Marketers, Stop Buying Tools & Start Building Them

For years, I was a SaaS power user. I collected subscriptions, learned every automation trick I could find, and figured out how to make one tool hand work off to another.

That still works. But I’ve started canceling subscriptions.

The reason is simple: AI has changed the economics of building software. I can now describe a process, start with an open-source foundation, and build a tool that does exactly what I need. It’s often cheaper than subscribing to a generalized product, and it fits my workflow better.

That’s the shift I see happening in marketing. The old digital marketer primarily bought tools. The AI-driven marketer increasingly builds them.

There’s one catch: you need to understand the process before you build the tool. AI can write the code, but it can’t reliably decide what a good workflow should be unless you can explain the outcome, the rules, and the edge cases.

That’s why experience with traditional marketing software still matters. You need a baseline for what the tool should do. But once you understand the process, you may not need to keep buying another subscription every time you want to improve it.

Start with a process, not a shiny tool

The temptation is to ask, “What can I build with AI?” That’s usually the wrong first question.

A better question is: “What repetitive process do I understand well enough to improve?”

That distinction matters. Building software from scratch is still difficult. Building on top of an existing, proven system is much more approachable.

For example, I wouldn’t ask AI to create an entire content management system from nothing. There’s no reason to spend thousands of tokens recreating features that open-source developers have already built, tested, and maintained.

Instead, I can use an open-source CMS such as Payload as the foundation. It gives me the underlying structure for pages, posts, media, categories, users, and forms. Then AI can customize the parts that are unique to my needs.

That’s a much better use of AI. It gets a head start from a framework that already understands how a website should be organized, while I focus on the experience and workflows I actually want.

This is also one reason I’ve been rethinking the traditional WordPress model. WordPress still works, and I wrote more about that in why WordPress isn’t dead—but the old way of building websites is changing.

Build a website that does more than publish pages

My current website began as a WordPress site. It has since moved to a custom system built with help from ChatGPT, connected to GitHub, deployed through Vercel, and backed by a database and file storage.

That sounds more technical than it feels in practice. The important part is that I’m no longer treating my website as a static brochure. It has become the home base for several marketing tools.

The site still has pages, posts, images, forms, and a content management interface. If AI disappeared tomorrow, I could still manage it much like I would manage a WordPress site. The difference is that the system can now include custom features without waiting for someone to build a plugin.

That changes the role of the website. It can become:

A publishing platform
A landing page builder
A content repurposing engine
A newsletter workflow
A social distribution system
A database for marketing operations

Those functions don’t have to live in one application. But putting them close together can create useful connections. The website knows which article belongs to which video. It knows the URL to use in a social post. It can access the same content database instead of passing information between six unrelated tools.

Use AI to build landing pages quickly

I recently needed a landing page for a small tool I had mentioned publicly. I realized the page didn’t exist yet, even though people were already beginning to hear about the tool.

So I gave ChatGPT a short dictated prompt explaining what the page needed to do. It created the landing page, connected the form to my CRM, added a thank-you page, and included an interactive preview of the tool.

It wasn’t just a generic page with a headline and a button. The preview let visitors open a card and see how the review process was organized. That kind of small interaction makes a page easier to understand and more useful.

Normally, this would have meant opening a page builder, choosing a template, configuring a form, connecting the CRM, setting up a thank-you page, and checking all the pieces manually. I still tested everything, but the first version appeared quickly.

This is where custom building gets interesting. If you have to keep your primary website on another platform, you can still build a lightweight landing page system separately. Connect it to your CRM through an API, and the page can create and submit its own forms without forcing you to rebuild the same form in your CRM’s interface.

The point isn’t that every landing page should be custom-coded. The point is that the cost of making a tailored page is dropping quickly.

Turn your existing content into a publishing system

One of the most useful tools I’ve built pulls new YouTube content into my website automatically.

When I publish a new video, the system can:

Find the new YouTube video.
Pull the transcript YouTube generated.
Capture the description and show notes.
Create a page on my website.
Generate an article from the source material.
Add the video, links, call to action, and transcript.
Prepare the content for newsletter distribution.

The code was written with AI, but it now runs on its own. I don’t need to open a coding tool every time a new video is published. The website checks for the content, processes it, and creates the page in the background.

That distinction is important. AI helped build the system, but the system doesn’t depend on me having an active AI conversation every time it runs.

I also include the transcript. Humans may prefer the article or the video, while search engines and AI systems may benefit from having the full text available. I’m making the content understandable in several formats instead of forcing everyone into one channel.

But the source still matters. The material begins with me thinking through an idea and speaking from real experience. AI edits, restructures, and repurposes it. It doesn’t need to invent the substance.

That’s the difference between useful automation and a content factory producing empty copy. The strongest input is still original human experience. AI helps you distribute and adapt it.

