The State of AI Use in Agencies

Interest in, and experimentation with, AI text generation systems has exploded in the last 2 years. The catalyst for this was undoubtedly OpenAI adding GPT-3 to their open beta. GPT-3 is 117 times as powerful as its predecessor*. This quantum leap in capability combined with an easy-access API created a gold-rush of interest in AI content generation.

Keen to explore the uses and limitations of these systems, I have undertaken 3 pieces of new research relating to AI text content generation:

Agency leaders survey

A survey of agency leaders regarding their adoption of these systems, and future expectations for them (n 64).

End user attitudes survey

A survey of 853 end users regarding their attitudes to different types of AI content.

Randomised blind trial

A blind split test of 289 users testing the difference in perception between human and non-human content.

* based on parameter count.

Why the report is no longer available to download

This research was carried out in 2022, before ChatGPT existed. A lot changes in a few years in the world of agencies, and far more in the world of AI.

I have taken the download down because people were still finding it and reading the numbers as current, without noticing how old it is. The data does not describe how agencies use AI today, and I would rather remove it than leave something misleading in circulation.

The report itself is still perfectly interesting as a snapshot of where agencies were before the tools most of us now use existed. If you would like a copy on that basis, ask me for one and I will send it over.