

5. Team

If you’re a marketing leader, who reports to you? What are they responsible for? What are their goals? What are their current challenges?

This becomes extremely useful for management skills we’ll build later.

6. Executives

Who are the people you need to influence? What do they care about? How do they make decisions? What questions do they repeatedly ask?

For some people, this may simply be your manager.

The point is to capture context about the people you need to communicate with.

What this replaces: repeatedly explaining your business to AI, repeatedly uploading docs that are likely outdated for the AI to pull context from.

Once these files exist, every skill you build on top becomes better.

There is one important detail.

I don’t want every skill reading every intelligence file.

That becomes clunky quickly. The model has to hold too much irrelevant context, uses more tokens, and can get confused.

Instead, skills should only pull the context they need.

A competitor analysis skill might read your competitors, ICP, and positioning.

A 1:1 skill might read your goals and team context.

Think of it as routing intelligence through your system.

→ [GET THE INTELLIGENCE LAYER + BUILD PROMPT]

Add an AI coach

The next thing I put inside the foundation of the system is an AI Coach.

In ChatGPT, make sure Memory and Reference Chat History are enabled. In Claude, use Memory or keep this inside a Project where Claude can build context about how you work.

Then give it a recurring job:

“Look across how I’ve been using you and coach me on how I could use AI better.”

A couple of good questions to ask are

1. What are 3 repetitive tasks I keep asking AI to do that should become skills?

2. Which files in my intelligence layer probably need updating?

3. What feature am I not using that fits how I actually work?

Why does this work so well?

Because the model already has a compressed representation of how you work. It has seen the kinds of problems you bring it, where you repeatedly get stuck, what you keep explaining again and what tasks keep showing up.

It does not have a perfect replay of every conversation. Both ChatGPT and Claude selectively retain useful context rather than carrying every historical detail into every chat.

For this use case, that’s fine.

You don’t need an audit log. You need the patterns.

If I keep asking AI to turn messy notes into executive updates, that probably wants to become a skill.

If I repeatedly explain who my ICP is, my intelligence layer is missing something.

If my customer research has started contradicting icp.md, that file probably needs an update.

And if Claude or ChatGPT launches a capability that maps directly to work I already do, my coach can point that out.

Record the way you work

You also need a repeatable way to build skills to help you do your work. The easiest way of doing this is to record yourself doing the work.

Claude now has functionality that allows you to screen record yourself performing a task and talking through what you are doing, then use that process to create a skill.

That means you can simply now just record yourself work and ask Claude to build that into a skill.

Layer 2: More of You

Layer 2 is all about using AI to extend your own capabilities to have more impact. I’d start with 3 skills.

1. People Brief

I want my 1:1s to be a good use of time.

So the People Brief reads from my goals and team context and prepares me before each conversation.

It can tell me what we agreed last time, what follow-ups I personally owe them, how they are tracking against goals, what blockers exist and what wins I should recognise.

Because I’m continually writing context to my intelligence layer, that intelligence layer tells me in advance of these meetings what are the most important things to spend time on.

2. Voice of You

The voice of you skill may seem a little dystopian, but it allows people to interact with a synthetic version of you to get thoughts, opinions, pushback, and clarity before actually coming to you. The purpose of it is to make sure that when they actually do interact with you, they have fully formed questions and can focus on the most important aspects of the problem, opportunity or work.

Because over time your context layer is gathering intelligence about how you make decisions and your work, it can actually get pretty accurate at giving someone a first version of your perspective on their work, their problem, their opportunity, and how they can improve things

3. The Storyteller

The most important part of a marketing leader’s job, but really any marketer, is to continually sell the value of marketing internally. Marketing is one of the more difficult jobs because it’s not always easy to see the return on investment.

You have to be your own best storyteller, and this skill helps you translate your outcomes into stories that you can tell internally so every other team understands the value that you are bringing to that role.

Layer 3: Better Thinking

Some of the best uses of AI aren’t to continually execute and create things. They’re actually to make you a better thinker. The skills in this layer are to help me be more impactful in my work and help actually better fine-tune my critical thinking and how I solve problems, or how I execute as a marketer, a growth leader, and go after opportunities.

1. Exec Panel

Just like you had a synthetic version of yourself for your team, colleagues, people who work with you, this is a synthetic version of your execs, your manager, or the people you have to manage up to. Managing up is one of the most important things you have to do as a marketer, whether it’s to a marketing manager, a marketing leader, or you are managing up to the C-suite execs.

