The first line on this website says the world has never had more information and never less clarity. I want to be specific about why that is getting worse, because I spend part of every week building the tools that are making it worse, and I think the honest version of the argument is more useful than the alarmed one.
A competent market report used to cost weeks of an analyst's time. It now costs a prompt. In an afternoon I can produce a sector overview of Brazilian agtech, a stakeholder map for a city, a regulatory calendar and a ranked list of partners, all of it fluent, formatted, plausible. I do this. We use these tools in Ocidente's own work and I would not give them up.
Something happens when the cost of an answer falls to near zero, though, and it is not that everyone becomes well informed. Every question gets covered in fluent text, and fluency stops telling you anything about whether anyone checked. Five years ago a confident twenty-page document implied twenty pages of somebody's effort. That inference is dead. The document may have been produced by a system that has never spoken to a human being in the market it describes, and it reads exactly like one that was.
So the scarce thing is no longer the report. It is what the report cannot carry.
Provenance, first. Where did this claim come from and when. In my own work every statement in a strategy document carries a grade: confirmed fact, estimate, working assumption, hypothesis, recommendation. It is a clumsy habit and it slows the writing down. It also means that when a plan goes to a founding team they can see which parts rest on a document and which parts rest on my judgement, and they can argue with the right part. A generated report has no grades. Everything in it arrives at the same confident temperature.
Freshness, in the places that move. Last year I evaluated venues in four cities for a client, and two of the four had changed operator within the previous twelve months. No database had it. The people who ran the buildings had it. A system that reads the internet will give you the old operator's name with total assurance, because that is what the internet says. The facts that decide outcomes, who is in the chair, who has stopped returning calls, which regulation is about to be reinterpreted, change faster than anything gets indexed, and they change through people.
And the reason behind the fact. A report can tell you a company left a market. It will rarely tell you they left because one particular partner made their life impossible, that the partner is still there, and that this is the person your client is about to approach. Somebody who was in the room knows that, or knows someone who was.
In practice we let the machines clear the floor. Anything that can be assembled from public, documented sources gets assembled fast, so no human hour at Ocidente is spent compiling what already exists. Every output carries its provenance, so nothing generated can pass as verified. Then we spend the human hours on the part the tools cannot reach: the calls, the visits, checking the three facts that would change the recommendation if they turned out wrong. The machine makes the first draft almost free. That is exactly why the verified second draft is now the whole product.
What the tools cannot do, and I say this as someone who builds them, is be told things. Nobody confides in a system. People confide in a person who has shown over years that the confidence is safe. Carlos has written about that better than I can. My addition is only that the technology has not reduced the value of that person. It has stripped away everything that used to be mistaken for them.