Data captures the output, not the work (Semafor)
Silicon Valley says AI will remake business within a few years. One of the people building the models says to think in decades.
Ioannis Antonoglou, co-founder and CTO of Reflection AI, told Semafor this week that the AI era will force every business to redefine what it actually does, then rebuild that with AI as the scaffolding. His own estimate, quoted in the interview, is "maybe a decade or two decades" before companies genuinely rethink how work gets done. He sells the open-weight models he hopes will do the rebuilding, so read the timeline as his view, not a neutral forecast.
The obstacle he points to is not model capability. It is data, and I think he has put his finger on the right gap.
Companies capture what is measurable: transactions, documents, finished work product. What they almost never capture is the work itself. The call where the decision was really made. The reason a senior person rang one client before another. The judgement that links a hundred small actions into an outcome. Antonoglou's own words: much of the context is "we jump on a call, or someone shared something or said something".
Ask anyone who has run a transformation program and they will recognise that immediately. The process map describes the workflow. It rarely describes how the organisation actually decides, and the people who carry that knowledge seldom think of it as knowledge.
That reframes most AI programs. They start with the data warehouse and ask what can be automated. The more valuable question is what the organisation knows but has never written down, and who holds it.
The strongest counter-argument is that models may not need our tacit knowledge at all. If capability keeps compounding, a machine could learn the job from outcomes alone and the missing context would stop mattering. The evidence is thin either way, and the people arguing context is the moat are also the ones selling the models. I am not settled on it. But even if the models get there, the organisations that can describe their own judgement will adopt them faster than those that cannot.
His prediction of uneven change follows from the same logic. Industries will not move at the same pace, which Semafor notes gives faster entrants room to spring ahead of incumbents still mapping their own processes.
Before buying another platform, it is worth asking a plainer question. Could you describe how the most important decisions in your business actually get made?
Source: Semafor Tech, "Rethinking everything", First Word on Reflection AI CTO Ioannis Antonoglou (14 October 2026), https://lnkd.in/gtSFAsYP
#Transformation #KnowledgeManagement #DataStrategy