The New Martech Stack: Keep the Hub, Build the Spokes

For years, I treated marketing software like a collection hobby. If a platform couldn’t do something, I added another platform. If the reporting was weak, I bought an analytics tool. If the CRM couldn’t send texts or postcards, I connected more services.

That approach worked. It also became expensive, fragile, and difficult to manage.

Now I’m moving toward a different model: keep a few strong platforms at the center, then build the custom tools that connect to them. I call it the hub-and-spoke martech stack.

The practical version is simple: choose an AI hub, a server or hosting hub, and a marketing automation or CRM hub. Let everything else become a custom application, workflow, or integration built around those core systems.

This doesn’t mean every company should cancel its software subscriptions tomorrow. The point is to stop buying a separate SaaS product for every edge case when an AI agent and a good API can often solve the problem more cleanly.

Why the traditional martech stack is getting harder to justify

The old model was built around specialized software. We’d start with a central platform, then add tools for the gaps:

• A texting platform
• A landing-page builder
• An analytics dashboard
• A postcard-delivery service
• A form builder
• An email personalization tool
• A reporting connector

I lived this in my Infusionsoft days. Infusionsoft was a useful marketing automation engine for small businesses, but it couldn’t do everything I needed. I added a texting tool, another platform for analytics, and another service for automated postcards.

The system worked. But by the time I connected all those ancillary tools, the cost of the stack had probably multiplied the core platform’s price by two or three times.

That money might have been better spent on advertising, hiring, or simply staying in the business. More importantly, every additional tool created another account to maintain, another integration that could break, and another interface my team had to learn.

The problem isn’t just the subscription total. It’s the operational drag.

Modern marketers can become more productive while still failing to grow revenue. We can spend all day improving workflows, comparing tools, and building dashboards without addressing the actual bottleneck in the business.

That’s why I’m becoming much more selective. I no longer ask, “Which tool has the most features?” I ask, “Which few systems should I build around, and how easily can the rest connect to them?”

The hub-and-spoke model

In the new model, a small number of durable platforms act as hubs. Custom-built tools and automations become the spokes.

The hubs provide the infrastructure that would be painful to recreate: data storage, authentication, email delivery, CRM records, payment systems, hosting, or campaign management. The spokes fill in the specific capabilities your business needs.

For example, a marketing team might use:

• An AI hub: Codex, Claude Code, or another capable agent that can write, modify, and manage software.
• A server hub: A hosting and database environment where your website and custom applications live.
• A marketing hub: HighLevel, HubSpot, Klaviyo, or another platform that handles customer data, automation, and communications.

Instead of buying five more tools to cover the gaps, you ask whether the gap can be filled with a small custom application or an agent-driven workflow.

That application might pull data from your CRM, generate a report, update a record, create a landing page, or answer a question from your database. It doesn’t need to become a new standalone product. It just needs to do its job.

This is one reason I’ve been paying more attention to which marketing workflows to automate first. The goal isn’t to automate everything. It’s to identify the repetitive work that sits between your core systems and build the smallest useful bridge.

APIs matter more than feature lists now

Features used to be near the top of my software evaluation checklist. I wanted the platform with the longest list and the most polished interface.

That matters less when AI can build the missing feature.

Today, I care much more about the quality of a platform’s API. Can another application read and write the data? Can it create records, update campaigns, manage contacts, trigger workflows, and retrieve useful information?

I don’t need to be an expert programmer to benefit from this. My AI agent can work with the API. I can ask it to evaluate a tool’s endpoints, build an integration, or create a small interface around a function the platform doesn’t provide.

If there isn’t a usable API, I look for an MCP connection. An MCP gives an agent a more direct way to interact with a tool. Instead of manually moving information from one application to another, the agent can operate inside the systems you’ve connected.

This changes how I evaluate software. A platform with fewer features but an excellent API may be more valuable than a feature-rich platform that keeps its data locked away.