The cool thing about AI is that, because you can capture all of the information the execs are creating internally, you can create a little executive panel: a synthetic version of your execs. You can pitch them things, show them things, and present to them to get feedback before you ever have to do the real thing. Pretty cool use of AI

2. Challenge Me

AI is a great tool to run different frameworks to challenge your thinking. Mine has:

Inversion: What would guarantee this fails?

Pre-mortem: It’s 6 months from now and this failed. Write the post-mortem.

Assumptions: Which assumptions does this strategy rely on that we haven’t validated?

Second-order effects: If this works, what happens next that we aren’t accounting for?

I’d start with one framework you like and just use that one.

Layer 4: Product Story

I deliberately put Product Story into the starter version of this system. If you are a demand marketer, performance marketer or SEO, you may swap this layer for something closer to your own work.

But I think product story is one of the most important things marketing teams need to have right now. Everything is so noisy.

Your product story needs to speak with your customers in a way that sounds different from everything else.

1. Voice of the Customer

The voice of the customer can be built from internal and external data. Yes, we are going back to more synthetic versions of things. The problem with the traditional ICP that most companies use to do some sort of fit and intent model is that it gets outdated pretty rapidly. It lives in a PowerPoint, it lives in a Google Doc, and it lives in some sort of place that no one ever really pays attention to. It gets outdated and isn’t usable.

The cool thing about doing a voice of the customer is that you can actually build a skill that continually updates it from all of the things it’s learning about your customer. And of course, you’re continually recording new learnings about your customers to the intelligence layer.

This is a true synthetic version of your customer that you can actually speak to and pitch new product ideas to, product positioning things like that, and actually get real feedback from your customers.

Now, it doesn’t replace speaking to customers. I cannot emphasise that enough. Speaking to customers is the number one job of any great marketer, but it is certainly a great way to gather research pretty rapidly

2. Voice of the Competitor

The voice of your competitor is pretty similar. This is the research skill that we talked about right at the start of this post. It’s able to go and scrape competitors’ websites, look at their positioning, look at the product updates. It can go ingest paid advertising messaging, ingest the latest case studies, and look at what they’re posting on social. It can actually give you the most important things that you need to know about:

• How are they positioning against us?

• Are there customer segments moving at all?

• Are they changing the value prop?

This gives you a really great perspective on the current competitive landscape that you’re in.

3. Are We Really Different?

Now you can have a skill that reads your positioning, ICP and competitor intelligence and grades whether you actually sound different.

Give you a core from 1 to 10. Some of the fun questions to ask are:

• Show me which claims every competitor can also make.

• Show me where our language has drifted towards the category norm. e.g. it’s just all the same

• Show me where we have a credible position nobody else owns e.g. where can we sound different.

You need a system that helps you fight back against the tyranny of sameness.

Layer 5: Run My Day

The last layer is how I use AI to run my day.

1. Priority List

This skill reads my goals, team context, executive needs and the relevant information coming through channels like email and Slack.

Then it tells me what I should work on.

Mine’s priorities are someone blocked waiting for a decision, a KPI that has moved off target, an important follow-up I owe or something an executive needs from me.

The key is prioritisation. You don’t want AI to just give you everything, you want to teach it what matters to you.

2. What Did I Miss?

This one looks for things I owe people - Emails, Slack messages, Meeting follow-ups, Google Doc comments etc.

Then it groups them by importance and how long people have been waiting and gives me enough context to respond.

3. Blocker Watcher

As a leader, one of the highest-value things I can do is unblock people.

So I want AI looking for signals like people waiting on other teams, missing decisions, missing budget, missing headcount, missing data and repeated slip dates.

Then tell me: Who is blocked? How long have they been blocked? What is causing the problem? What can I personally do to help?

I’ve found this has been hugely impactful.

4. Custom Dashboards

Finally, use AI to change how information gets presented to you. Custom dashboards is a tier 1 use case for AI.

Aggregate the data you care about and build interfaces around the questions you actually need answered. Claude and ChatGPT are perfect for building live artifacts around custom data.

Copy the whole system

This is a good starting system for you to build on top of. You can simple go here, copy the prompt and start asking ChatGPT or Claude or personalise to your needs.

So, when someone online tells you:

Systems builders are going to survive AI.

You should now have a much clearer idea of what that means and a system of your own to amaze them with :)

Until Next Time,

Happy AI’fying

Kieran

That’s just how things worked.

