What happens when marketing teams start building their own AI-powered tools

Most marketing teams use AI to write things. Blog posts, social captions, email subject lines. 73% of marketing teams now use generative AI (Gartner ), but the vast majority still treat it as a content tool. The teams pulling ahead are the ones that have moved past writing and started building their own AI marketing tools . Working prototypes. Interactive demos. Event activations. Personalised sales materials. Internal platforms. Things that used to need budgets, timelines, and specialist teams that most B2B organisations simply didn’t have.

I’m Barnaby Ellis , Head of Strategy at Fifty Five and Five. I’ve spent nearly 30 years in digital agencies, and the hardest part has always been the same: turning an idea into something someone can actually see and react to. You can describe a concept brilliantly in a deck, and half the room still won’t get it until they can pick it up and play with it. AI hasn’t solved the thinking. But it has removed the resource constraints between a sketch on a whiteboard and a working thing in someone’s hands. That changes everything about how we pitch, how we prove ideas, and how quickly we can move from “what if” to “here, try this.”

From idea to demo in a day: AI prototyping for marketing teams

A few weeks ago, I had coffee with a contact at an existing client. They were new to the company and we hadn’t met before. Through a natural conversation about what we do, I found out they had an event coming up: the Retail Tech Show in London. Off the back of our TCS London Marathon activation , which won a Bronze Drum Award, I offered to go away and think about how we might support with an activation for their stand.

The first step was research. I spent half a day looking into the retail industry and the common challenges it faces. “Returns” stood out immediately. Whether it’s people ordering multiple sizes and sending one back, wearing something once and returning it, or the environmental cost of processing and shipping returns, it’s a problem everybody can relate to. You don’t need to work in retail to understand the frustration.

The game concept came from an old Nintendo handheld device from the late 80s that I remember playing. I wanted to create something that challenged the player to use both hands, each controlling a different part of the game. I sketched the idea on a whiteboard, drew a detailed diagram with labels, and described each label in a brief.

Then I went straight into Claude Code. I gave it the diagram and the descriptions, and within a day I had a playable game.

But I didn’t send it to the client. Not yet.

I took it to Owen Steer first. Owen is a keen gamer, and he brought a pure gameplay perspective: how do you win, how do you lose, do you have lives, how do you score points, does it speed up, do you get rewarded at certain levels? All things that I’d made some assumptions about, but Owen brought that logic immediately.

Then I took it to Fergus Hannant, one of our designers. I showed him the wireframe and a mood board of what I wanted it to look like. He got it straight away, not just the vision, but what needed to be done to bring it to life. In under two hours, he turned my rough prototype into a polished interface: colours, shapes, and animations. It went from something scrappy to a demo I was confident putting in front of the client.

The process is repeatable: research, sketch, build with AI, layer human expertise, test, deliver. I remember working at an agency years ago where we came up with a concept for a credit card campaign, and we had this incredibly skilled art worker who sketched beautiful diagrams of a convenience store set in a traditional Moroccan market. It took ages and they were stunning. Now you can describe that concept and get it back as a working prototype in an afternoon. That’s what’s changed. Not the need for good ideas, but how quickly you can make them tangible.

Why the best AI sales enablement is showing, not telling

The most effective sales enablement isn’t a better slide deck or a smarter CRM plugin. It’s showing a prospect the thing you’re proposing rather than describing it. A working demo changes the conversation from “imagine if” to “what if we took this further?”

I saw something recently from Kieran Flanagan that stuck with me: we are in the era of show, not tell. That’s exactly where we are as an agency.

Back in the day, we would prepare huge slide decks that took days to put together. You’re trying to bring ideas to life in a way that clients can relate to, backed with data. That hasn’t changed. But there’s a moment of truth, the penny drop moment, when you want them to just get it. And that’s where this approach helps. We can describe what we’re doing, but then quite quickly show them the actual thing.

With the game, I sent the client an email with as little information as possible. The whole point is that you should be able to walk up and play with minimal instruction. I gave some context, explained what it was designed to do, and challenged them to beat my score.

They loved it. They played it, told me their score, and engaged with how it connected to their business. Their technology solutions map to every stage of the retail returns problem, and I’d made that connection explicit in the game itself: subtle information highlighting the issues around returns, with gentle nudges towards the client’s technology.

Not a slide that says “we can do this.” A working thing that proves it.

So what happens when these tools become ubiquitous? When everybody has the same AI capabilities, the differentiator won’t be the output. It’ll be whether you built the right thing to address the right problem. AI removes the resource constraints that used to stop you going from telling people about an idea to showing them exactly what you mean. But what you choose to show, and why, still requires strategic thinking.

When to buy AI tools and when to build your own

Not everything should be built from scratch. MIT’s State of AI in Business 2025 report found that purchasing AI tools from specialised vendors or building through partnerships succeeds roughly 67% of the time. Fully internal builds succeed at approximately half that rate (MIT NANDA ). Meanwhile, 42% of companies scrapped the majority of their AI initiatives in 2025, up sharply from 17% the year prior (S&P Global ). The buy-versus-build decision matters because it determines whether your AI investment ever reaches production.

Buy when the problem is already solved. If the market is mature and a vendor is actively investing in the space, there’s no competitive advantage in rebuilding it yourself. AI-powered meeting transcription, email automation, CRM enrichment: these are commodity problems. Buy the tool, integrate it, move on.

Build when the advantage is yours. If what makes the solution valuable is your proprietary data, your unique process, or the way your team thinks about a problem, build it. Your personalisation engine, your content workflow, your client activation: those aren’t jobs to hand to a vendor who serves your competitors equally.