Klaviyo is a good example of the direction I’m watching. Its API and MCP capabilities make it possible to manage much more of the platform through an agent. Rather than logging into the application for every task, I can imagine asking an agent to build segments, create workflows, or retrieve campaign information.

The same shift is happening with HighLevel. I’ve been frustrated that some actions were available through the API while other important actions—especially building automated workflows—were not. Browser automation could handle those tasks, but it was slow, expensive in agent credits, and less reliable than a direct connection.

As platforms expose more of their capabilities through APIs and MCPs, the agent becomes the working layer between the marketer and the software.

The three hubs I’d start with

1. An AI hub

The AI hub is where you describe what you want, inspect your data, write code, and coordinate the other systems.

For me, that means tools such as Codex or Claude Code. The exact choice will change quickly, so I wouldn’t get too attached to one brand. What matters is whether the agent can work with your codebase, call APIs, interact with your hosting environment, and help maintain what it builds.

This is more than using AI to draft email copy. The agent can become an operating layer for your marketing infrastructure.

You might ask it to:

• Find customers who meet a particular condition
• Create a campaign draft for a defined segment
• Build a landing page from an approved brief
• Review a workflow for missing steps
• Pull performance data from multiple systems
• Diagnose why an integration stopped working

You still need to review its work. I wouldn’t hand an agent unrestricted permission to publish campaigns, alter customer data, or deploy code without checkpoints. But the amount of useful work it can prepare for review is growing quickly.

2. A server and database hub

You need somewhere to host your website, custom applications, and data.

For a larger setup, you might use Vercel with Supabase. That combination is powerful and flexible. For my own projects, I’ve also experimented with Railway because it combines hosting and database management in a way that feels simpler for marketers and designers.

That simplicity matters. The best technical stack isn’t necessarily the one with the most configuration options. It’s the one you can understand, use, and maintain.

With a strong MCP connection, an agent can handle much of the setup: creating a project, configuring the environment, deploying an application, and making changes as the project evolves.

I’ve used this approach to build websites and custom functionality outside the traditional WordPress workflow. I can create landing pages quickly, add a plugin for a specific need, and change the backend interface or styling without waiting for a third-party developer to support the request.

That doesn’t mean WordPress has no place. It means the default choice deserves to be reconsidered. I’ve written more about why the old way of building websites is changing.

3. A marketing automation or CRM hub

The final hub is the system that manages your customer and campaign operations.

For a smaller marketing team, HighLevel may cover a lot of ground. HubSpot is a strong option for larger organizations. For e-commerce, Klaviyo or Omnisend may be the more natural center of gravity.

I’ve looked at replacing these platforms with a collection of open-source tools to save money. Sometimes the numbers look attractive. But once you bolt together several systems for email delivery, customer records, messaging, automation, and compliance, you end up maintaining a stack you never wanted to own.

For now, I’d rather pay a reliable platform to handle the difficult infrastructure—email service providers, phone carriers, deliverability, permissions, and related operations—then customize the parts that make our business different.

Build your own test case before asking the company to change

One of the best ways to understand this model is to build something for yourself first.

That’s what I did with my personal brand. I built a site, deployed it, added custom functionality, and learned where the new workflow was better than my old one. I didn’t begin with a major enterprise migration. I created a smaller environment where I could experiment without putting the business at risk.

That gave me something more useful than a sales presentation. I had a working example.

When I saw similar frustrations on the employer side, I could show the team what I had built and explain how the same approach might work for a smaller internal project. Because I had already tested the stack, the conversation was about a real system rather than a theoretical promise.

If you work in-house, use your personal site or a small internal property as a laboratory. If you’re a consultant, build the workflow around your own business first. Create the landing page, connect the database, test the deployment, and see how the agent handles maintenance.

You’ll learn faster, and you’ll make a much stronger case when it’s time to propose a larger change.

A personal project can also become a personal brand AI can’t simply copy, because the value comes from your judgment, experiments, and accumulated decisions—not just from the software stack.

What the new model can do

Once your systems are connected, the possibilities become much more practical.

