Sandeep Krishnamurthy on recombinant AI fluency

Near the end of the hour, Niki tried to get Sandeep Krishnamurthy to name the one AI tool every student should learn first. He wouldn’t do it.

“I refuse to answer the question. It’s not one. I just said AI is not a tool. Stop thinking like this — it’s going to take you down the wrong path.”

That refusal is the entire argument compressed into a rapid-fire answer. Sandeep runs the College of Business Administration at Cal Poly Pomona, one of the largest business schools in the country at more than 5,000 students, and before that spent two decades in Seattle forging industry partnerships at the University of Washington and UW Bothell. For the past year he’s been writing about a concept he coined — recombinant AI fluency — and the fifth and final Provn Talent Scout webinar was an hour spent on what it is and how you prove you have it.

AI is a presence, not a tool

Sandeep’s starting point is that AI has been miscategorized. Parents and business leaders keep asking him whether this is just another wave like the browser or the smartphone — another item for the stack. He thinks that framing is the trap, because the moment you call AI a tool, you slip into a training mindset: drill the procedures of one system and assume you’ve made it to the other side.

“AI is best understood as a presence rather than a tool. It’s like having Einstein in your living room. It’s up to you how you extract information, how you engage him. Einstein might get bored and leave your house, or you can get amazing information from him.”

Intelligence is plural now. In his framing there is no such thing as becoming an expert in ChatGPT, or Gemini, or any single product — the people creating real value are the ones who stop trying to master one system and start orchestrating several.

“There’s no such thing as becoming an expert in ChatGPT or Gemini. We encourage students to be problem-oriented — to get the strengths of different AI systems to work together to make something real happen.”

That is recombinant fluency: weaving outputs from multiple systems — ChatGPT for one step, NotebookLM for another, Gemini or Elicit for another — into a single workflow pointed at a problem. He notes that most students don’t even know NotebookLM exists, because the hype and their school’s site license funnel them toward one product. The move he wants is from an analytical mindset to an integrative one: stop drilling the analysis of one small thing, and learn to integrate across many forms of intelligence.

From certifications to production

For years the answer to “how do I stand out” was a certification — SAP, supply chain, pick a field. Sandeep thinks that era is closing.

“We’re in the middle of a profound shift from certifications to production. We want students to produce something. Can you build an agent? Go figure it out.”

It doesn’t have to be impressive. Build an agent that sends one joke a day to your network. The artifact isn’t the point — the demonstrated understanding is, and so is the willingness to make the thing instead of reading about it. He’s candid that this is the harder road, and that it takes initiative most students don’t show. When they do, he moves.

“If a student takes two steps, we’ll take 98. We’ll go to them and help. But they have to take the first two steps.”

This is also where his philosophy and the Provn model converge. A challenge you actually solve, paired with a record of how you thought about it, is production rather than certification — and it’s the thing he keeps telling students to go build.

The signal isn’t on the resume

Production matters so much precisely because the resume has stopped carrying information. A company posts an ad for an entry-level analyst and gets tens of thousands of AI-written applications; an ATS running on AI then reads cover letters written by AI. Nobody can break the tie.

“They get 50,000 applications all written using AI. They don’t know how to break the tie.”

So the signal has moved — to artifacts, to events, to the room. He told two stories. In the first, his school ran an AI hackathon with 350 students; one team was handed a database of 10 million phishing emails and asked to build something useful and present it to cybersecurity leaders. The company that set the challenge came back and said it wanted to hire them. In the second, a student in a high-visibility role simply talked with an industry visitor — explained her work, asked questions, listened — and the visitor said “you’re hired” on the spot, because he’d just come out of 25 interviews where everyone was socially awkward.

“We’re focused on the AI effects, but personality still matters. The way you speak, the way you connect, the way you go through situations in a fearless way.”

The phrase he and Niki kept circling was that AI is sending hiring back to the future. When knowledge is a prompt away, the differentiators are the old ones — hustle, communication, initiative, whether someone actually wants to work with you. A generation that lost some of the art of small talk during COVID has to rebuild it, because a business leader will hand you a nugget out loud that they’d never put in writing. And asked to choose between two otherwise identical candidates — one of whom submitted a portfolio of videos and things they’d built — his answer was immediate.

“The second person. There’s a series of artifacts I can look at — videos, code, whatever. It signals effort at some level of quality, and a problem-solving mindset.”

Islands of green in a sea of red

A surprising thread ran through the hour: students are chasing stability. Sandeep met a 4.0 student who took time off because he was so confused about what to do, and landed on a single criterion — stability. Business students are now applying to government accounting jobs. One told him he was joining the army; another, the LAPD (“125K plus benefits,” and Sandeep couldn’t argue).

“People are looking for islands of green in a sea of red — small, protected walled gardens where they can pitch a tent and thrive.”

In tech specifically he steers students toward two durable paths: cybersecurity, where the perimeter fight is about to get enormous, and DevOps, because someone has to integrate and maintain all the code that interns and agents are now generating. More broadly he frames the opportunity as “AI plus” — AI plus marketing, AI plus accounting, AI plus finance — contextual intelligence rather than AI in the abstract. He also thinks the romance with cheap AI labor is about to meet a spreadsheet. Where CTOs once raved about interns who did six months of work in a week, they now watch those same interns burn through tokens.

“Now they tell me: you can’t believe how fast he blows through tokens. I could have paid a developer a full salary for what this kid used in two weeks.”

His bet is that humans make a comeback — not as rubber-stampers in the loop, but because companies are about to reckon with the true cost of having let go of the people who understood how the pipes connect.

The rapid fire

The hour ended where this post began, with a round of fill-in-the-blanks.