We saw this play out with our Avalara project. We were delivering GEO optimisation work manually each month. Rather than signing a long retainer to keep doing that, we proved the process worked with human delivery first, then built a tool to replace ourselves in that process. The client gets better consistency, faster turnaround, and a capability they own. As I see it, rather than replacing roles within a company, which tends to be what people fear about AI, what we did is use AI to replace ourselves in a process we’d already validated. More teams should consider that approach. If you’d like to explore what that might look like for your business, get in touch .

What does it actually cost to build custom AI marketing tools?

The honest answer: less than you think in tools, more than you think in thinking. I built a working game prototype in a day using Claude Code. That’s from whiteboard to prototype to Fergus’ polished version ready to share. The AI platforms themselves cost very little compared to traditional development.

Where the real investment sits is in the time before you start building. Understanding the client. Researching the problem. Coming up with the idea. Sketching the concept. Defining what good looks like. You also need the right skills on your team to validate the output, because without that, you end up with something that looks polished but doesn’t hold up.

A prototype is fast. Getting the prototype right takes strategic thinking, domain knowledge, and people who can spot when the AI has missed the mark. The investment isn’t in the technology. It’s in the expertise to use it well.

Real examples of B2B teams creating with AI

It’s easy to talk about what’s possible. Here are real projects where teams used AI to build things that weren’t content.

TCS London Marathon: 1,500 personalised finish-line videos in three days

TCS asked us to bring an AI activation to their stand at the Running Show, the expo all runners attend to collect their packs. The solution: capture each runner’s photo, process their likeness in real time using nine different AI tools, generate a 3D avatar, and composite it into a personalised finish-line video delivered to their phone. All before they’d even run the race. 40,000 machines rendering in parallel. Over 1,500 personalised videos generated across three days. BBC News covered it. It won a Bronze Drum Award for Best Use of AI in a Campaign (TCS Marathon activation ).

What Fifty Five and Five built in the time they had was amazing. 1500 runners had a photo realistic AI video created in near real time. The technology really supported what we were trying to communicate on the stand, the TCS Digital Twin program.

Anmol Patel Social Media & Insights Manager, TCS

Avalara: an AI-powered social engagement engine

For Avalara, we built a tool that identifies opportunities across Reddit, LinkedIn, Quora, and Medium where their existing content library can answer real questions people are asking. It analyses conversations, matches them to relevant Avalara resources, and crafts responses grounded in their expertise. The tool covers the full workflow from picking a product to focus on, to finding engagement opportunities, to writing responses. What started as a manual service became a capability the client owns and runs themselves.

TCS Compass: 80+ AI-generated social posts per month

We deployed Compass , our AI-powered SaaS platform, fine-tuned on TCS’s brand and tone of voice. The team now relies on it for social media content and converting video into copy quickly. Over 80 posts generated per month, 60+ hours of production time saved, and 3x faster from ideation to posting (TCS Compass ).

What changes when everyone on your marketing team has AI

Roles blur. The strategist builds prototypes. The designer writes code. The marketer interrogates data without waiting for a report. When everyone on the team has AI tools, the traditional lanes break down.

Once upon a time, everyone stuck to their lane. You’d do your part and hand over to the next person. Now everyone is able to approach challenges with the same creative freedom. When I can show you what I mean rather than just describing it, it’s easier to build on. It’s easier to say “I don’t get it” or “what if we tried this instead?”

“What if” is the golden question. That’s when you know someone’s engaged. If we can break down those lanes so that everybody brings a “what if” question, we’re getting a diverse set of thinking from people who’d previously have waited their turn.

McKinsey’s research backs this up at scale: a human team of two to five people can already supervise 50 to 100 specialised AI agents running an end-to-end process like launching a product (McKinsey ). Marketing executives who describe their AI use as mature have already seen 22% efficiency gains (McKinsey ). Those aren’t projections. They’re numbers from teams already working this way.

But more capability needs to be matched with the judgement to use it well. You need to know what good looks like. Otherwise you have no way to judge whether the output is right. That comes from research, good briefs, and understanding the client.

We learned this the hard way. About a year ago, we were creating AI-assisted content for a client where the case studies were confidential. The constraints on what we could quote led to a stretch, a fictionalisation of evidence that went beyond the actual work. We made up case study examples that weren’t accurate. The person who created them didn’t spot the problem because they didn’t have enough context about the client to know. I flagged it. The client flagged it. We’re well past that now, but it stuck as a clear reminder: AI output without domain knowledge to check it against is a liability.

AI tools give your team more capability. But that capability only works when it’s paired with the skills to know when the output is right, when it needs polishing, and when it needs scrapping entirely.

What can B2B marketing teams actually create with AI?

Far more than content. Working prototypes, interactive demos, event activations, personalised sales materials, and internal tools. The resource constraints that used to sit between an idea and a finished product are disappearing.

But this isn’t a technology story. It’s a thinking story. The teams that get the most from AI do the research first, understand the problem they’re solving, and know what good looks like before they start building. AI needs polishing, not rescuing. And when every team has the same tools, the differentiator won’t be the output. It’ll be the quality of the thinking that shaped it.

Build when it’s your competitive advantage. Layer human expertise at every step. And ask the questions other people aren’t asking. That’s where the breakthroughs come from.

The best AI-powered marketing tools aren’t the ones you subscribe to. They’re the ones your team builds to solve a problem nobody else has thought to solve yet. For more on how AI is shaping the future of marketing, explore our guide to AI in marketing and our picks for the best AI marketing tools .

If any of this has prompted a “what if” question, I’d love to hear it. Drop me a line , and let’s explore the possibilities.

Frequently asked questions