How AI Changed the Evolution of Marketing Software Development

It’s Monday morning at a B2B SaaS company. The growth lead opens the dashboard and trial signups look like they fell off a cliff. Trials didn’t actually drop. An AI agent someone wired into the CRM last week started rewriting lead statuses on a schema it didn’t fully understand, and now half the signups are filed under a stage that doesn’t exist. The lifecycle emails are firing off that bad data. The number the CMO is staring at is just wrong, and the chart kept drawing anyway.

All of that is software, not marketing, and it never stops.

Marketing tools have always broken. What’s new is the cause: an AI agent doing work that used to take a small team, wired in by someone who isn’t an engineer, running against data nobody owns. That’s the whole story of where marketing software development went in the last two years. These days the marketing stack changes at the speed of whatever your team wired up last Tuesday.

I’ve watched software reshape marketing for twenty years, and I helped build one of the earlier waves of it. So when I say AI changed the job, forget the LinkedIn doom version where everyone’s out of work by Friday. The real change is more specific, and a lot more useful if you’re the one deciding who to hire.

Marketing Software Now Changes Faster Than You Can Buy It

For a long time, the marketing “stack” was a mailing list and a mail merge. Then email service providers showed up, then the first CRMs, then marketing automation, then a landscape so crowded it stopped being funny. Scott Brinker’s marketing technology landscape counted 15,384 tools in 2025, up nine percent in a year and roughly a hundred times bigger than it was in 2011.

You don’t need to memorize that number. Just count how many of those tools are billing your card right now. A lot of companies run somewhere between twenty and a hundred of them. That’s a small software operation, and most companies staff it with nobody.

Each of those earlier waves took years to arrive and years to absorb. A company adopted a CRM and lived with that decision for most of a decade. The pace was slow enough that marketing could treat software as something you bought, configured once, and forgot.

AI broke that rhythm. The tools change month to month, the good practices are six weeks old, and the thing you can build yourself this quarter was a four-engineer project last year. The job quietly turned from buying software into building it, and that’s a different muscle entirely.

I’ve Lived Through One of These Inflections Before

I’m Matt Watson, and I’ve started four companies. The one that matters for this story is VinSolutions, the CRM I co-founded that became the number one CRM in the automotive industry and sold for around $150 million. We bootstrapped it to $35 million in annual revenue with no outside funding before that exit, which mostly means I couldn’t hire my way out of any problem and had to build my way out instead.

VinSolutions was marketing software. It ran dealership email campaigns, tracked leads, scored them, and tied all of it back to who actually bought a car. Building that in the 2000s meant real engineering: data models, integrations with systems that did not want to be integrated with, deliverability, and reporting that had to be right because a dealer was making payroll decisions off it. Even then, before anyone said the word “martech,” the work that decided whether the marketing worked was software work. I ended up owning it because the job was engineering and I was the engineer in the room, not because I wanted it.

That’s the part I want you to take from my resume, not the exit. I watched software swallow marketing once already, and AI is doing it again in a fraction of the time.

What AI Actually Changed for the Marketing Stack

For most of the last decade, the sane move for marketing technology was to buy. Building your own data platform or attribution model meant hiring four engineers and waiting nine months, and by the time you shipped, a vendor had eaten the feature anyway. So you paid the bill and moved on.

That math changed. AI-assisted development means one capable developer can now ship a working version of things that used to take a team. Adoption backs this up: about 85% of developers regularly use AI tools as of JetBrains’ 2025 survey, and at Google, Sundar Pichai said 75% of the company’s new code is AI-generated, with engineers now supervising agents instead of typing every line. The build vs buy decision for martech finally has a real answer beyond “just buy it.” For a growing number of companies, it’s to build the parts that matter and buy the rest.

I’ll use my own company as the example, because I’d rather show you a real one than describe a hypothetical.

At Full Scale, our engineering team rebuilt our entire website as a custom application with Claude instead of running it on a WordPress theme like we used to. That was a real rebuild done by engineers using AI, on a timeline that would not have been possible two years ago.

Then we kept going. We’ve built more than 40 AI “skills,” small automated tools that run pieces of our content marketing, from research to drafting to publishing. We built an AI agent trained on my own newsletters, LinkedIn posts, blog articles, and my book, so it can draft marketing content that actually sounds like me instead of sounding like a press release wrote itself. And when we wanted a data report on the state of the industry, we asked Claude to deeply research the whole thing and write it, and our State of IT Staffing guide is the result, original data and all.

None of that came from a SaaS subscription. We built it, because we could, and because the off-the-shelf versions either didn’t exist or didn’t fit. The ability to use AI to create and run marketing content right now is genuinely unlike anything I’ve seen in twenty years of doing this.

Here’s the catch, and it’s the reason this is a hiring story and not a “fire everyone” story. AI made building the marketing stack cheap. Running it without quietly breaking things is still hard, and still needs a person. Someone has to understand the data model the agent is writing to, catch the webhook that fails silently, and notice when the AI confidently ships something nobody asked for. Marketing teams are pouring into this fast: 62% of organizations are experimenting with agentic AI and another 23% are already scaling it. The agent handles the easy 80%. The 20% that breaks is where the revenue leaks out, and closing it takes a developer.

The obvious pushback is that better AI tools and no-code builders should shrink that 20%, not hand it to an engineer. They don’t. The seams between tools, and the data those agents quietly write to, are exactly the part no single vendor’s AI owns. Someone has to hold the whole thing together, and that someone writes code.

The Job This Creates: Marketing Needs a Developer

When a marketing dashboard breaks like that, most companies reach for another marketer. That hire won’t fix it. The work that decides whether a campaign converts is the pipeline behind it: the integrations, the attribution, the event tracking, the automation, and now the AI agents everyone wants to bolt on. There’s also the first-party-data plumbing that replaced third-party cookies: server-side tracking, consent handling, customer data platforms. The headline and the offer matter far less than all the plumbing that carries them. Someone has to build and maintain it or it rots.

That someone is a developer, and the role has a name now: the MarTech developer, a software engineer who builds and maintains the marketing technology stack instead of just running the tools. What the role does day to day, the skills to screen for, and how to tell whether you actually need one is its own subject, and I cover it in that post.

One thing matters more than any technical skill, though: whether the developer will talk to your marketers. Software is about communication, a point I make constantly because it is the thing most hiring gets wrong. It is the philosophy behind Product Driven, the book I wrote on building software people actually use.

Who Actually Builds and Runs It

So who builds all this? There are three honest ways to staff the role: a senior US hire when the work is core to the business, an agency for a project with a fixed end date, or a dedicated offshore developer for the ongoing version. I weigh the tradeoffs and the real cost of each in what a MarTech developer is, and whether you need one. The short version: for continuous work, offshore is the structural fit, as long as you do not cheapshore and buy the cheapest person instead of the best one.

That is how we build teams at Full Scale: a dedicated developer who works inside your marketing team for the long haul, not a vendor you file tickets with. Our marketing software development page lays out how we staff it, inside our broader staff augmentation model.

The bottom line

AI ended the days when you could treat marketing software as something you bought once and forgot. The stack changes faster than any purchasing cycle, and the companies that keep up are the ones that can build their own. That takes a developer embedded with your marketing team.

Customers don’t buy cool code. They buy cool products. The marketing teams that ship are the ones with a developer behind them.

At Full Scale, we put a dedicated MarTech developer on your team for the long haul, in your standups and on the hook for the same numbers you are. Book a discovery call.