Today, that intelligence is the most valuable asset you have.

Your accumulated knowledge is the number one thing you need to begin monetizing with AI. The frameworks you’ve developed. The marketing campaigns you’ve run. The experiments that succeeded. The experiments that failed. The counterintuitive things you’ve learned. The benchmarks you trust. The mistakes you don’t want to make twice.

For all of us, that experience is something you lived through, at an incredible fast pace, but rarely get all the benefit from.

Some of it we can apply subconsciously, some of it we can prompt to an AI assistant, but so much of it is lost.

AI drastically changes the value of that experience.

Turn it into structured context, and suddenly every AI system you use can reason from what you’ve learned.

So I built something to do exactly that.

I call it my Marketing Intelligence Layer.

And I’m giving you the skill and template to build your own.

This is solving a forever problem. It dates back to Darwin.

Charles Darwin accumulated a huge amount of information over his lifetime. Observations. Quotes. Ideas. Questions. Learnings. Things he’d notice when traveling. Things he’d pick up from conversations.

He kept commonplace books and notebooks to organise what he was learning. It meant ideas captured years earlier could eventually contribute to much bigger pieces of work.

His system mattered because human memory is terrible infrastructure for accumulated knowledge.

Everyone has a version of this problem, including marketers.

You’ve run experiments, campaigns, paid advertising, SEO, brand, product marketing. You’ve built teams. Learned hard lessons across multiple companies.

You know what works, what hasn’t worked. You understand the flaws of attribution. You have strong opinions on the impact of brand and the importance of product positioning because you’ve seen what works, what hasn’t worked.

But where does all of that intelligence live?

For most of us, everywhere.

And nowhere.

That’s becoming a much bigger problem because your AI can only reason from the intelligence it can access.

Every time you open Claude or ChatGPT without that context, it starts cold.

You re-explain your experience. Re-describe your frameworks. Provide context you’ve probably provided 100 times before.

Or worse, you don’t provide it at all.

Then we complain that the output is generic.

So I wanted to make my intelligence portable.

The Marketing Intelligence Layer

The idea is pretty simple.

I wanted one structured artifact containing the transferable marketing intelligence I’ve accumulated across my career.

Frameworks.

Wins.

Losses.

Metrics.

Formulas.

Counter-intuitive learnings.

Operating principles.

All organised by topic and available to whatever AI assistant I’m using.

Here’s the system I built:

The workflow is deliberately simple.

I drop raw material into a folder.

I run /build.

The system extracts the useful intelligence, classifies it, strips confidential company information and adds the transferable learning to my Intelligence Layer.

Then I can give that layer to Claude, ChatGPT, Gemini or an agent I’m building. I can do this by either pointing it at the artifact we’re going to build on my local laptop, or a Google drive folder.

The AI starts with my accumulated context instead of starting from zero.

But that raises the obvious question.

Does it actually make the AI better?

I tested the same problem with and without my Intelligence Layer. A real marketing problem and ran the exact same prompt twice.

One with a clean new AI session. One with my Marketing Intelligence Layer loaded.

First, here’s the prompt:

I'm taking on a new CMO role.
We're trying to grow the pipeline by 100% over the next 12 months for a vertical SaaS AI agent company. It’s a Series B company that is at $50M ARR and wants to double to $100M ARR within 12 months.
Give me the most impactful recommendation you have, one that most CMO’s wouldn’t know and would fail to suggest.
Please start with the TL;DR bullet points.

It’s hard to show in a simple graphic how different the outputs are between these two things, but even the TL;DR does highlight how much better the ‘with intelligence’ version is.

Here’s the summary without my intelligence layer loaded.

Good, but here is it building on top of my intelligence gathered from previous roles, advisory roles, and conversations with smart people.

There is a huge improvement.

Here’s how I built it

There are 4 parts to the system.

1. The Inbox

What it replaces: knowledge scattered across dozens of places.

I don’t want another knowledge-management ritual, another note-taking process, another productivity tool; I already know those things end up in the same scattered mess.

Instead, I drop everything into a /raw folder.

Meeting notes.

Docs.

Transcripts.

Voice memo transcripts.

Bullet lists.

Old presentations.

I don’t need to clean the files, tag them, or decide where anything belongs. The system will take care of that.

2. The Extractor

Running /build tells the skill to read everything in /raw and identify the distinct transferable learnings inside each file.

