Since 2001, we have watched four technology waves sweep through digital marketing: paid search, opened by GoTo.com in February 1998 and Google AdWords in October 2000; social networks, with Facebook opening to everyone on 26 September 2006; the smartphone, with the first iPhone on 29 June 2007; and generative AI, with ChatGPT on 30 November 2022. Each was announced as the end of agencies. Each shifted the job. From Luxembourg, where we went through all of them, the lesson has stayed the same: a technology is worth what it changes in the client's results, and someone has to remain accountable for that result.
The facts
- Paid search was born with keyword bidding on GoTo.com in February 1998 and went mainstream with the launch of Google AdWords on 24 October 2000. Grégory Camus ran his first pay-per-click campaigns on GoTo.com in 2000; he joined the Luxembourg market in 2003.
- Social networks became a mass channel when Facebook opened to anyone aged 13 and over on 26 September 2006, and the smartphone made mobile the first screen from 29 June 2007, the US release date of the first iPhone. In that same year, 2006, Xavier Lesueur co-founded BuzzParadise, an influencer marketing platform.
- Generative AI entered everyday use with the release of ChatGPT on 30 November 2022. We founded AIxH in 2023 in Differdange to work with that tool, under the same rule as in the previous waves: AI produces, humans steer, results are measured.
Four times, the same announcement: the end of intermediaries
In 2000, paid search was supposed to make agencies redundant: an advertiser could buy its own keywords in a few minutes. In 2006, social networks were supposed to let every brand speak directly to its audience, with no media and no agency. In 2007, the smartphone was supposed to simplify everything through apps. In 2022, generative AI was supposed to write, design and analyse in place of the teams.
Each time, the tool did go mainstream, and the promise of simplicity held for six months. Then complexity came back in another form: automated bidding and quality scores, reach algorithms, attribution across screens, and today models whose every output has to be checked. What disappears with each wave is repetitive execution. What remains is judgement: where to put the money, what to measure, when to stop.
What Luxembourg teaches that a large market does not
A market of 700,000 inhabitants, several working languages and a large share of cross-border workers impose a discipline that large markets can afford to ignore. There is no volume to hide a poorly targeted campaign. An ad served in a single language leaves out part of the residents; a campaign designed for Paris or Brussels ignores the daily commutes from Thionville, Arlon or Trier.
That constraint has an upside. Decision-makers are reachable, results can be read quickly, and a method that works here works elsewhere. The home insurance campaign we designed for AXA Luxembourg, adapted into 20 languages to cover 95% of the resident population, would not have been imagined from a monolingual market. Nor would the image tags for MONDORF Domaine Thermal, produced in three languages with human approval.
The lesson that has not moved since 2001
A technology is only worth what it changes in the client's results. We applied that sentence to paid search, to social networks, to mobile, and we apply it to generative AI. It has a practical consequence: someone has to remain accountable for the result, with a name and a phone number. A platform is not, and neither is a language model.
That is why we founded AIxH in 2023 with AI in daily production and a human on every decision. Our story since 2001 fits in one line: the tools changed four times, accountability for the result stayed in the same place.
AIxH's view
We support organisations in Luxembourg and the Greater Region that approach generative AI the way we approached the previous waves: real use, measured, with an identified owner. Our AI audit and integration service in Luxembourg starts with the tasks that already exist in the business, and our AI training sessions take place in person, in Differdange or on the client's premises. If your organisation is torn between waiting and starting, the question to ask is still the one from 2001: what does this tool change in your results, and who answers for it?
