Most marketing innovation dies in the gap between “that worked well” and “let’s use this properly.”
You run a pilot. The results look promising. Everyone’s excited. Then six months later, it’s still sitting there as a proof of concept that three people use occasionally. The idea was good. The execution was fine. But somehow it never became part of how the business actually operates.
This is the innovation graveyard that most marketing teams build without realising it.
Door4’s Founder, Leon Calverley, sees this pattern clearly: “I like to think that we’ve always been an innovative operation – it’s in our DNA. We build services that solve problems for us and our clients – and GenAI suddenly adds a new dimension of capability.”
But adding capability isn’t the same as creating value. GenAI makes it easier than ever to build something quickly. The hard part is turning a quick win into something that compounds.
Leon points to a specific example: “Our content planner – which drafted the idea for this article I’m replying to – was the subject of a low-risk, quick idea that has now been built into an almost enterprise grade platform.” meta!!
The journey from “quick idea” to “enterprise-grade platform” is where most innovation projects fail. Because when the technology finally works, no one plans for what happens afterwards.
The content planner began as a lightweight tool to generate article titles for Door4’s scrapbook posts. Nothing complicated. Just a way to speed up ideation.
Then it evolved. It became a synthesis tool. Then it added third-party contribution management. Now it’s integrated into the broader workflow – tracking conversations, feeding data into custom GPTs, connecting to the agency’s centralised data repository.
This is the bit most teams miss. They build the tool. They prove it works. But they don’t think about how it connects to everything else they’re already doing.
“I think it’s important to remember that your tools need to work alongside other platforms and tools that you’ve built or bought,” Leon says. “It’s not hard to build a new tool with its own data – but without consideration of integration, you’re walking into a silo problem.”
This is why so many AI experiments stay experiments. They work in isolation, but they don’t feed into your CRM, your analytics, your reporting, or your team’s actual daily workflow. When something only works if you remember to use it separately, it stops getting used.
Leon’s point about GenAI needing context is particularly relevant: “Ensuring that your GenAI ‘brain’ operating systems have sight of the data and the story that your customers/team are telling.”
If your innovation doesn’t connect to the rest of your marketing stack, it’s not an innovation. It’s just another login to forget.
“Our ability to draft new content ideas, execute topic research, write video scripts has been 10x if not 100x. We can get much more done – and GenAI allows us to enrich with detail in ways that would have taken much longer in the past.”
But speed without standards creates a different problem. Raw output isn’t good enough.
“Understanding that raw LLM output is almost never good enough to align to your brand is central to being a responsible marketer,” Leon explains. “We have devised models for building tone of voice and author styles – both for our agency and for individuals where we help to draft content.”
This is the discipline that separates useful AI adoption from chaotic experimentation. The technology makes it easy to generate. The hard part is maintaining quality, consistency, and brand alignment at scale.
Leon’s advice? “It’s crucial to keep a human in the loop.”
Based on the team at Door4’s experience, three things separate the prototypes that scale from the ones that sit unused:
1. Integration from day one. Don’t build in isolation. Think about how this connects to your existing data, tools, and workflows before you start. If it requires a separate login or manual data transfer, it won’t last.
2. Standards before speed. GenAI can produce content faster than ever. But without tone of voice models, brand guardrails, and editorial oversight, you’ll just create more work cleaning up inconsistent output. Build the quality controls early.
3. Measure what matters, not what’s easy. Leon’s point about ensuring GenAI tools have “sight of the data and the story” is crucial. If you can’t connect the tool’s activity to actual business outcomes – leads, conversions, revenue – it stays a side project.
Most marketing directors ask: “Will this work?”
The better question: “If this works, what needs to happen for us to actually use it properly?”
That forces you to think about integration, training, process changes, and measurement from the beginning. It stops you from building prototypes that succeed technically but fail commercially.
The Door4 content planner story shows what’s possible when you get this right. What started as a quick way to generate article titles evolved into a platform that now manages ideation, contribution, synthesis, and data capture across the entire content operation.
It worked because we didn’t stop at “this is useful.” We asked, “How does this become part of how we work?”
That’s the difference between piloting and shipping. Between innovation theatre and actual competitive advantage.
If you’re sitting on a pilot that proved itself but hasn’t scaled, ask:
If the answers reveal gaps, you’ve found your work. Fill them before you build the next prototype.
Innovation isn’t about how many pilots you run. It’s about how many of them become permanent advantages.
Door4 helps e-commerce businesses turn marketing investment into measurable growth. We combine advertising, experience, and technology to show what works and what doesn’t – so you can spend budgets wisely and prove results to your board.
Our latest articles, features and ramblings.
We explore performance marketing, AI, communications and optimisation.
“Our ability to draft new content ideas, execute topic research, write video scripts has been 10x if not 100x. We can get much more done – and GenAI allows us to enrich with detail in ways that would have taken much longer in the past.”