This is an important distinction.

I don’t want summaries of my documents. I’ve seen people make this mistake time and time again.

I want the intelligence buried inside them.

A single meeting transcript might contain:

• a useful benchmark

• a channel learning

• something we tried that failed

• a framework I used to make a decision

• a counter-intuitive insight

• an operating principle worth remembering

The Extractor pulls those apart into individual pieces of intelligence.

3. The Router

Once a learning has been extracted, the system needs to decide where it belongs.

My default Intelligence Layer covers 8 areas:

Demand Generation
Paid
SEO
Content
Attribution
Product Marketing
Activation
Leadership

But you can add whatever matters to you.

AI in Marketing.

Partnerships.

Community.

Brand.

Developer Marketing.

Whatever you’ve spent your career getting unusually good at.

The important part is that you don’t have to organise it manually.

If the system finds:

“How we structured the paid team...”

That might belong under Leadership, not Paid.

If it finds:

“Changing our activation flow increased...”

That’s Activation.

The routing lives in the build.md skill. It also populates specific learnings attributed to the different companies you’ve worked at. You’ll see this structure in the HTML template as part of the giveaway below.

4. The Confidentiality Layer

It’s going to be fascinating to see how ‘intelligence’ evolves over time. Who owns what intelligence. When working in a company, you have access to proprietary data that you can’t take with you.

Confidential company information belongs to the company. But, the learnings you acquire at that company, the intelligence you gather, is yours to own and capture.

That’s why I built a confidentiality filter into /build.

It strips things like:

• exact company revenue

• internal targets

• employee names

• customer names

• unreleased products

• company-specific sensitive information

But there’s an important distinction between company data and the transferable lesson you learned from working with that data.

For example:

The goal isn’t to create a personal archive of company information.

It’s to preserve the judgment you developed while doing the work.

What you actually get

After /build finishes, I have a single HTML file containing my Marketing Intelligence Layer.

I can open it in a browser.

I can give it to Claude.

I can give it to ChatGPT.

I can use it as context for an agent.

And because it’s mine, I’m not rebuilding my marketing brain every time the hot AI tool changes. You’ll see the template as part of the package I’m giving away below.

You can build mine

I’m giving away the actual system I use.

→ [DOWNLOAD: THE INTELLIGENCE LAYER HTML TEMPLATE]

The template gives you the self-contained artifact that stores and organises your intelligence.

And you can take the skill that builds it:

→ [DOWNLOAD: THE /build CLAUDE CODE SKILL]

Drop the skill into your Claude Code commands folder:

/.claude/commands/

Then run:

/build setup

It asks for your name, your role and whether you want to add custom topics beyond the defaults.

Drop your source material into:

/raw

Then run:

/build

That’s it.

I also built it to work with Google Drive

I didn’t want the system tied to one laptop.

So there’s another version:

/build drive

That finds your Drive folder, reads the source material from /raw and writes the updated Intelligence Layer back to Drive.

Same system.

Now your intelligence is accessible wherever you’re working.

The first version of your Intelligence Layer probably won’t be particularly impressive.

It takes time to build that intelligence. The more you use the system, the better it gets.

The value compounds.

Every meeting.

Every campaign.

Every experiment.

Every failure.

Every framework you develop.

Every time you learn something you don’t want to learn twice.

Drop the source material into /raw and run /build.

Your Intelligence Layer gets richer; it becomes a huge asset.

This is the most impactful way to monetize your experience.

Until now, most of that intelligence has been trapped inside your head.

It’s now feeding your AI assistant.

Build the intelligence layer first.

→ [DOWNLOAD: COMPLETE MARKETING INTELLIGENCE LAYER BUNDLE]

Skill + HTML template + setup instructions. Free, no email required. Available on GitHub

If you build one, I’d love to see what you create. Drop me a note.

Until Next Time,

Happy AI’fying,

Kieran

I’ve recently spent a bunch of time with founders and other marketing leaders. These are my unfiltered notes on the state of B2B marketing now. What’s working and what’s not.

1. Brand / Product Positioning.

The challenge for all brands right now is that AI is consuming all the online oxygen. ChatGPT and Claude alone have created an army of content creators who cover every single feature they launch. Every channel is awash with content about AI and content generated by AI. It’s hard for any brand to make noise.

To stand out, the best companies are positioning around the buyer’s immediate problem state rather than static persona descriptions. What does that mean?

