By Alexander Concannon

For most startups, above-the-line marketing feels like playing with higher stakes. The reach is bigger, but so are the bills. Production, agency fees, voiceover talent, sound engineers, licensing, long lead times…it all adds up, and suddenly it feels all too hard and expensive. Especially if it doesn’t work.

But, we still wanted to give it a shot. 

Small businesses are doing it pretty tough right now, and Valiant’s message of connecting SMEs to the right funding has never been more relevant. To get that message beyond our usual performance channels, we needed a different approach. That’s what led us to radio, a channel with broad reach, but one we could only realistically tackle by rethinking the financial realities of how the creative got made. So we tried something new – our first radio campaign, built entirely in-house with AI.

The experiment

The idea started small. Could we repurpose existing content into something that felt more alive? Blog posts into podcasts, for instance. 

We played with a few solutions and landed on ElevenLabs, whose software was able to create voices that sit somewhere between eerily natural and slightly uncanny. Add ChatGPT into the mix for scriptwriting, and suddenly we weren’t just testing text-to-speech, we were producing actual ads.

The first attempts were rough. Scripts had the right words but no personality. Voices flattened sarcasm, and desperation was deadpan. When we tried switching between characters, it sounded like we’d clipped two disparate sound bites and mashed them together.

But with more context, things improved. We found we needed to write for AI. A longer setup before a line, or extra words after, gave the AI a rhythm. Adjusting prompts helped the voice land closer to our brand’s dry humour. Sometimes we just relied on luck and clicked ‘regenerate’ until it sounded right. Bit by bit, the spots started to feel broadcast-ready.

Why we tried it

The economics were impossible to ignore. Traditional radio production makes experimentation costly and slow. You brief an agency, wait weeks for scripts and voice casting, book the studio, get the edits, review the mix, and by the time it’s live, your original idea already feels old.

With AI, the cost dropped to a few hundred dollars in generative credits. The turnaround shrank from weeks to days. Most importantly, though, we’ve created a process that’s turned radio into a testable channel, not just a retrospective brand project.

What worked and what didn’t

• Worked: Quick iteration. We could generate three or four script versions in a morning, voice them and decide which one to expand upon by lunch.
• Worked: Lowering the barrier to experiment. Radio went from ‘too expensive to play with’ to ‘worth testing alongside digital’.
• Didn’t work: Sarcasm, subtle humour and emotional delivery were a bit hit and miss. Humans still win on craft.
• Didn’t work: Sound design. Sound effects were clunky. A keyboard click? Passable. A whistling kettle? Closer to bad sci-fi than breakfast radio. Backing tracks were non-existent.

 

• Worked for us, but maybe not for others: Approval process. There’s an element of randomness in the output. Using a small, in-house team meant we could tweak on the fly to deliver a better final product than the original concept might have been. If your process requires multistep approvals, you would struggle to create content quickly.

The lesson for us was that AI can’t replace the polish of a professional studio, but it can open the door to more frequent experiments. 

The bigger picture

For years, fast creative loops have been the domain of performance channels. You could A/B test a digital ad because the production cost was near zero. Radio, TV, out-of-home? Too expensive to iterate. You run a big campaign and hope it lands.

AI is shifting that balance. It doesn’t replace the ideas or the craft, but it makes production cheap and fast enough that you can apply the same rapid-test mindset to offline channels. Suddenly, a brand team can trial multiple scripts, rotate versions based on website traffic and feedback, and re-cut campaigns in near real time.

That’s the real change – in-house teams becoming faster, more experimental and more data-driven, not just online, but everywhere their brand shows up.

What’s next

We’ll keep working with human talent. There are campaigns where nuance and polish matter too much to compromise. But now there’s a new tier of possibility: the ability to test, learn, and refine at a fraction of the cost and time.

Right now, the tools are also absurdly cheap. In my opinion, AI is the cheapest it will be in the short- to mid-term. Eventually, commercial realities will catch up. But for now, with a few credits and a bit of editing, you can create a national radio campaign from your laptop.

For us, the value lies in creative freedom (and, of course, saving some money). We can act on an idea without asking for permission or a bigger budget. We throw out old work without worrying about sunk costs. And we can learn quickly what lands with listeners and what doesn’t.

That, to me, is the future of in-house marketing: not replacing imagination, but removing the friction between ideas and execution. AI is giving teams the ability to move faster, loop tighter, and keep improving with every cycle.

Alexander Concannon is the chief marketing officer at Valiant Finance.

Read more: Why your marketing strategies are missing a beat without audio