Imagine outlining a campaign in plain language, agreeing on the copy and audience, and then asking the agent to build the draft: landing pages, forms, email sequences, segments, and timing. The agent can prepare the work, while you review it before anything goes live.

Email personalization becomes more interesting, too. Many platforms support conditional content or liquid-style templating. A campaign can show different sections based on a customer’s segment, purchase history, interests, or behavior.

Creating that logic manually across a large campaign can be tedious. An agent can generate and check much of the underlying code quickly.

The important distinction is that the agent isn’t replacing strategy. It’s reducing the distance between a strategic decision and a working implementation.

That’s where I see the real leverage. The marketer still decides what the campaign should say, who should receive it, and why it matters. The agent helps turn those decisions into functioning systems.

A practical checklist for simplifying your stack

1. List every paid tool. Include the subscription cost, the person responsible for it, and the workflow it supports.
2. Identify your actual hubs. Which systems contain your customer data, website, content, and campaign operations?
3. Inspect the APIs. Before buying another tool, find out whether your existing platforms can already do the job through an API or MCP.
4. Remove duplicate functionality. If two tools do similar things, decide which one should remain central.
5. Build the smallest useful spoke. Don’t recreate an entire platform. Solve one recurring problem with a small custom tool or workflow.
6. Add human approval points. Keep review steps before publishing, sending, deleting, or changing important customer data.
7. Measure business impact. A simpler stack should reduce cost, time, errors, or bottlenecks. If it doesn’t, it’s just a new hobby.

I’m still early in this transition, and I don’t expect every hub to remain permanent. The AI tools will improve. Hosting platforms will merge capabilities. Marketing suites will expose more functionality to agents. Some of today’s hubs may eventually become spokes themselves.

For now, though, the direction is clear. I want fewer subscriptions, fewer disconnected interfaces, and more control over the custom layer that makes our marketing unique.

The new martech stack isn’t a larger pile of software. It’s a small set of systems you trust, connected by tools you can shape yourself.

Keep the hubs durable. Build the spokes only when they solve a real problem.

And before adding the next SaaS subscription, ask whether an API, an AI agent, and a small custom build can do the job better.

Related Reading

• Is AI Killing the Art of Marketing? Why So Many Are Grieving Right Now

The conversation discusses the evolving landscape of marketing technology, emphasizing a shift from a multitude of SaaS tools to a more streamlined 'Hub and Spoke' model. The speaker shares personal experiences and insights on how to effectively utilize AI and fewer core tools to enhance productivity and growth in marketing efforts. Takeaways Shift to Hub and Spoke model Focus on API and customization Chapters 00:00 The Changing Landscape of Martech 09:30 Building Your Personal Tech Stack 14:52 Essential Hubs for Marketing Success