• Slack didn’t position itself as “chat software.” It positioned itself as “where work happens,” creating an entirely new mental category that made feature-by-feature comparisons irrelevant. The result: dominant market share in a space crowded with messaging tools, and a $27.7B acquisition by Salesforce.

• Notion escaped the “note-taking app” comparison trap by positioning itself as a flexible workspace. Instead of competing on features with Evernote and Google Docs, it attached to a broader job-to-be-done—result: 100M+ users and a $10B valuation.

Everyone wants to elevate their brand campaigns and positioning in this way (because all products & features are starting to sound the same).

The person who can do this is in very high demand. The challenge is, there’s a significant gap between the top tier in this role and everyone else. It’s hard to find and hire for this role.

2. AEO (Answer Engine Optimization)

Everyone I talk with is bought into optimizing their content and brand to be more visible in LLMs (predominately Google’s AI overviews & ChatGPT). Most have been doing it for some time.

What I’m starting to see is a growing % now asking, “Wait, what am I getting back for all this work?” Today, all you can measure is referral traffic (people who clicked on links). What I’m hearing is the quantity of traffic you can measure directly from LLMs is < 4% of what you used to get from Google’s blue links. This is a huge change.

The interesting nuance, though: Conductor’s 2026 State of AEO Report found that visitors arriving from LLMs convert at twice the rate, in one-third the sessions, compared to traditional search. And Semrush found that brands appearing in AI Overviews saw a 9% increase in branded search volume within 90 days. So the traffic is smaller, but it’s higher intent, and it’s seeding downstream searches you might not be attributing back to AEO.

The problem is most teams aren’t set up to measure this way. 85% of marketers say AI has changed their search strategy, but only 22% have fully integrated SEO and AI search into a single measurement framework. Most are still looking at the AEO line in their analytics dashboard, seeing a small number, and questioning the investment. The real return is showing up in branded search lifts, higher-converting visitors, and pipeline influence that sits across multiple touchpoints. Much harder to measure.

There is little anyone can do about this other than OpenAI, Google, and other LLM providers. The hope is that, as these platforms begin to expand advertising within their LLMs, they’ll provide more visibility to help customers optimize their ads.

3. Paid Advertising

There is a renaissance in paid advertising. At HubSpot, we invested near zero dollars in paid advertising in our first 10 years. There was a period of time when all companies I spoke with were primarily investing in inbound marketing (SEO & Content). Paid was something they did, but wasn’t held in the same regard. It was something you had to do, but the goal was to limit the amount you spent.

AI has turned this on its head. There is a renewed focus on paid ads.

The best teams are moving to value-based bidding (telling the ad platforms what a good lead actually looks like, not just optimizing for volume). That means ingesting way more first-party data back into the platforms. They are using AI to scale creative testing, generating and iterating on ad variations at a speed that wasn’t possible before. And measurement has become meticulous about attribution, incrementality, and proving what’s actually driving revenue.

This is already showing up in real results. One SaaS company reported that AI-generated ads now drive 60%+ of their Meta spend, without buying a separate tool. An agency ran an $85k side-by-side test across clients comparing AI-generated creative against human-produced creative, and the AI ads performed comparably on ROAS, with significantly faster iteration cycles. The question has shifted from “can AI make good ads?” to “can you afford not to test at AI speed?” This is the lever most companies are pulling right now.

And it’s aligned with where the platforms are heading. Google is consolidating inventory under AI-first campaign types like Performance Max and AI Max. Meta’s Andromeda update strengthened automatic audience creation, conversion prediction, and creative adaptation. The ad platforms are essentially saying: give us your first-party data, let our AI handle the targeting and creative optimisation, and get out of the way.

4. Social/Content

AI has eaten informational content. People don’t want to read your ‘how to’ article. They want to ask an LLM. It means the focus of content has shifted from blogging to social. And like point one, it’s not easy to stand out on social.

The majority of people I spoke with are bought into ‘creator-led’ social, e.g., individuals > brand. That means there’s a lot of resources going towards founder-led content - turning your founder into a personality online / and some select employees.

LinkedIn’s own 2026 guidance makes this explicit: activate “people close to the work” and create “with people, not for them.” LinkedIn is favoring individual voices over brand accounts in the algorithm. This is the trend across all social channels, podcasts, YouTube, and community forums. And now those very things influence how you appear in the LLM results.

Creator-led content strategies are a must-have.

