There are two kinds of AI programmes in the wild right now.
The first is AI theatre. A flashy pilot, a steering committee, a few demos to the board, a press release. The organisation learns almost nothing. The systems don't change. The work keeps running the same way it always did, with a thin layer of AI sprinkled on top for credibility.
The second is AI-first adoption. Quieter. Less newsworthy. Usually runs for 18 to 36 months before anyone notices. But at the end of it, the organisation operates differently — its workflows, its decisions, its skills, and its platforms.
What AI-first actually costs
Real AI adoption costs three things at once: attention, architecture, and accountability. Attention, because leadership has to actually understand what they are buying. Architecture, because the systems have to support AI as a first-class citizen, not a bolt-on. Accountability, because somebody has to own whether the AI is actually producing value.
Most AI theatre skips all three. The work gets delegated to a team with a mandate but no authority, a budget but no architecture, and a scorecard but no real accountability.
The test
The simplest test is this: in a year, will this organisation operate differently because of this AI work? If the answer is "we'll have some pilots and some slides," that's theatre. If the answer is "these three processes will run differently, these two decisions will be made by different people, and these six skills will be in-house," that's real adoption.
Thakhutse's advisory work is built around the second answer.