What 10,000 Trained Professionals Taught Us About AI Adoption

Later this year we will publish The State of AI Adoption in Thai Enterprises — a benchmark report built on measured data rather than vendor surveys: aggregated results from our AI-readiness assessments, capability assessments, and the training programs behind 10,000+ professionals trained and 3,450+ co-built AI assistants.

The survey side of that report is collecting now (more on how to contribute below). But some patterns do not need to wait for the charts — we see them in almost every engagement. Here are the four we would bet on before the data lands.

1. The gap is almost never the tools — it is integration

The most common shape we find when we assess an organisation: foundations score high, workflow integration scores low. People understand what AI is and can use it in isolation — but it has not entered the shared, repeatable processes where work actually happens.

In one engagement with a ~400-person professional services firm, the literacy assessment showed exactly this: high foundations, low integration. The fix was not more general training — it was co-building one owned workflow (report drafting).

^◦By the numbers≈60%drop in average first-draft time over one quarter, after co-building one owned workflow

^◦RelatedRead the full case study

Individual skill is necessary but not sufficient. Adoption lives or dies at the level of the team process, not the individual prompt.