5. Creators

We bought theHustle (for 27m) back in 2021. Our thesis was that B2B would start to look a lot more like B2C over time. That meant the attention would start to divert from brands to individuals in the form of creators/influencers.

This was correct, and the trend is accelerating. Today, creators are a core part of HubSpot’s marketing engine. Nearly every AI-native startup I speak with prioritized their creator program before they began to look at traditional marketing channels like SEO and paid.

The numbers show how fast this is moving. Creator marketing currently accounts for roughly 5% of global advertising spend, but budgets are accelerating; CreatorIQ reports influencer marketing budgets grew 171% year over year, with nearly two-thirds of that new spend reallocated from traditional paid and digital channels. LinkedIn’s own buyer research shows 6 in 10 B2B buyers now discover new brands through creator content, two-thirds use creator perspectives during consideration, and nearly half visit a vendor site after engaging with a creator. Creators now influence a huge part of your buyers’ journey.

The challenge is, demand for B2B creators is accelerating, and the pool of B2B creators to work with is still relatively small. It’s going to result in the cost of creators increasing. And there isn’t a straightforward way to measure the return on creators. Most brands are measuring their return on impressions/engagement, as part of their brand metrics. Others would like to measure on a mix of impressions/engagement and direct conversions. An increasing number are looking for creators to help influence their positions in the LLMs, given platforms like LinkedIn, YouTube, and Reddit are widely used by LLMs to determine their results.

That last point is worth underlining. Creator content is becoming a cornerstone of how LLMs build their answers.

6. Prospecting (AI)

I’m increasingly seeing marketing go further down the funnel. Most commonly through prospecting. This has been a huge AI success story for us at HubSpot, and I see it being replicated at the most AI-pilled companies I speak with. I’d break this into three buckets:

• Triggered Prospecting: this is integrating AI into your triggered email workflows against high-intent actions that result in meetings booked for your sales team. At HubSpot, we were able to drastically improve conversion rates into meetings booked using AI to personalize the copy.

• BDR Prospecting: This is integrating AI to help BDRs increase their coverage by automating emails, generating call scripts, and social outreach.

• SMB Prospecting: A large part of our business is selling into small companies. We don’t have BDRs in our SMB segment for obvious reasons. So, we do the same automation as point 2 for our SMB reps.

Most companies I’ve spoken with are still surprisingly early in how they’re using AI across prospecting.

7. Marketing Leadership

Nearly all founders don’t want a traditional marketing leader. They want a marketing leader who has depth of knowledge in GTM (worked in different motions), depth of knowledge in working with sales, product, and engineering, is incredibly strong in AI, is a systems builder, and isn’t a ‘brand awareness’ CMO. The profile they’re looking for is incredibly hard to find.

And the data backs this up. A recent survey of 75 SaaS CMOs found the top priorities are now pipeline, conversion, ARPU, and retention, not awareness (Gripped, 2025). That’s a different mix than a traditional CMO has been asked to own.

One founder on X put it bluntly:

“AI is changing marketing as fast as eng, so hiring a CMO who can’t IC anymore is like hiring a VPE that can’t code anymore. Personally, I just hired ‘head of’ for demand gen and appointed myself CMO. Much lower risk.” — @calcsam

Founders would rather stay as CMO themselves than hire someone who can’t get their hands dirty. CEO’s want a marketing leader who they believe can do the work.

8. AI (Native)

You’ll need to be strong in AI to get hired onto the marketing teams I’ve been speaking with. All have integrated AI into their hiring process. Many are mandating the use of AI as part of the interview process; e.g., you’ll have an exercise to complete and need to demonstrate how you used AI to solve it.

The job postings tell the story. 78% of marketing job descriptions now mention AI, but only 11% name a specific tool; employers want AI-fluent marketers, not tool specialists (Dexity, 2026). In B2B specifically, demand gen roles lead the way with 50% of postings referencing AI skills, followed by content marketing at 43%.

One hiring manager on X shared exactly the kind of exercise founders are now running:

“If I were hiring a marketer, I would make them build a micro app. I would also require them to complete the job application and prepare for interviews using AI, and provide a link to the chat thread they used to do so. This is the best way to verify that you can trust someone to use AI on the job.” — mfishbein.

Hopefully this rundown was helpful.

Until Next Time,

Happy AI’fying

Kieran

It turned out his account was hacked, but we’re seeing a growing anti-AI sentiment when it comes to using AI to create content. People call it ‘slop’.

