To see that the systems he trusts have blind spots and interests of their own, his included, and to put his own interest in writing.
Stands to lose
His uncle's trust and, later, his standing at his firm, where part of his enthusiasm is a nephew's.
Strength
Patient, thorough and truthful. He asks the one question that lands, and he does not press the point when the records prove him right.
Flaw
He advises everything at once, and takes a platform for the end of middlemen.
The question their story asksCan the nephew who believes in the system make his uncle examine himself without losing him?
Herald. He carries the machine into Rajan's office and keeps bringing the question Rajan cannot answer.
The arc
FromA graduate with a laptop and a cost spreadsheet.
→
ToA venture analyst who writes down his own conflict of interest and tells his uncle what the model had already found.
Karthik starts as the bringer of software and hard questions, and his records correct Rajan more than once. His own confidence is checked too: the all-at-once rollout he advised silences the floor, and the platform he cites turns out to be a new middle. By 2026 he is the go-between who discloses in writing that the adviser is his uncle.
part of the arc is hidden until you have read Chapter 27.
On a Saturday in November he shows Rajan three systems and a spreadsheet comparing the cost of the software with the cost of the people. He waits for more than thanks. It does not come.
He puts a Copenhagen buyer's questions to an AI trade assistant and reads the answers aloud. When Rajan dismisses a mill over a bad lot, he asks whether Rajan has tried it since.
Early 2025. He runs ten years of Rajan's records through a supplier-scoring tool, which puts Senthil's mill at the top. He asks for a checklist of Rajan's instincts and is refused.
He is pleased when Rajan orders every system at once. In July 2025 he leaves for a venture firm in Singapore and joins the Friday meeting by video for a few weeks.
September 2026. He draws Rajan into his fund's decision, writes a note to the lead partner stating that the adviser is his uncle, and tells Rajan the screening model had already flagged two of his three objections.
later steps are hidden. They appear as you read on.
Respectful, persistent and kind. He phones every Sunday.
He challenges Rajan with records and questions, advises a rollout that goes wrong, and in 2026 handles him patiently into a room where both of them have to declare an interest.
What they carry
Karthik speaks for the record and the system. The book gives him its strongest case for the machine and for writing knowledge down, and the question that tests whether Rajan's instinct reads a man or only where he comes from. It also marks his limits: a rollout advised all at once, and a platform taken for the end of the middle.
Part IShows Rajan three systems and then a spreadsheet: the cost of the software beside the cost of the people who do the same work.
Chapter 4: Why This Matters NowShows the systems and the cost spreadsheet. Leaves Rajan a login for the supplier-checking module. On the Sunday call asks what Rajan saw in Erode, and whether he would have seen the same if he had known the man's father.
Chapter 5: When Intelligence Became CheapPuts the Copenhagen buyer's questions to the assistant and reads the answers aloud. When Rajan dismisses the Karnataka mill, asks whether he has tried it since.
Chapter 7: What AI Cannot ReplaceRuns ten years of records through a supplier-scoring tool. Asks Rajan whether he would have seen the same thing in the Erode owner if he had known his father. Asks for the checklist; does not answer when refused.
Chapter 18: The AI-Native OrganisationIn November 2024 shows Rajan the software and takes the laptop home. Sends articles. Is pleased when Rajan orders everything. Joins the Friday meeting by video for the first few weeks, then only his Sunday calls.
Part VArgues that machines will cut out every middleman.
Chapter 22: Own the Layer AI Cannot DisintermediateAsks for a video call to show a chart. Explains disintermediation and cites Weiss's platform. Describes his firm's screening model and the Jakarta decision, and does not pretend to know who was right.
Chapter 27: The Investor’s PlaybookAsks Rajan to read the deck. Answers each objection with 'come anyway'. Writes a note to the lead partner, copied to Rajan, stating that the adviser is his uncle. Tells Rajan the screening model had already flagged two of his three problems. Sits beside him in the meeting and does not point out that the records judged Senthil better than Rajan did.