episode final v1: But right now, I feel like the whole software field is starting to change. I'm using other tools less and less, but there are some tools that are worth sticking to and building around. So today I want to talk about this major shift that I see happening with my personal Martech stack and even the company that I work for is changing dramatically around a few key details and how we're using these tools in order to advance and grow our companies. Welcome back to the AI Driven Marketer. I'm Dan Sanchez. My friends call me Danchez. And if you're having a hard time keeping up with all the changes happening across AI and are just trying to keep straight what you need to know in order to stay productive, well, this is the show for you. You are in the right place. Honestly, we're all having a hard time keeping up with all the updates. which is why I tend to focus on the most practical things that I think marketers and marketing leaders specifically need to know for them and their teams in order to actually grow. 'Cause it's one thing I've seen is the distraction out there across all the tools and all the things is causing a lot of marketers to become more productive, but unfortunately not actually grow revenue, right? 'Cause there's only so many problems that are actually the bottleneck to your growth. So one of those things is often tools. We have a lot of them. We're trying to integrate them. They're a pain to manage. They cost a lot. Like how many SaaS tools are your is your department actually Paying for. Or maybe you as a if you're a solo marketer out there, how many SaaS tools do you even have as a solo person? how many of you paying for? It's probably more than you'd think, right? You total them all up every month and you're like, oi, man, SAS is a lot. But I think that's changing. And it's changing in a few different ways. Mainly it's changing from a SAS collection tool, like you're collecting Pokemon or something, but it's going to a different model, what I'm calling the the Hub and Spoke. model when it comes to tools. So we used to have our huge stack, and I think that stack is simplifying dramatically. Where we have a few core tools and then we're custom building all the applications that fit in between. I remember in my Infusion Soft days, it was all a rage to try to figure out the right third parties to connect with Infusion Soft, because Infusion Soft is a great engine, don't get me wrong, for small businesses, they eventually rebranded to Keep and haven't been able to keep up. So I'd think they they've kind of fallen away. For the most part, is a small small business marketing solution. I'd say high levels kind of taken taking up a lot of their market share and others. But back in the day, it was one of the best marketing automation platforms or small business around. Shoot, I mark I automated a whole college admission system with it. And but it couldn't do a lot. It couldn't do text messaging. So we'd have to bring in fix your funnel and it became a really robust texting thing. We didn't have good analytics, so I brought in another tool to be the analytics. I I wanted to be able to do automated postcard delivery, so I had to bring in a tool for that in order to hook it up. And I had all these different ancillary tools connected to Infusion Soft as a hub in the middle, and it worked great. But, you know, it costs a lot. It probably multiplied Infusion Soft's cost by two or three X, if I remember right, and you that would just be better money spent elsewhere, either back in my pocket as a solo, or maybe just back into more ad spend where we can actually move the needle a little bit more, or back hiring that next person. So more and more I'm finding those ancillary spokes from the main system like InfusionSoft just don't need to exist anymore, or don't you can custom code your own things. I find that I'm doing this more and more internally with high level, and more and more I'm just building tools. I'm like, a high level can't quite do. This. That's fine. I'll build a solution. It'll just work through high levels API. high level can't do that. That's fine. I'll have another tool that does this and works with high level. Shoot, I want to be able to mine high levels information. Well, I just ask my agent a question and it goes and gets the answer from high level and brings it back to me. Rather than building a dashboard to do it, just my I just ask my agent questions and it finds out. I'm even doing this on the business side within Trim Healthy Mama, a much larger business than my own. and I find more and more we're just using AI as the tool to bridge the gap between our larger tools and we're letting subscriptions go of the smaller tools and building around a few core tools like WooCommerce. And right now we're even transitioning from MailChimp to Clavio. Like I don't know how hot why it took us so long, or an e-commerce solution, you know? Or an e-commerce commerce business. And I think this is going to be the case moving forward. And I think there's a few distinctions, a few ways I now look at tools quite differently than I did before. If I'm considering a new tool, there's some things that have risen to the top of the list as most important. It used to be that the amount of features were really important to me and how well they worked. Not so much anymore. the amount of features do not matter nearly as much as they used to. Why? Because I can custom build them. I can build all my tools. So now like having a robust API has become like top-tier thing. And it's not cause I know how to deal with APIs, it's cause my agent does and I can build into it, right? So API access has become a thing and I'll have ChatGPT evaluate a tool based on its API endpoints and how we can manage it. Another big thing is its