We have a client who took what turned out to be a far-reaching decision about 20 years ago.

They’d grown big by doing massive direct mail campaigns, making use of new digital methods which enabled personalised content. In the Netherlands, everyone knows their name. And they’ve spread their wings wide, with sister companies throughout Europe.

But back then, the archiving methods for those successful campaigns reminded of a slower time: print proofs kept in physical portfolios. Cumbersome, and most importantly, difficult to share.

I was working at the company at the time, and I’d proposed and built other digital products for use within its walls. Among them their intranet, a mix of facebook-before-the-fact and a news site.

And so I was introduced to their ideas early on to digitise the cumbersome portfolios.

The object was to include all their marketing channels: direct mail, billboards, commercials on tv and radio. And yes, also their fledgling email campaigns.

When I left a couple of years later to start what would become Django Web Studio, this was one of the projects I took along.

Fast forward to 2025, the digital archive is a responsive web application in React, with highly automated content management in our Django backend. More than twenty thousand campaigns, more than a hundred thousand unique media items.

For our client, it’s become a valuable asset for inspiration and guidance. Because new campaigns are built on creative variations of highly scoring campaigns of the past, it makes the digital marketing archive one of the linchpins of their ongoing success and growth.

But.

As a company we're always looking to help. We saw a problem coming, came with a solution.

The problem, of course, the sheer amount of data. As mentioned, thousands upon thousands of campaigns. There's a search engine, what if we could do better?

A chatbot could help marketers understand the vast content of the archive better, even help users create new campaigns. It would be AI driven, obviously. We’d RAG index the entire content and use an AI chatbot to source the index when questioned.

But because we know that a proposal is just another piece of paper on a decision-maker’s desk, we went ahead and built a demo. Good enough to show off its capabilities, looking good enough to fit in the existing design.

We presented to client, who was amazed and enthusiastic. And was particularly surprised by the marketing campaigns the chatbot could outline as response to a short question.

Next step is to include the chatbot into the production environment and get some real user feedback.

And the step after could be to take on the task of indexing media: formidable, given the number of hi-res images and video. But the returns would also be formidable: users could enter an image and a text and get back a campaign mockup.

Now wouldn’t that be cool.