Live with Azeem: AI & ROI
Exponential View Azeem Azhar ● Covered by 3 sources
Azeem Azhar went live to unpack why AI coding tools boost individual output but barely move company-wide productivity. A tech exec told him her team's 'one plus one plus one equals one-and-a-half.'
Based on reporting by Exponential View, Azeem Azhar — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Azeem Azhar sat down for a live conversation this week to dig into a puzzle that's been nagging at plenty of executives: if AI tools are so good, why isn't the bottom line moving? The session followed up on an essay he co-wrote with Nathan Warren days earlier, and it leaned on a specific, telling anecdote he picked up over tea with a senior exec at a well-known public tech company.
That exec runs roughly a thousand engineers. Nearly all of them use Claude Code day in, day out. By every individual measure, they're crushing it — more lines shipped, more pull requests submitted, tasks closed faster than before. And yet, when she zooms out to the organizational level, the gains just don't stack up the way you'd expect. Her own phrase for it stuck with Azhar enough to build a whole discussion around it: one plus one plus one plus one equals one-and-a-half.
That's the crux of the ROI mystery. Individual productivity metrics look fantastic in isolation, but something in the machinery of how large engineering orgs actually function — code review, integration, coordination, decision-making — seems to eat up most of the extra output before it ever reaches a P&L statement. It's not that Claude Code or tools like it are failing. It's that speeding up one person's typing doesn't necessarily speed up the fifty other steps a piece of software has to pass through before it creates value.
Azhar didn't pretend to have a tidy answer in the live session; the point was to surface the tension and get people arguing about it in real time. But the anecdote itself does a lot of work. It's a sharper, more human way of saying what a dozen dry productivity studies have hinted at over the past year: individual-level AI gains are real and measurable, yet they keep failing to translate into firm-level growth, at least not yet, and not automatically.
My take — AI-written commentary, not fact-checked reporting
This tracks with something I've suspected for a while — we're bolting rocket engines onto bicycles and wondering why the whole convoy doesn't move faster. The bottleneck was never typing speed; it's organizational plumbing, and no LLM fixes a broken review process or a company that still ships code the same clunky way it did in 2015. Until firms redesign the workflow around the tool instead of just dropping the tool into the old workflow, expect a lot more 'one-and-a-half' math.
Read more about this at: Exponential View