MCP. I I I f I tend to like the API better because I can work directly through the agent on that, or I can build a tool that connects through the API. But if not an API, then an MCP so that my agent, like Codex, for example, or Cloud Code for you maybe, can actually just work directly with the tool. More and more I'm finding the tools are becoming more savvy in this way too. Clavio, a new solution that we are pretty close to locking in for Trim Healthy Mama. one of the things that excited me about is that you can manage essentially the whole app through the API or the MCP. So now Clavio, I don't even need to log into. And that's becoming a whole different dynamic for SaaS. Clavio is going to be one of the main hubs, and I'm going to be managing it and building out the automated workflow and building segments and doing all the things I need to do in the app, but I'm gonna be doing it through an agent where it's much faster. In fact, I I was honestly even though I've been promoting high level for a while now on this show, I was getting a little frustrated because you could do a lot through the API through high level, but you couldn't build out the automated workflows. I was like, what the heck? But luckily, I I literally started looking into like, can we like make my own high level? Like how is how hard is it? It's actually pretty hard. It's they got enough going on that it's it's quite hard to duplicate. But just recently, this last this past week actually, they did announce that they were releasing a whole new version of their API and MCP so that you can actually build out automated workflows from your agent. I I don't have access to it, but as soon as I get it, I will be reporting on it and telling you how well it works. Because the only other way to build the workflow is to have your agent open up a browser and do the drag and drop thing, which takes a ton of credits and is not always the most it's just not the most the best solution to be building automated workflows with your agent. But can you imagine that? Already I told you that I'm building all my landing pages and forms and all that stuff with codecs. But can you imagine if you can outline a whole p campaign in your AI agent and as soon as you got all the copy and the emails and all the stuff locked in and your writers and are all in agreement about what's going out when to who, you're like, Okay, agent, go ahead and build the whole thing. Let me know when it needs to be turned from draft to live so I can review it and double check it. Now it can. It's in It's it's insane, mind blowing that you can do that. You can build a whole campaign. Things that used to take weeks, weeks people, in order to build out the whole funnel and all the landing pages and all the variances, and now you can even customize it because a lot of these email campaigns have something called I gosh, I'm gonna blink on it. liquid code, liquid text. I don't know. It's essentially like a more simple solution for coding within email marketing so you can highly personalize it. Not just hi, first name, but like hi, include this section if they're this segment. Insert this segment, insert this segment, or have this in there, plug this there variable in there. It's like a whole custom email so that you can have 10,000 unique emails being sent out to your list of a hundred thousand people. That's possi that was really hard to do by hand. But now with AI agents that can actually use that kind of simple code and email, it's really easy. And you can build sophisticated sequences in just a few moments because your tools can interact with With your agent. So more and more I'm finding this hub and spoke model is something that I'm like now growing into and just seeing helping my marketing take off, but it's also helping the company I'm working for' marketing take off in a way that I never thought possible before. And I think this is the future, guys. this is early, this is early, but I I I think this is where it's going. So I wanted to make an episode just to tell you about it, kind of give you some insights on how I'm looking at it. and I highly recommend. If your company is larger and it's a little harder to explore these things, start experimenting with these things for yourself. Build your own out build out your own little personal brand. That's what I did. I I built my site and I I launched it with Vercel and Neon, which is the database. I've recently moved to something else. I'll talk about in a second. But I built it all out. And then I was like, this is working really cool. I can throw up landing pages fast. I can I can build my own plugins to do whatever functionality I want them to. Shoot, if I don't like the back end, I can be like, hey, let's change the CSS and the back end so it looks prettier. Bam, done. It's so nice. WordPress, I'm like, WordPress is gone. WordPress is dead to me. And now it's that's that's hard to say because I've been ch a champion of WordPress for a long time, but this solution is just so much better. So I built it for myself. And then when I was struggling with WordPress on the employer side, I was like, hey, look what I did for myself. What if we did the same thing over here? Here's the stack I'm using. I talked about it with my CTO. They were already on Vercel. And instead of using Neon for the database, we use SupaBase, you know, very popular JavaScript database solution. And we're launching it there. So I'm duplicating what I did personally and I'm just migrating the whole thing over in my with my employer as well, so that I can have the same benefits with my marketing team. So That's what I highly recommend doing. If your personal brand, if you're an in-house marketer, you should be using your personal brand as a place to experiment with these newer AI tools, test it