5 Years on from a ‘Global Media First’ what have I learned about AI?

5 years ago, during lock-down & home schooling, our CEO, Charlie Cadbury and his wife Maria Cadbury allowed a masked film crew into the kitchen to witness the annihilation of a massive pile of carrots whilst Charlie listened to, and engaged with, Alexa.

Say It Now, alongside an all star, multidisciplinary, multi agency team featuring EssenceMediacom, MullenLowe U.S. (TBWA\Worldwide) & AMA launched something that felt mildly bonkers at the time: an interactive smart-speaker radio ad that let people order Bayer‘s Berocca Boost just by speaking to Alexa during a standard ad break.

This was the first ever audio ad in the world that allowed the listener to buy a physical product using their voice.

The insight: Start with the channel people already trust and use every day, then make the next step frictionless.

Here’s the part that feels relevant again in 2026.

There is a lot of anxiety about AI taking a load of jobs in a very short time frame. I get it. But I am more optimistic, mainly because I think human adoption is the rate limiting factor. We have a lot more time to adapt than you’re being told by the endless AI hype.

The last five years, for us at least, have been a live experiment:

The capability curve shows a hockey-stick…but behaviour change is lumpy, slow, and social.

Even when something works, it still has to earn trust, fit habits, navigate organisations, budgets, regulation, and everything that exists between “possible” and “normal”.

That gap matters, because it means we have time to adapt. Time to redesign roles, retrain people, update incentives, and build systems that let humans do more of what humans are best at.

There’s never been a better time to build the future, you have time. Keep your head.

Time to reminisce with the OG team:

Omar Oakes‘s ‘Media First’ piece in Campaign: Link here
So much support and trust from the DAX team Faye Trinquart, Julia Connaughton CMgr FCMI Global Case Study: Link here
Say It Now steady nerves from Sander Siezen Case Study: Link here