A running argument about cross-border tax practice, AI, and the architecture of professional judgment.
11 Essays so far
Ten essays building toward the case above.
The IRS's own AI policy draws a line most readers walk past: a machine that writes gets a person on every output, while a machine that chooses gets an accountable official for the system as a whole. The taxpayer at the other end usually meets only the second kind.
13 min read Read
Every proposal to tax AI is reaching for a worker who is no longer in the room. The structure that generated the tax also generated the profession, and only one part of it survives.
12 min read Read
The industry believes that more information, and better information, is the path to a better answer. The harder discipline is the opposite one: knowing which few of a thousand sources bear on the question, and having the judgment to leave the rest out.
11 min read Read
Every vendor now calls its AI native, so the label no longer carries information. The distinction underneath survives it: which assumptions a system carries forward, and which professional questions it can honestly answer.
14 min read Read
The EU's draft guidance on AI transparency had to answer a narrow question: when is a human's hand in AI output real enough that the output needs no warning label. From a motive that has nothing to do with professional ethics, the regulator drew the same line this series has spent five essays drawing — that a signature is worth something only when it carries substantive ownership, not a rubber stamp.
10 min read Read
Bain and BCG agree: AI pays off only when the people accountable for the outcome own the redesign, not when a technology team installs a tool. In a regulated profession that prescription turns stricter — the person who will sign the work has to own how the system reasons, because the architecture of judgment cannot be delegated to anyone who will not be in the room when it is tested.
9 min read Read
McKinsey's case for the AI assembly line is right about almost everything. But the canonical line diverges toward autonomous output, while a tax opinion converges toward a single position a named professional must own. The last station, the signature, can never be automated.
13 min read Read
What one client folder actually does to ktAi, and what ktAi does back. The memory the system builds is the actual product — the chat, the compliance reports, the opinion drafts are outputs of that memory.
9 min read Read
Mid-market multinationals cannot economically reach partner-grade cross-border opinion work — and the engagement model itself fragments the company's accumulated context across firms and files. This is the structural gap, and the wedge ktAi is built for.
14 min read Read
The debate over whether AI will replace lawyers and accountants has been stuck in the same place for sixteen months. Here's what both sides are missing, and what it means for the practitioners who'll actually have to navigate the shift.
9 min read Read
ktAi is open to inquiries from tax attorneys, advisors, accountants, tax directors, in-house teams, and firms thinking about AI for their practice.