If you’re trying to create more content from your best ideas, I’ve also written about why good ideas often fail to land—and where AI can help.

Once the article exists, the system can turn it into social posts and schedule them through my CRM’s social tools.

For each piece of content, I can have it create several posts and distribute them at different intervals—for example, one immediately, another three days later, and another ten days later.

I’m not trying to replace the social platform. That would be a lot of work, especially when the existing tool already handles scheduling, analytics, comments, and social listening well.

Instead, my website prepares the content and sends it to the tool through the API. The social platform does what it’s good at, while my custom system handles the part that’s specific to my workflow.

This is a useful pattern:

Build the unique workflow yourself.
Use established tools for complicated infrastructure.
Connect the two with an API.

You don’t need to recreate every product you use. You need to decide which parts create a meaningful advantage and which parts are better purchased.

Open-source software can replace expensive subscriptions

We’re exploring this at Trim Healthy Mama with an open-source collaboration tool called Buzz. The goal is to replace Slack, which costs the company roughly $15,000 per year.

Slack is a good product. But if an open-source alternative covers the basic collaboration needs, we can host it ourselves and customize it instead of paying for a growing collection of add-ons and integrations.

The bigger opportunity is what happens after the replacement. If the tool is open source, we can build on top of it. We could add automated reports, internal dashboards, or AI agents that participate in conversations and workflows.

That kind of customization isn’t always possible—or economical—inside a closed SaaS product.

This is not a weekend project for every company. A major change to internal communications should involve IT, security, migration planning, and user testing. We’re moving content over before making the final switch.

Still, open source gives teams another option. You don’t have to accept every pricing increase or workflow limitation as permanent.

Build a personal CRM around the relationships you actually have

The next project I’m exploring is smaller and more personal: a CRM for managing relationships.

I came across a simple CRM concept aimed at solopreneurs and started thinking about how useful it would be to manage more than business contacts. I’d like a system for professional relationships, community connections, friends, and people I want to stay in touch with more intentionally.

Most of us can maintain a few dozen relationships without much help. Beyond that, people disappear into the gaps. You remember that you meant to call someone, but not when you last spoke or what you discussed.

A personal CRM could track goals for relationships, surface people I haven’t contacted recently, and suggest a thoughtful next action—a phone call, a text, or a useful introduction.

The goal isn’t to turn relationships into a sales pipeline. It’s to make it easier to be attentive when life gets busy.

I’ve seen how much my own life has been shaped by the quality and quantity of my relationships. Podcasting has been valuable partly because it gives me a reason to meet and learn from more people. A better personal CRM could help me maintain those connections afterward.

Know what not to build

Building your own tools doesn’t mean canceling every subscription.

I still pay for ChatGPT, my CRM platform, and Riverside. Riverside handles recording and media workflows that would be far too complicated for me to recreate efficiently. I’ve built some supporting features around clips and social sharing, but I’m not going to spend two years trying to rebuild a mature media product.

That’s the discipline required here. Custom software is valuable when it gives you a better fit, lower cost, or a workflow you can’t buy elsewhere. It’s a poor choice when you’re rebuilding a complicated product that already works.

I think about the decision in three categories:

Build it when your workflow is unique

If your process is specific to your business and existing tools force you into awkward workarounds, custom software may create real leverage.

Build on open source when the foundation already exists

Start with a CMS, CRM, collaboration platform, or other framework that provides the underlying structure. Let AI customize it rather than asking AI to reinvent it.

Buy it when the infrastructure is difficult

Media processing, security, reliability, compliance, and large-scale integrations can be difficult to reproduce. Pay for the mature tool when it saves you from a long and risky rebuild.

The skill that matters most is process design

The future of marketing tools won’t belong only to developers. But it also won’t belong to marketers who expect AI to make every decision for them.

The valuable skill is knowing what should happen.

Can you describe the steps? Can you identify the exceptions? Do you know what a good result looks like? Can you review the output and recognize when the system has made a bad assumption?

Those skills come from doing the work. They’re why experience with traditional SaaS tools is still useful even as the tools themselves become less central.

My recommendation is to find one repetitive workflow you understand deeply. Don’t begin with the most important system in the company. Pick something contained, such as a landing page builder, content repurposing flow, internal dashboard, or follow-up reminder.

Document the process. Find an open-source foundation if one exists. Build a small version. Test it with real inputs. Then improve it one step at a time.

AI gives marketers more control over their tools, but control comes with responsibility. The system still needs human judgment, taste, empathy, and oversight.

The next time you’re about to buy another subscription, pause and ask a different question: “Could I build the part I actually need?”

Sometimes the answer will be no. That’s fine. But increasingly, the answer may be yes—and the tool you build may fit your work better than the one you were about to rent.

YouTuber Qualifications: See What You Need To Get Started