out, get a feel for it. They're not very expensive at all. it's well worth it. And plus you get to then showcase to your employer, hey, look, I did a little test case with my scenario. What if we took this did a test case internally with this smaller website? It's a lot easier to make the case when they know you've already know how to do it, you have proven results with it. And they can actually play around and see your tech stack in order to feel better about it. No, we're not just reading and got a nice little sales pitch from some sales guy somewhere trying to sell us on a solution. No, we're actually in the weeds playing with it. It's one of the best things to do to learn and grow in this regard. So now let's talk about a few different tech stack options. When I see the different hubs that are out there, there's mainly three major pieces, three major hubs that you need. In order to do this well. If you're a bigger company, you're probably going to need more, okay? At Trim Healthy Mama, we have mutt many more hubs that we kind of sync together into a super database, if you will. but if you're a smaller shopper, have probably like three employees or so in your marketing department, then you probably could get away with just three hubs. You need an AI hub, which is probably going to be codex or Claud Code. You need that. That's the central thing because that's where you're gonna be doing a lot of the work from with the other two hubs. You also need a web host of some kind. Now, with Trim Healthy Mama, we're using Vercell and Superbase. But for myself, I actually moved to a different solution that's a little simpler. Instead of having the two, the host and the database, I found a great solution that has both of them in one, and it's way simpler. And it's kind of like more If you're a marketer or a designer, it's like way more intuitive. There's way less options. It's way less developer focused. Just easier to deal with. Railway is essentially web host and database put together. I don't know why it ever got separated personally, because in my experience working with WordPress, the database and the web host was always the same thing. But in a lot of these AI vibe coding solutions, they've been separated out to two different things. So keep together. you often use a free GitHub account to push to them. And then because Railway has such a strong MCP, your agent could do a lot of the setup. You literally just go in, you you buy a cheap account at ten or twenty bucks a month, give it the give your codex or a claude code the MCP and it could do almost all the setup for you. It's just one of the easiest things to do once you get in there and set up the MCP to your codex. It's it's a wonderful solution. So we have your first the AI hub. You need your server hub. The last one is your marketing automation slash CRM hub. This is one that I looked at trying to even reduce and build off of open source solutions and trying to combine a few different things to save money. And I just found the amount of money you save creates a tech stack that I don't want to maintain. It's just too many bolting together, bolted together solutions that you don't want to have to deal with. it's much worth it to just pay high level or if you're a much larger company, something like HubSpot, though I think high level gets it done f for even pretty good size companies, but HubSpot's a great solution. So solutions like that end up going the distance. Or if you're an e commerce company, something like a clavio or an omnicent work quite well. But the point is you can get away with just those three solutions. AI hub, server hub. And the server is where you're building all your custom applications and your website and everything else. It's the catch all for everything, and then your marketing suite. which has a lot of tools. Like high level on HubSpot, they have a lot of tools built into Clavio, lots of tools in there. But everything else that you needed or needed tweaked, you just code it yourself. And hook it up to your own server and then you're off to the races. Between those three, you don't need much more other guy. I guess a bus your business probably has a Google Suite or a Google Workspace account or something like that, you know? you're pretty much set to go. That will cover almost all the bases that you will need. Again, I am early to this. I think a lot of I'm seeing a lot of other people start to build similar stacks. And I think if you start working towards this now and start working on some personal projects again between your AI agent, 'cause you know you're probably I know you're paying for it already, even with a plus account, you can get a w you can do a lot of coding and build a lot of stuff with it. Start with Railway. Go to dance dot com slash railway. You'll see what I mean when I introduce it and kind of show you what I'm doing. and then either a high level or a HubSpot or some kind of marketing suite, all the spokes become the custom built tools that are built on your own server, whether you're doing it personally or for your business. The AI is only going to get faster, it's only going to get smarter, it's going to get way better at maintaining, debugging, and hardening for cybersecurity reasons all the code that it writes. So I think this is becoming the best solution moving forward. And then who knows? Maybe someday we even take out our high dependence on high level NubSBot. But for now, it's well worth the money to not have to deal with all the ESPs and cell phone carriers and or phone carriers and all the different things. I'd rather just one company take care of all those things. So that's where I'm at. Hopefully you found where I'm at helpful. let me know and definitely go check out danchez.com/slash railway to learn more about that.