Here’s the fascinating thing: people prefer AI-generated content before they know it’s AI.

Bynder ran a blind test with 2,000 consumers. 56% preferred the AI-written copy over the human version. Then they told participants it was AI. 52% reduced their engagement immediately.

People aren’t rejecting AI content because the writing is bad. They’re rejecting it because it feels like you outsourced the thinking. It’s inauthentic. Instead of reading your perspective, you could have just handed them a Claude prompt.

This is the problem with the AI content systems I’ve seen. The system attempts to automate both the typing and thinking.

What people call AI slop is just great writing patterns done at scale. Great copywriters used these patterns for decades. AI regurgitates them at scale.

The patterns aren’t ‘slop’. The patterns are ‘saturated’. Too much of anything becomes irritating. People want originality.

But AI is producing more ‘slop’, just not in the way everyone talks about.

The ‘slop is the perfectly structured post and zero original insight.

A great post needs a genuine idea, something interesting to say. If you have that, perfect writing doesn’t matter. It can be rough around the edges. The problem is, people are using AI to produce saturated patterns without any quality of thinking.

AI is making them lazy. They’re using AI as a copy-and-paste tool. The perfect AI content system still puts the majority of the work on the human.

And this is why the AI content systems that win help the person to craft content with better insights, opinions, points of views that resonate with their audience.

The worst AI content systems cause the person to skip both the typing and thinking.

The ‘Tasty’ AI Content System

To show this, I’m going to show one skill of the updated version of my “Tasty” AI content system. Soon to be released.

The system acts like a content team, available to use in ChatGPT or Claude, with all context kept in a shared layer that is continually updated so no matter where you choose to work, the AI assistant always has your context.

For this post, I’m going to show one of my favourite skills from the system because it highlights the difference in AI helping you craft real content vs. AI creating real slop.

AI for Content Ideation:

This is one of my favorite things to do at the start of the week. I type “give me content ideas,” and the skill does the rest. It loads my audience profile, checks my winning patterns for each platform, then goes out and researches what is trending across Reddit, X, and the web. What comes back is a ranked list split into two sections: ideas mapped to patterns that already work for me, and ideas riding a live trend. Those I like I add to my content queue (claude/chatgpt artifact). Here is how it works.

Step 1 - Load Context: The first thing the skill does is load my content audience profile. This is not an ideal customer profile. An ICP tells you who to sell to. A content audience profile tells you how to create content that people actually react to.

I built mine using a skill called the Content Audience Profiler. It goes out and researches the person you are creating content for. It reviews their footprint on the web, on social, and looks at the creators they are already consuming content from. Then it builds a detailed profile from all of that. I update it every month based on what is actually performing. A small example of what the profile includes:

Their default emotional state: Cautiously impatient. They believe AI is real and important, but they have been lied to enough times by vendors, consultants, and content creators that their default is mild skepticism until proven otherwise. They are not cynical. They are evidence-hungry.

Validation hooks: “You have read a hundred posts about AI in marketing. Most of them were written by people who have never had to hit a pipeline number.” / “Your team says they are using AI. You are not sure you believe them.”

It arms your research skill with the freshest context of your audience’s content consumption habits and what stops them from scrolling.

The second thing the skill loads is my winning platform patterns. This is where it gets interesting.

I have a skill called Learn What Wins. You feed it your past content and performance data, and it figures out the patterns behind what actually works for you. Not what works in general. What works for you, on your specific platform, with your specific audience.

My LinkedIn profile is built from 160 posts over a single year. All of the content. All of the performance data. Ranked by engagement. The same for Substack. The same for YouTube, including full transcripts. Here’s a sample of what the LinkedIn patterns look like:

Top patterns ranked by engagement

1. News drop and your take

2. Builder story (on the ground, shipping, practical)

3. Numbered how-to with clear outcomes

4. Identity reckoning (”how this role will change”)

5. Comedy / satire

6. Contrarian take (”why people are wrong about X”)

Confidence notes: Builder story is trending up over the last 3 months. News drop and your take may be trending down. Comedy has not been tested recently.

The key insight here is that the skill maps ideas to patterns, not specific topics. It is not regurgitating old posts. It is finding new ideas that fit into structural formats that have already earned engagement.

If I ask for ideas across all platforms, it checks all the winning profiles. If I say “give me Substack ideas,” it only checks my Substack patterns. If I say “LinkedIn,” it only checks LinkedIn. The ideas come back engineered for the platform I am trying to grow.

