Everything else an organization publishes about itself is written to impress someone.
The website is positioning. The annual report is a performance for people who have already decided. The press release exists because someone needed a press release. All of it is drafted, reviewed, softened, and signed off by people whose job is to make the organization look like it has things under control.
Then there is the job ad, which has a different constraint: it has to attract someone who will find out.
Why the constraint changes the writing
A job ad that oversells the role gets a hire who leaves in four months. Everyone who has written one knows this, so the writing pulls in an unusual direction — towards accuracy, because inaccuracy is expensive and the bill arrives quickly.
So the ad describes the work. Not the mission, not the values paragraph nobody reads, but the actual day: what arrives, how much of it, which systems it moves between, what the person will spend the morning doing. It names tools by name because a candidate needs to know. It says how much volume there is because someone who cannot handle the volume must not apply.
And it is dated, public, and linkable. It is a primary source about how an organization’s work actually runs, published by the organization, on purpose, with a reason to be truthful.
What it also proves
An ad is a budget that has already been approved.
Nobody advertises a role they cannot pay for. Somewhere behind that posting, a person with authority agreed that this work is worth a salary, indefinitely — which means the problem is not hypothetical, it has an owner, and the owner has already lost the argument about whether it matters.
That is a stronger signal than almost anything a survey would produce. Ask an organization whether it has an operations problem and the answer is a considered opinion. Read that it is hiring a third person to route inquiries between two systems, and the opinion is unnecessary.
What it does not tell you
Being honest about the method matters more than being impressive about it, so: an ad is evidence of the work, and it is not evidence of the cause.
The same posting can mean the volume is growing, or that the last person left, or that a process quietly broke eighteen months ago and nobody has had time to look. The text cannot distinguish them, and neither can anything reading it. The disciplined version is to treat the posting as a reason to ask, never as a conclusion to state — a rule this project learned by getting it wrong first, which the next article is about.
The part that should make this less strange, not more
Reading job ads to understand an organization sounds like surveillance until you notice where the information came from: they published it. Deliberately, publicly, to be read by strangers, with an invitation to respond attached.
The alternative — arriving with a generic message about digital transformation, addressed to a role rather than a person, referencing nothing — is the one that should feel worse. It requires the recipient to do all the work of figuring out whether it could possibly apply to them.
A message that quotes the organization’s own sentence back and says this specific work is the kind a system does now has done that work already. It can be judged in four seconds and dismissed in one, which is a courtesy, not a trick. And the recipient can see exactly how they were found, because the answer is a link to something they wrote.
Totomoko re-architects how organizations work so that AI does the job — inside the systems they already run, on an architecture built for portability: model routing and a data layer the organization owns. Built that way, swapping a vendor is a configuration change rather than a migration.