And this data does not go stale. I log my results every month, and the patterns update. What works shifts over time. Builder stories are climbing for me right now. News drops are cooling off. The system sees that and adjusts.

Step 2 - Research Driven by Context Above: Now that the skill knows my audience and what patterns win, it goes out and researches what is happening right now.

This is the part that makes the ideas timely, not just structurally sound. The skill connects to MCP servers that search across the platforms where my audience is already consuming content. Reddit, X, the web. You can add whatever platforms make sense for your audience. It searches the last 30 days for conversations, viral threads, and emerging takes in the topics my audience cares about.

And because it already loaded my audience profile and winning patterns, it knows exactly what to look for.

Step 3 - Match & Rank: What comes back is split into two sections.

The first is Proven for you. These are ideas mapped directly to patterns that have already worked for me. Every idea tells me which platform it is recommended for, which pattern it maps to, and why the system thinks it would work. You can see idea 3 here, “Yes, you are behind,” mapped to Identity Reckoning on LinkedIn. 610 likes, 50 replies on the source post. Idea 6, “One agent won’t get you to real work,” mapped to News-drop + Your Take. 2,618 likes on the source. These are not random suggestions. They are engineered to fit formats that have already earned engagement.

The second is Trending upside. These are ideas pulled from live conversations happening right now across the platforms my audience is consuming content from. Idea 8, “Data fragmentation is the real agent bottleneck,” came from Box CEO, Levie’s viral thread with 1,961 likes. Idea 10, “McKinsey: agents could power two thirds of marketing,” is a clean Data Nugget with a built-in action framework. The system found these because they are trending in my audience’s world and they map to patterns I know how to execute.

The system scores and saves them in my content queue. Stored context I can always view weather I’m in ChatGPT or Claude.

Don’t do the lazy AI thing. James Clear said in Atomic Habits - “You do not rise to the level of your goals. You fall to the level of your systems." Your systems really matter.

Until Next Time,

Happy AI’fying

Kieran Flanagan

The Pangram Labs study (published July 9, 2026) scanned over 1 million posts across five platforms. LinkedIn was the most AI-saturated b…

Read more

Not more productivity. More chaos. A tangled web of ‘AI Sprawl ‘.

Everyone turned into an AI builder overnight. Some did it because it was now a big part of how they got promoted. Some did it because their team was…

Read more

But I still believe there is real value in AI for content, in its ability to build a profile of your audience, better unders…

Read more

I’m one of them. So is Brian Halligan, the former CEO of HubSpot and now partner at Sequoia, who revealed he’d built an AI agent trained on his entire life that now attends his Zoom meeti…

Read more

These have been learned through years of grinding, and iterating to figure out how to scale a personal audience through content.

Where applicable, I’ve added in skills from …

Read more

The title of the post was “A frontier without an ecosystem is not stable.” It sounds like something written for a CTO.

But it’s as applicable for anyone working in marketing.

Buried in the post is the clearest description I’…

Read more

However, Fable 5 does deserve a different conversation. It’s a new class of model - the first Mythos-class model Anthropic has release…

Read more

Almost all of it is being written by a group of people who’ve already made their money, had great careers, and aren’t under the same pressures as people still trying to build one.

I want to tell you why this nar…

Read more

That’s a mistake.

If you integrate AI into how a team works, you’ll get minimal results. I believe you have to rethink how the team…

Read more

What shipped in the las…

Read more

We’ve seen this happen before.

When factories first got access to electricity in the late 1800s, most of them didn’t redesign their operations. They just replaced the steam engine wit…

Read more

I recently took over a group of 400+ people.

I’v…

Read more

“I’m totally overwhelmed by AI. I know you wouldn’t understand, but it’s how I feel.”

I do understand. Everyone working in the industry is overwhelmed. No matter what they’re posting about on…

Read more

Read more

The workers there were, by their own union’s admission, the worst in the American auto industry. Absenteeism ran at 20%. Cars rolled off the line with missing parts. Workers left empty bottles inside door panels to rattle and annoy future owners. GM closed it down.

Two years later, Toyota and GM re…

Read more

Having Ogilvy’s principles codified into a grader that scores your copy in real-time does.

Everyone consumes expert content. Newsletters, podcasts, YouTube. With AI, it’s now possible to instantly turn that content into action.

As Naval Ravikant put it:

“Reading is not the same as understanding. Understanding i…

Read more